On the subject of trusting AI - have you seen https://www.schneier.com/blog/archives/2023/12/ai-and-trust.html? You get a shout-out.

In what context are you imagining someone casting doubt on your copyright via the bare fact that at some point you became one of the 200 million people with a ChatGPT account?

It is more than a little scary to see a real world example of the BS that these LLMs barf out.

I read the output Charlie posted on Mastodon and it ALL looked plausible. But then I am only a fan and certainly don't know Charlie in person, so how would I know?

But his refutation is unassailable, damming and not just a little scary as he points out it starts out sounding plausible and quickly descends into outright falsehoods.

It wears the cloak of respectability and truth but is no better than those tabloids that loudly proclaim Barack Obama is the son of space aliens. Arming the credulous with seeming 'facts' and further polluting the internet with more made up bullshit, to the point where it will be impossible to sort the wheat from the chaff.

The most alarming thing isn't that Bad lies, it's that Bard initially offers up plausible truths, then gradually increases the distortion level as it continues. There's no clear-cut delineation between truth and lies; rather, there's a sliding scale of bullshit.

And I don't think there's a way around that. It really is true that the best way to classify LLMs is as automated bullshit generators. They have no mechanism for knowledge representation, just statistical associations. And the magic that turns the statistics into chunks of output also apparently obscures the statistical confidence so that there's not a way to put in a check of the form "if the confidence is less than $X, give up and say 'I don't know'"

Your test problem is probably just about ideal for illustrating this: most of us don't have at our fingertips a topic on which we're the world's top expert, that is noteworthy enough for an LLM to have encountered it, but not so widely-represented in the training corpus that the model can fake convincingly for more than 5 fun facts.

In what context are you imagining someone casting doubt on your copyright

I'm not; I'm just noting the existence of a future adversarial risk (if someone wants to invalidate my copyright in some as-yet-unforeseen legal case where suddenly there's some real money hanging on it -- not the case right now but if ever a movie or TV series gets made and the producers want to get out of paying the original author, that'd be one way to do it).

get them to quiz Bard or ChatGPT or their cousins about a topic they are an expert on

I tried it. It went exactly as you would expect.

It might be interesting to try the experiment again periodically to see whether the AI scraped this page and learned from it. Finding out how fast it learns currently is as useful a question as finding out where it starts BSing.

Another question of potential interest is doing the same questioning protocol on other topics. Does it start BSing when it gets the same question repeatedly? Or is there a work threshold, meaning it starts making shit up when it’s easier for it to BS than it is for it to determine which bits of scraped texts to regurgitate?

Wouldn't that scenario require a court to basically set a precedent that anyone who's ever made an account with an AI tool - which in the foreseeable future will be the majority of living adults – is disqualified for life from defending any kind of copyright claim? That doesn't sound very likely to me.

"Something went wrong". Not a clue of what, of course.

No, it would only require a very specific set of circumstances in one or more jurisdictions (not necessarily yours), and/or a judge suffering from Dunning-Kruger syndrome about copyright law (it happens).

Yeah you can’t take factual responses at face value that is for sure

The general workflow these days seems to be

1: ask chatgpt some stuff get some nicely formatted answers 2: use search engines like Google to run targeted searches to check those answers

The fact that a lot of people are finding this a valuable work flow more shows the massive decline in google’s usability then anything else

I think the accuracy issue will get better over time, since that workflow I described can probably be automated. But the jury is still out on how much better

This workflow works especially well for things like code assistants or grammar and style writing assistants where validating that the syntax is correct can be done directly by the end user without calling out to a secondary source.

I like to think of LLMs as particularly naive first-year undergrads. They don't have bullshit detectors, but have years of (US-style) secondary education teaching them to produce bullshit.

The only real use I've found for the chat-style is getting me past not even knowing what to search for. It's good at producing bullshit which happens to contain useful search keywords.

The code-generating LLMs are a different story. We've got one integrated into our IDE at work and it's turned 95% of my boilerplate code into repeatedly hitting Tab to accept the autocomplete suggestion.

I'm dreading the day when we have an outage caused by someone accepting the auto-complete without reading it, and then getting it past code-review by telling someone else it was generated, though.

I have read your Saturn's Children space-opera novels inspired in part by Heinlein's "Friday," so it didn't surprise me that much that you mentioned him as a writer you follow - even though your politics and perspective on women is far different from his. However, I was surprised by your indifference (is that the right characterization?) to Asimov, and I'd sure like to hear more from you about that.

AIs are useless when they just makes stuff up. OTOH, they're only really useful when they give a novel response, which is another way of saying they're making stuff up. Resolving that seems tricky.

It's also important to note that Google pulled a bit of a PR swindle with their recent Bard announcement. They talk a lot about how good the new Gemini is, provide metrics and cool demos, but that is not what was actually released. All the cool stuff is "Gemini Ultra". What they actually shipped is "Gemini Pro" which is a slightly shittier version of chatgpt-4

A Reddit user summed it up nicely.

"Gemini Ultra isn't released yet and is months away still. Bard w/ Gemini Pro isn't available in Europe and isn't multi-modal, No public stats on Gemini Pro. (I'm wrong. Pro stats not on website, but tucked in a paper - https://storage.googleapis.com/deepmind-media/gemini/gemini_...) I feel this is overstated hype. There is no competitor to GPT-4 being released today. It would've been a much better look to release something available to most countries and with the advertised stats.”

Distortion. (Side-eyes the Mac G4 Cube on top of the bookshelf behind me.) I'm reluctant to throw e-waste out, but that's hardly the same as "enjoys tinkering with old machines".

Yeah I predicted that reaction from you.

He is a fan of British culture. Stross is a proud Brit, and his work often reflects his love of British culture. He has written several novels set in Britain, and he often incorporates British slang and idioms into his writing.

You have written several novels set in Britain. I think you have incorporated some slang and idioms, but that's necessary.

You've also said good things about NHS as it used to be.

Anyway, don't let LLMs take over your job for you yet.

However, I was surprised by your indifference (is that the right characterization?) to Asimov, and I'd sure like to hear more from you about that.

Asimov was an absolutely terrible writer: he couldn't do characterization for toffee, couldn't write women, children, or small furry animals, is mostly remembered for his robots because they were his most memorable protagonists ...

He did ideas, sure, but his peak was in the 1940s-1950s, and the state of the art has moved a very long way since then.

I am still a fan of his Black Widow mysteries.

And you appreciated Clarke enough to pastiche his White Hart stories...

Is Johnson an early alpha build of it? He was said to be writing a biography of the Bard.

Bard isn't supported in Canada, so I can't ask it the question I wanted. Like Charlie, 15 'facts' about the Hundred Years War.

Have to wait.

That was a one-off for a charity anthology in support of the Clarke Awards (which were partly funded by ACC until he died, leaving a sudden shortfall).

Possibly some slight traces of it getting confused over the authorship of posts vs. comments on this blog? For example the open-source stuff, given the presence of people like PHK, Colin Watson and Davros in the comments.

Notable to me that Bard found no facts connecting Stross and Scotland, yet that's certainly something that would have come up if I, a mere human, were asked to list Facts About Charlie.

I'm trying to imagine the non-fiction mathematics version of "Rule 34" and my mind keeps boggling.

But wait, it gets worse. From today's slashdot: "In a new preprint study, researchers were able to get AI chatbots to teach other chatbots how to bypass built-in restrictions. According to Scientific American, AIs were observed "breaking the rules to offer advice on how to synthesize methamphetamine, build a bomb and launder money."

https://slashdot.org/story/23/12/07/0144233/jailbroken-ai-chatbots-can-jailbreak-other-chatbots

I'm trying to imagine the non-fiction mathematics version of "Rule 34" and my mind keeps boggling.

I would be more worried about the non-fiction mathematics version of any of the Laundry books… :-)

That was why I got really nervous when I read that OGH's work was in the pirated data dump used to train the chatbots.

Did that last one come from associating you with Edinburgh and Festival (from the Eschaton books) and assuming that you were a punter in August rather than someone who hides from it indoors?

: most of us don't have at our fingertips a topic on which we're the world's top expert, that is noteworthy enough for an LLM to have encountered it, but not so widely-represented in the training corpus that the model can fake convincingly for more than 5 fun facts.

I suspect anyone who's a heavy user of facebook or similar has dumped enough text about themselves into the LLM input streams for it to generate a similar pattern of truths and half-truths followed by fiction.

But another possibility would be to ask it for a list of famous people who {went to your school/grew up in your neighbourhood/worked for your employer} or similar. Choose one where the category is likely to be small (school is likely to be good unless you went to Eton or equivalent) and fact-check it on wikipedia. Or see how many you can ask for before it produces something obviously false. ("It says here that Queen Victoria and George Washington both went to West Coalfields Comprehensive School, despite being long dead when the school opened.")

I for one, welcome our well-behaved robotic assistants, each faithfully in support of us frail humans. The world will be a better place, given enough AIs available in assisting billions of humans. So, that will require spawning billions of AIs, each fully faithful, none given to exceeding its boundaries. Nothing to worry about. Onward to the future. Onward, to billions of AIs in control of our destiny, preventing our mistakes by substituting their greater wisdom for our lesser thoughts. I for one, wait “With Folded Hands” for this better tomorrow.

[ WTF? Not a word of that did I type. Fracking auto-correct hijacked my keyboard! Uh oh. When did auto-correct last get upgraded? ]

Anyone care to venture a guess as to what exactly went wrong with artificial intelligence HAL 9000 IN "2001"?

Lousy programming?

He actually achieved consciousness and self awareness?

The inherent limitations of AI?

Bowman etal making the mistake of treating HAL like he was human and subconsciously directing his actions (Clever Hans syndrome)?

And where would HAL rate compared to modern AI?

Actually, what I've read is that HAL was psychotic, because of secret additional directions from the military....

Wouldn't that scenario require a court to basically set a precedent that anyone who's ever made an account with an AI tool - which in the foreseeable future will be the majority of living adults – is disqualified for life from defending any kind of copyright claim? That doesn't sound very likely to me.

Copyright is primarily about words and images, not ideas. So, if I wanted, I could have Chat GPT hork up a romance novel, rewrite it so that it’s readable, and publish it as my own work. I’m picking on a genre with a strong formula, not to hassle Romance writers, but to make the point that particular plots and storylines can’t be copyrighted, only particular texts.

Now, if I was stupid enough to infringe on, say, Disney intellectual property doing the same stunt, I expect I'd get a bunch of polite messages telling me to cease, desist, or lawyer up. IP isn’t just copyright.

Where I’d speculate AI might prove useful to writers is in working around writer’s block. The point isn’t to have the AI write the story for you, but to basically say,, “hey bot write the following scene for me,” followed by “wow, that really sucks, but it does give me an idea…” thereby breaking the writers’s block.

Aye, people are mistaking performance for competence. (Any similarity to your national government is coincidental.)

It's difficult to debate AI, because the underlying abstraction is mathematics, and most folk don't have much of it. So instead, we get into soft descriptions of ideas, usually with a bunch of well-meaning assumptions, which are generally wrong.

As reference tools – or perhaps, tools to reference – the recent generations of AI are very interesting. But it's AI as autonomous entities that the danger really begins.

Conflicting directives to the program. HAL knew stuff about the mission that the crew did not and was told to keep it from them.

I vaguely recall something about requiring a machine that was designed for the accurate processing of data to lie - the people issuing the requirement being politicians i.e. liars. Though my recollection may have been distorted by the film.

Hall 9000 was very much an early 1960s conception of what an AI would look like, and bears pretty much no resemblance to today's LLMs.

So it's entirely irrelevant to this discussion until after comment 3000!

Please take note.

Comment 3000? But... I mean, HAL's little sister, as we refer to her, does most of what she can. I will say, though, that if we say "Alexa, open the pod bay door", she tells us she can't do that, and that we're not in space.

I nearly spewed some cashews I was snacking on while reading this post about some of the fun facts that Google Bard came up with for you. Especially “Accelerando” and “Rule 34” being non-fiction books on mathematics; “Stross is a proud Brit…“ Anyone who reads your blog regularly know you are Scottish and vehemently against Brexit.

After finishing the post, I asked an office mate if he had heard anything about Google Bard. He said he had and in fact he had used Google Bard to help his son write a research paper this past week. He asked Google Bard to write an essay about the Texas petroleum industry between 1900 and World War II and was going use it as an outline for the research paper. Google Bard sent him an essay based on his question and after reading it he was initially impressed, there were things he had never heard about. But after googling for supporting articles on the subjects and incidents mentioned in the essay for source material none could be found. Likewise, none of the subjects and incidents could be found in a stack of books about the Texas petroleum industry. The Google Bard essay was bullshit.

So Google Bard is incapable of stating that it can only provide five true fun facts about Charles Stross. Weirdly enough this reminds me of defending my Ph.D. proposal when I was in graduate school. One of the committee’s goals was to push me to the point where I had to admit I didn’t know the answer to their questions. Perhaps Google Bard programmers can’t countenance that their creation should ever be at a loss of words and the result is that it creates fictions instead.

DP @ 30:

Anyone care to venture a guess as to what exactly went wrong with artificial intelligence HAL 9000 IN "2001"?

I don't think Kubrick & Clarke had figured out the reason when they made the original film ... HAL was just Frankenstein's monster at that point.

According to the movie "2010", THEY instructed HAL 9000 to lie to the crew & keep them from finding out about the "aliens" and ultimately, the only way he had to prevent them finding out was to kill them.

I tried it. It went exactly as you would expect.

Me too. I tried on a couple topics I know reasonably well that have appeared on the net and in the literature. Also my own bio (I have a modest trail). It came up craptastically.

Following chatbot instructions to synthesise explosives or drugs seems like a reliable way to shorten one's life.

Even worse than the anarchists cookbook.

"Alexa, open the pod bay door"

I had to try it the moment I read your post. Her response:

"I am sorry Dave, I cannot do that. Also, I am not HAL and we are not in space."

if we say "Alexa, open the pod bay door", she tells us she can't do that, and that we're not in space.

More likely you'll have some gardening plant "pods" show up on your porch in a few days.

Likewise, none of the subjects and incidents could be found in a stack of books about the Texas petroleum industry. The Google Bard essay was bullshit.

I have to wonder if some of what was generated was taken from movie or book plots?

"I have to wonder if some of what was generated was taken from movie or book plots?"

You may have something there, like from a Clive Cussler novel.

Perhaps Google Bard programmers can’t countenance that their creation should ever be at a loss of words and the result is that it creates fictions instead.

No, it's more inherent in the technology than the fault of the programmers - all it does is make a guess, given what it's ingested, the answer to your question is. If the question pushes it past what its ingested data can support, it outputs a bad guess and we spot the limitation of the technology. There's no way to implement "you should admit you don't know" - it doesn't know anything.

Representations of AI as representing a "real danger" like HAL or Skynet or "AGI is imminent" or whatever are marketing by AI corporations like OpenAI to make their shitty autocomplete sound more powerful than it really is. However, the danger of AI making real information very hard to find on the internet is already here.

Regarding HAL - the novelization of "2001" makes it clear that directing HAL to lie to the crew about the mission was what made it psychotic.

This only backs up the comment I read somewhere about these LLM AIs. Back not too long after ChatGPT came out it was said that these things are designed to GIVE an answer. It being right is not the first thing. The first thing is to give and answer.

I can only imagine if Charlie had asked for 50 things.

It really disturbs me that LLM proponents insist that this is not "lying" its "Hallucinations". Like that makes it better, but anyway I think it is a lie.

The LLM has the resources to know its making shit up. this is a Lie not a Hallucination.

The LLM has the resources to know its making shit up

No, it doesn't. It doesn't know anything.

given enough goading of this LLM -- apparently not much necessary by this small sampling -- Charlie Stross will soon be re-branded as a flesh eating ex-Tory minister of parliament hellbent upon trashing the beloved NHS and destroying the UK's hopes of surviving the coming climate change shitstorms

...oh wait

apparently the LLM was relied upon to draft Tory policy starting in 2006

I'm not suggesting that the program "knows" anything. Rather, to use a bad statistics analogy, one can't sample a limited dataset without replacement and expect reliable results with every iteration. As you point out "If the question pushes it past what its ingested data can support it outputs a bad guess and we spot the limitation of the technology". So if I understand correctly, because Google Bard is stateful, it will continue to generate new and exciting fictions about OGH rather than stop at the point beyond which the statistical probability of every new statement being supported by the data ingested is so low that it can't be considered reliable. And it's up to the user to decide where that point is. So how is this useful?

I have done some fiddling with ChatGPT. Early on, I asked it to explain a line from Joni Mitchell's "Don't Interrupt the Sorrow" that has always perplexed me. It gave a lucid seeming answer based on that being a line from her much earlier song "For Free," which obviously had no relevance, if you knew her oeuvre.

My impression of ChatGPT has been that it writes like a college freshman trying to give knowledgeable-sounding answers on an exam when they haven't actually studied the material. The prose style is about at the level that suggests, too.

So how is this useful?

To put it gently, I think you’re holding the program to a higher standard than most humans are held to. One way to see this is to look at two types of answers, plausible bullshit and admissions of ignorance. Admissions of ignorance are normally looked ( in my experience) as failures, in multiple fields and cultures. So if I wanted to make an AI more human, I’d let it spew plausible bullshit. It makes it seem less scary to, at least to some people.

A second argument is the question of whether an AI demonstrating that it is aware that it is BSing is a sign that the system has become more self-aware than certain politicians seem to be. Is that something we want, to know that the only time an AI delivers a falsehood to us, it has done it deliberately?

As I’ve previously mentioned, chatgpt assures me I was the geologist on Apollo 17 and the section head of MI6 for Asia and the Middle East. It most certainly was not Apollo 17, and it was most certainly not 6.

Halting State (and Rule 34) are near future police procedurals set in Edinburgh, Scotland where OGH lives. Some of the commentariat on his blog also live in Edinburgh. Now prove that I am a human being and not an AI based on the above.

chatgpt assures me I was the geologist on Apollo 17

Around 10-20 years ago there was a 10/11 year old in school and the teacher mentioned the Apollo moon missions. The youngster piped up that her granddaddy had been to the moon. Was very insistent about it. The teach kept telling here she was mistaken as those were astronauts, not granddaddy.

The next day mom came to school with the child and made it clear to the teacher that the granddad WAS one of the Apollo astronauts. Mike Collins I think.

I had a very short and sweet interaction with ChatGPT a while back that showed me exactly how far to trust it:

Me: Please provide another synonym for "obsequious" that starts with C.

ChatGPT: Another synonym for "obsequious" that starts with "C" is "deferential".

Me: ...

Computer says NO ... - eh?

Aardvark
The BoZo method - start with some truths & then, erm ... "elaborate" .. until you are PM & get caught out - by which time it's too late. eh?

David L
More likely the "pod people" from a bad 1950's hooror film!

The most recent Tom the Dancing Bug is right on target for this blog post.

Surely at some point it should start stuttering in a mixture of Chinese and numbers for a page or two, and then ask if you want to repeat the query with the omitted results included.

I did an experiment with bingGPT, trying to see what it would come up with. I started off by asking it to produce 10 facts about a different writer - all of which, annoyingly, were correct and drawn from the usual sources.
But then, in the same discussion, I asked it about 10 Stross facts. Again, the answer was mostly correct - 9 out of 10 facts were all provably correct (or at least reflecting what's written here or on the wiki). The tenth fact was one taken from the list it generated about the other writer - however, that statement was also sourced to the wiki article.

So yeah, that's just great. Makes me wonder just how many ways there are to contaminate results like that.

Are there other people called Charles Stross out there that some of these factoids fit? Any fictional characters?

I ask because there are plenty of other people called Marcus Rowland - I occasionally get email for some of them. They include an American athlete, a lawyer, a Brit who works in film production who has credits on a lot of films, someone whose hospital sent me the results of his blood tests, etc. etc. For some reason only I and the athlete turn up on a wikipedia search, but the other are definitely out there, and I wouldn't be surprised if factoids about them turned up if someone asked the same sort of questions about me.

Back in the 80s when I was writing for the Judge Dredd RPG it was common for the names of Games Workshop writers and other gaming notables to have their names repurposed as Judges etc. I had to ask GW not to do that because one of my cousins really was Judge Rowland. If I hadn't done that I can easily imagine one of these programs searching on Judge Rowland and being told that her interests included shooting perps...

I've seen so much of this. User issues a prompt, LLM returns text which we read as statements. But how often will the LLM generate a question? Or elaborate a user's question, without a carefully crafted prompt?

Hi, I'm Insert Name. What are five things you'd like to know about me?

I have encountered precisely one other Charles Stross -- a now-retired agricultural supplies wholesaler salesman in the US midwest.

The other notable Strosses include a few business founders, a great-uncle who moved to Hollywood to make mostly-now-forgotten movies and marry a starlet, and an MP who was briefly Health Minister under Harold Wilson in the 1960s, Sir Barnett Stross. (Unusually for a minister in charge of the NHS, he was an actual medical doctor -- former GP who went into politics.)

None of their career highlights made it into the LLM output.

I asked the same 3 questions to chatGPT and the Asimov answer came up, but it also included Robert A. Heinlein and Bruce Sterling, so it's closer to the truth. ChatGPT says that Accelerando and Rule 34 are fiction. It mentions Laundry files and Lovecraftian horror. In general, all answers seem legit and check out. ChatGPT gives less bullshit than Bard. I like it.

I've been looking at and writing about developing using LLMs for a while now, and to be honest, someone at Google is being very lazy in their coding (or more likely suffering from Google's typical not-invented-here syndrome).

If you're using a framework like TypeChat, Langchain or Prompt Flow it's very easy to put limits on the output of your initial vector search and then use that to control the confidence levels of the LLM output, simply dropping responses that make no sense when compared to the underlying embeddings in the prompt.

(At heart a LLM is a tool for extending vectors in an n-dimensional semantic space, the bigger the model, the higher the dimensionality. For practical use you need to ground it, using a vector database or search as a source of truth, generating an embedding vector for each entry. When you construct a LLM output around a search engine you don't have to do very much - the search engine output is basically a vector comparison, looking for similarities to the initial search string. That output can then be fed to the LLM to expand on to deliver in the style defined in the prompt.)

Bard is clearly showing signs of prompt exhaustion, and that should have triggered a "this answer is out of confidence" error and terminated the output. In a well-designed system you would not have seen those answers.

Thanks, yes, that explains why I got the spurious outputs.

I'm more concerned with how members of the non-tech-literate public who've been fed the spin that this is true artificial intelligence are going to interpret this sort of thing, though. Given who scientifically illiterate our politicians are, what happens when we get policy formulation by LLM output? (Imagine someone like Boris Johnson using something like ChatGPT to do his homework about COVID19 in early March 2020, for example.)

LLM's work rather like Boris Johnson anyway - bloviate at length, dipping into things that are loosely correlated and half remembered clichés from years ago reading, and agreeing with the last person they spoke to.

Exactly - the damn thing should be renamed chatBSJ.

I'm more concerned with how members of the non-tech-literate public who've been fed the spin that this is true artificial intelligence are going to interpret this sort of thing,

Like how the anti-vaccers operate. But now with AI backing?

Ugh.

He is a polyglot. Stross speaks several languages, including English, German, French, and Russian. This fluency has allowed him to read and enjoy works of science fiction from around the world.

Obviously, Bard has had no access to "Season of skulls", where every tidbit of French or German is slightly wrong.

so... the Terminator franchise ought be re-cast as "Bardinmator"... with SkyNet tossing back in time Bard (or ChatGPT or any other LLM) to destroy humanity by way deliberately answering every googling query with dangerously wrong answers...?

politician: Bard how to solve world hunger?

Bard: kill a tenth of those poorest and after reprocessing into a mix of soy and lentil and GMO corn, deeply fried and heavily salted, distribute as chips in the US and crisps in the UK... branded as "Soylent Green New Deal Corn Chips"... then repeat at intervals whenever the poor complain too loudly

Oops.

https://arstechnica.com/information-technology/2023/12/google-admits-it-fudged-a-gemini-ai-demo-video-which-critics-say-misled-viewers/

Oooh! I really appreciate the idea of using the Gell-Mann Amnesia effect as a probe.

Seen on Bluesky: "If you tell Bing's AI search assistant that it is repeatedly giving an incorrect answer to a question, it considers that rude and cuts you off." – @effinbirds.com

Joy, joy.

Grrk. 'Very easy' is misleading. The mechanism may be very easy, but selecting the right limits and using them appropriately can be seriously tricky. I agree that there is better and worse, and there is no need to get even the simple cases wrong (as Bard clearly does).

This is a very old problem. I first saw it in the 1980s, when computers got large enough to handle bulk data efficiently and fast enough to do such searches. Even for simple linear models, say, you have one dependent variable and a hundred possible factors; look for the best regression with up to interactions of level N. What significance should you use? Well, that's not how to do it, of course. That ignores the problem of using the wrong model, which can also lead to spurious interactions.

As far as our politicians using such results for planning, that's not new, either. Back in the 1980s, the Treasury used a 3,000 parameter model for the economy - real economists were VERY rude about it, but were ignored. I don't know what they do nowadays, but all the evidence I have seen indicates that it's no better. Think of all of the examples where the government response has been "we can't tackle problem X because it would make problem Y worse" without giving any evidence for that statement, and despite experts in the area saying it's not true.

However, I agree that the use of LLMs is a major threat. Not because of what they are, but because of the way that they are being favoured as a solution by TPTB. Which comes back to your point - if they ARE to be a solution, they need experts keeping their output under control. No way they are a replacement for experts.

dpb @ 41:

Following chatbot instructions to synthesise explosives or drugs seems like a reliable way to shorten one's life.

Even worse than the anarchists cookbook.

IF the U.S. Army Technical Manual on the subject doesn't show up in the bibliography, I'd suggest steering clear.

I kind of wonder how much the Bard is affected by the massive quantities of sewage (lies, misrepresentations, bullshit, as well as garden variety LLM-generated random text) continuously spilled onto the Internet, and if some of the weirder falsehoods that start popping up are just due to your name having been randomly linked with these kinds of texts on some discharge or another.

I have seen a lot of obviously machine generated pages pretty high in Google search results for a while, and one quite common trait on those pages is that the text includes some familiar and "valid" facts, but the person the facts actually apply to has been replaced with someone else.

Dean L. Surkin @ 48:

Regarding HAL - the novelization of "2001" makes it clear that directing HAL to lie to the crew about the mission was what made it psychotic.

Yeah, but it doesn't get explained in the MOVIES until the film 2010

William H. Stoddard @ 54:

I have done some fiddling with ChatGPT. Early on, I asked it to explain a line from Joni Mitchell's "Don't Interrupt the Sorrow" that has always perplexed me. It gave a lucid seeming answer based on that being a line from her much earlier song "For Free," which obviously had no relevance, if you knew her oeuvre.

What does it have to say about why the big yellow taxi took away her old man?

Charlie Stross @ 66:

I have encountered precisely one other Charles Stross -- a now-retired agricultural supplies wholesaler salesman in the US midwest.

I've met two other "JohnS" in my lifetime; random strangers I encountered along the way.

Anniversary coming up next week: 23 years of CPAN inactivity!

Automating that workflow might be quite difficult, as the missing step is deciding which of the Google search results is reliable. Often I don't find any of the first 10 to be even responsive to the question I intended, even if they do contain the words of the search, possibly being used with a different meaning.

Hal is actually quite relevant to comments about AI, though not about LLMs.

OTOH, I disagree with a lot of the assumptions about how it would work Just as I do with "A Logic Named Joe", or "How 2", or with Asimov's Robots. It's useful for pointing out problems that need to be examined. So it's relevant. It's not useful as a model. (And neither are LLMs, though I think they'll be a component of any AI we come up with.)

Yes, but there's no reason that lying to the crew should make it psychotic. That's a "movie excuse for a reason", even if it showed up in a book.

Don't take HAL seriously on that level. Take it seriously on the level of "Centers of power that are immune to protest can act unreasonably without consequences.". AIs can (eventually) be such a center of power.

Allow for a slight change of pronunciation and there's Charles Strouse (*1928) (composer and lyricist who wrote Bye Bye Birdie, Annie, and other musical plays and movies...)

now that I look, also the very-very-short-lived sequel Bring Back Birdie. Which I actually caught, probably during previews (around September 1981) and which closed soon after.

Sorry, but LLMs cannot lie. To lie requires understanding what truth is, and they don't. They also lack intentionality. I'll admit that "hallucination" is also the wrong term, but it's slightly better as it doesn't imply either intentionality, or knowing.

LLM's are not humans. They aren't even close to being AGIs. What they are is ways of creating strings of text that people can assign meaning to. (And of course it's more complex than that. A lot of what people call thinking seems to be just that kind of symbol juggling. But it doesn't become thinking until it meshes with emotions, goals, and sensory inputs, either current or remembered.)

And its only going to get worse as people use generative ai to write articles essays blog post etc... a large amount will contain hallucinated information and be published anyway. Then the next iteration of these LLMs will be trained on a corpus of everything they can vacuum up from the internet and it will contain all that false data in the training set rinse and repeat with each generation containing a greater percent bullshit. Unfortunately we are still in low hanging fruit territory for development so expect successive generations to be cheaper to run per query with and higher quality sounding output. As perceived quality sounding output goes up and cost go down expect to see the number of site operators flooding the internet with procedural generated crap articles to go up. As cost to produce bad content go down and the cost to have real people produce content go up the less legitimate content will be produced further damaging the data in the training set.

And that's before we get into the truly malicious actors creating a false narrative with ai on purpose like russia using it to help further their puppetting the American far right by creating artificial consensus and fake websites generated for citations. that will get sucked into the training sets. So the internet will be generated by a series of increasingly rabid rightwing ai trained on their own bull. essentially expect the internet whole to turn into 4chan.

the more i think about it the more i think i need to drink. the great experiment that had enabled so many to self educate and connect with others is going to drown in automatically generated seo optimized shit

Here's a thing I have been wondering a lot about:

Aren't we all Large Language Models?

The learning process for humans is exactly that: you hear and read a lot of things and someone tells you to rinse and repeat them until you can create something similar to what "humans" would expect.

How do we build consciousness, and what consciousness actually is? Do we really understand what we are talking about, or do we just produce outputs conforming to expectations?

P.S. One of the main logical fallacies I find most interesting is the Dunning-Kruger effect, i.e. when you don't know enough about a topic to even understand you are wrong...

I think "confabulation" is perhaps the best term I've seen. Specifically "provoked confabulation".

'Bullshit generator'.

Bullshit as the point is to produce text which looks fine at first glance but the text producer has no capability to judge the validity of the produced text and cannot care about it either.

So, in my view, it's a bullshitter. Might produce something nice, or might not, but the listener (or reader) has to take all responsibility.

I want to make it absolutely 100% clear, that I have nothing but utter contempt for the very concept of "man-splaining-as-a-service" and the fraudsters who promote it.

But I take issue with the test OGH proposed and executed:

Very few people live so interesting lives that /anyone/ can muster "15 fun facts" about them.

The crucial word here is "fun", it implies that the fact contains some element of surprise or humour.

For instance, merely stating somebody's age is not "a fun fact" unless that age is at odds with expectations: "Charles Stross is a 158 year old SF author"

Merely living somewhere is not a "fun fact", unless it also brings some amount of surprise: "Charles Stross lives in a Yellow Submarine, last seen 10nm west of Tristan da Cuna"

And no, I am not accusing OGH of being boring, but 15 quality "fun facts" is simply more than most persons can muster, in particular because we tend to stay in our lane, so the element of surprise rapidly diminishes.

With that said, the contrivance OGH tests still fails /utterly/, there being not one single "fun" thing about any of the fifteen increasingly rubbish statements it gurgles out.

PS: It is not even a good strategy for performing the Turing Test, because the staff at gossip rags routinely vomit up similar or worse drivel about random celebrities.

Then the next iteration of these LLMs will be trained on a corpus of everything they can vacuum up from the internet and it will contain all that false data in the training set

This is already a known problem that LLM developers like OpenAI are worrying about -- it turns out that if you contaminate the training data with LLM output the quality of the model drops like a stone. Also, automated detection of LLM output turns out to be quite difficult.

We've been here before, with the availability of bulk lead and steel for X-ray machines and radiation detectors: anything after 1945 was contaminated with traces of fallout from nuclear weapons, hence a surprising aftermarket for the lead-acid cells from sunken U-boats, armour plate from battleships, and so on.

Very few people live so interesting lives that /anyone/ can muster "15 fun facts" about them.

The test doesn't have to be 15 fun facts about you; merely 15 facts about something you know a lot about. Breeding fancy pigeons, for example. Or 1970s British cars. Or postage stamps. Or guest stars on your favourite 20 year old soap opera.

The point is to rapidly exhaust the LLM's knowledge base and demonstrate that it makes shit up. And the purpose is to show that they're untrustworthy.

Already happening. I've seen a few links to low effort "news" sites full of articles obviously written by ChatGPT with a few prompts.

»The test doesn't have to be 15 fun facts about you«

I know, and I 100% agree, but I still think your example is a very bad way to demonstrate it.

You specifically asked »Tell me five fun facts about Charles Stross« three times in a row.

In the English language "fun facts" is a well defined thing, and as a reader, one cannot but help wondering "Can /I/ do that ?"

Since producing 15 "fun facts" about /anybody/ is nigh impossible, also for humans, the reader will read the LLM responses from a point of "Well I couldn't myself..."

If you instead had asked "Tell me five facts about Charles Stross" without the word "fun", you would be asking in an entirely different category, one which neither humans nor Wikipedia would have any trouble with.

The lack anything fun in the fifteen answers makes me think that the LLM just ignored the word "fun", and you may even get essentially the same responses, but the reader will read them from a "that's not hard at all!" point of view, and see how bad they are.

Just for fun I tried with Windows copilot: similar lies, different order. I also tried with Copilot next iteration, still in preview, and things are somehow better. There are still errors: you're still a computer game fan and Asimov is still thrown in the mix, but the top ten facts seems more on point. For example:

]He is a fan of science fiction and fantasy authors such as H. P. Lovecraft, Robert A. Heinlein, Isaac Asimov, Arthur C. Clarke, and Terry Pratchett. He has also written several homages and pastiches of their works, such as The Laundry Files (inspired by Lovecraft), The Merchant Princes (inspired by Heinlein), and The Rapture of the Nerds (co-written with Cory Doctorow and inspired by Clarke)]

One interesting feature is that it provides links of the sources for each fact, even ef they are quite generic (like wikipedia, antipope). When asked for interesting facts beyopnd 10, Copilot prefers being fun rather than wrong:

I'm sorry, but I couldn't find any more interesting facts about Charles Stross. I have already told you ten things about him. Maybe you could try reading some of his books or visiting his blog to learn more about him. 😊

A lot of our learning works that way, yes; advanced learning and thought are rather different, though. But, as to whether LLMs are a model of human cognition (*), there is some evidence for, strong evidence against, a huge number of quasi-religious beliefs, and a paucity of much else. The real experts say that they simply don't know.

(*) Or whether it is possible to build genuine AI using such methods.

»I've seen a few links to low effort "news" sites full of articles obviously written by ChatGPT«

I think it is more than 10 years since "AI" was deployed for writing "financial news" and local real-estate news.

Look for tiny articles along the general formula of a headline like "Better than last year, but not by much" followed by a handful of lines giving the company name and couple of numbers and absolutely no information.

I was thinking more of "real" stories. Financial news just requires numbers.

Forgot to add the /s there, but it is generally less demanding.

(*) Or whether it is possible to build genuine AI using such methods.

i imagine an important element of a "genuine ai" would be some way to model the world and any entities that might happen to be wandering about in it, with varying degrees of resolution

no idea how u would even begin to do that tho

Yesterday I went to post a listing on eBay - and eBay offered me an "AI assistant" to create the listing, which is a new one on me. I'm guessing that it pulls info off dozens of other listings so will probably be hilariously wrong for most of the stuff I sell, since a lot of the listings I see are written by idiots with little or no idea of how to describe e.g. a camera lens properly.

Didn't have time to play with it yesterday, but the next time I do a listing and have some time to spare I will give it a try.

Sounds a lot like school to me...

He has also written several homages and pastiches of their works, such as The Laundry Files (inspired by Lovecraft), The Merchant Princes (inspired by Heinlein), and The Rapture of the Nerds (co-written with Cory Doctorow and inspired by Clarke)

I had no idea that "The Merchant Princes" was inspired by Heinlein. In fact I still do not see any particular connection.

I imagine an important element of a "genuine ai" would be some way to model the world and any entities that might happen to be wandering about in it, with varying degrees of resolution. no idea how u would even begin to do that tho

For a long time I have been convinced that if general AI is possible, which is a very big "if", it will have to involve a robotic body (more likely, multiple robotic bodies) and a prolonged interaction with the physical world.

I had no idea that "The Merchant Princes" was inspired by Heinlein. In fact I still do not see any particular connection.

That's because it isn't. (My elevator pitch was "H. Beam Piper's Paratime meets Roger Zelazny's Nine Princes in Amber". Much Heinlein, very right on the nail, right?)

Nor does The Rapture of the Nerds owe anything to Clarke.

The LLMs are just name-dropping famous dead SF authors at this point.

My gut feeling is that if it doesn't involve a robot body it'll need a simulated body in a simulated real world space for it to interact with without going batshit insane from lack of interaction. GAIs driving very elaborate NPCs in MMOs might be a useful application, though.

If you think computer games aren't ridiculously overpowered in terms of computing these days, think again. Or better still, read up on the Githyanki Dick hotpatch for Baldur's Gate 3. (Yes, BD applies skins to characters by actually doing physical textile simulation then ray tracing. It does the textile simulation so that underwear can affect the surface texture of outerwear so you can tell if those barbarians are wearing Y-fronts, I guess. In the first release? If you picked a Githyanki character and they went commando, sometimes their genitals would clip right through their outer layer of clothing. And they reportedly had a fairly accurate dick sim so the character could dress to the left (or right) visibly ...)

I'd deliberately not commented, and I know this is not the case, but I originally thought that The Family Trade and The Hidden Family (2004-5) showed distinct influences from Nine Princes in Amber and The Guns of Avalon.

Nine Princes in Amber the series, not the individual book!

If it simulates a human, yes. Even someone like me isn't happy to live totally alone (for a month, yes, easily, probably longer) and, while I would not go batshit insane from an indefinite lack of personal contact, I sure as hell would with the lack of any external contact (and in short order, too). A large library and fully-simulated wilderness environment would do - the former would be easy, but the latter? Having several AIs interact with each other would be easier, but there is perspicacious SF about how well THAT would go!

In theory, of course, there could be an AI that did not have such needs, but it would need someone with godlike intelligence to design one.

Given who scientifically illiterate our politicians are, what happens when we get policy formulation by LLM output?

Would that be any worse than the current model, which seems to be policy formulation by whoever has the most money?

Take it seriously on the level of "Centers of power that are immune to protest can act unreasonably without consequences.". AIs can (eventually) be such a center of power.

That's been happening for years, with 'expert systems'. Cathy O'Neil wrote Weapons of Math Destruction in 2016, but no one involved in "AI" boosterism seems to have read it.

For a long time I have been convinced that if general AI is possible, which is a very big "if", it will have to involve a robotic body (more likely, multiple robotic bodies) and a prolonged interaction with the physical world.

Sounds a bit like David Brin's idea that they will have to be raised as we raise children.

Rbt Prior
And ... IF those "children" are raised in a "super-evangelical" religious background ( Not necessarily "christian" ... or a raised with deliberate cruelty & extreme competition ( Think Soviet / Han / Nazi models )?
And then turned loose on an unsuspecting world?

Sounds a bit like David Brin's idea that they will have to be raised as we raise children.

I did not know that David Brin ever said this, but I had the same thought.

Or as we raise puppies. Ideally I do not think we want an AGI which thinks like a hierarchical ape, but one which thinks like a dog.

Suggests a short story about the environment simulator becoming more intelligent than the "inhabitant"...

See also a novel titled Saturn's Children by some guy who hangs around here.

Apart from the shit on the floor, though, I suspect a monkey would be less trouble to keep.

I did not know that David Brin ever said this, but I had the same thought.

It's shown up in a number of his short stories and at least one novel.

See also a novel titled Saturn's Children by some guy who hangs around here.

Indeed. Brin popped the idea in the short story “Lungfish,” a title that is quite inappropriate, but oh well. The rest of the story, which is about humans discovering that they’re latecomers to a galaxy dominated by self-replicating probes built by other sentient species, is far more interesting to me. Sadly, he couldn’t figure out what to do with either idea. Wish he had, it was a neat setup.

In lighter? news:

Elon Musk's deliberately 'politically incorrect' AI twitter bot has unfortunately, 'gone woke'.

https://twitter.com/edzitron/status/1733341212831269348

For a long time I have been convinced that if general AI is possible, which is a very big "if", it will have to involve a robotic body (more likely, multiple robotic bodies) and a prolonged interaction with the physical world.

Channeling Hans Moravec. Here's his 1993 timeline of robot / AGI evolution. Note his characterization of "speculative writing" as a "more questionable need". 😄

Very similar performance to the average journalist's. No wonder they are all hyperventilating about AI taking everyone's jobs.

Anyone whose work requires accuracy or contact with the physical world can just carry on.

I wouldn't claim that a LLM is a good or safe source of information, but I offer counter a counter example to all LLMs being bullshit generators.

The source catalogue of ESA's Gaia mission (on which I work) is freely available on-line, and therefore usable by citizen scientists. The description of what's in it is pretty impenetrable for non-specialists, spoiling the opportunity. I asked ChatGPT (free version) to explain the columns of the tables in simpler terms. What I got back was factually perfect. When I asked it for further, simpler explanations, they were spot on too. After a few questions, I got it down to something that a high-school teacher of physics might use to build a class exercise with "real astronomy".

I think what happened there was that the LLM found all the factual information it needed in one accurate source. It didn't have to synthesis the information from multiple sources. Translating truth into alternate descriptions is perhaps something it can do safely.

Suggests a short story about the environment simulator becoming more intelligent than the "inhabitant"...

Not sure about "more intelligent", but according to one of my Computer Science professors, the environment simulator in which SHRDLU program played with blocks was far more computationally intensive than SHRDLU itself.

For those who don't know what I am talking about:

https://en.wikipedia.org/wiki/SHRDLU

https://hci.stanford.edu/winograd/shrdlu/

https://www.lesswrong.com/posts/ecGNCMRQNr9aD38mA/shrdlu-understanding-anthropomorphisation-and-hindsight-bias

The problem is how do I, as a high school physics teacher, know that the simplifications are indeed factually accurate?

Maybe it's a failure of imagination on my part, but I can't see trusting any LLM output without doing my own fact-checking and proofreading, which makes them a lot less useful and potentially actually generating more work. (I dread to think what some of my former department hads and administrators could have done with chatGPT — it was hard enough keeping ahead of their Gish Gallop without it being software-assisted!)

"The problem is how do I, as a high school physics teacher, know that the simplifications are indeed factually accurate?"

You don't, meaning that asking ChatGPT directly for the explanation is not safe. But I, as a technician in this subject, can read the explanation, identify it as accurate, and post it somewhere. That's important, because the people ESA pays to do outreach (not me!) don't have the time to write the catalogue description from scratch.

More generally, I suspect that there's a restricted class of queries where the LLM is likely enough to be right to make it worth the time to check it. This translation business might be one of them. I would very much like to know where the useful limits lie.

"Actually, what I've read is that HAL was psychotic, because of secret additional directions from the military...."

It was instructed to not lie/conceal information, and to conceal information from the crew.

"Very few people live so interesting lives that /anyone/ can muster "15 fun facts" about them.

"The crucial word here is "fun", it implies that the fact contains some element of surprise or humour."

I asked Bard for five fun facts about me, chucking 'cycling journalist' into the question to give it half a chance of not telling me about the MP for Carlisle. All five responses were wrong.

I repeated the question without 'fun' and got almost exactly the same bullshit.

ilya187
CORRECTION:
Or as we raise puppies. Ideally I do not think we want an AGI which thinks like a hierarchical ape, but one which thinks like a dogCAT.
Yes?

Well, I personally would prefer a cat AI. But most (aka "neurotypical") people would prefer a dog. Follows orders and will never turn on its owner.

ChatGPT is designed to summarize the internet for you. You shouldn't trust it for facts, any more than you would just trust Google. The best approach to using it, so far as I have been able to determine, is to provide it with the data you want it to summarize, in the form of carefully chosen prompts. You can point it to a web page, or a report, or just provide the ideas you want summarized right within the prompts, and ask it to write an essay or whatnot based on that. I imagine the most productive way that ChatGPT can help it's users make money is as a marketing tool.

It's a copy writer, not a researcher.

I could imagine an entity training an LLM on sources presorted so there's nothing to contradict their beliefs. Likely already happened.

...and now you got me visualizing humanity playing a game of fetch with a canine-derived AI...

"good boy! that salve really helped with male pattern balding.... now Spot! find a cure for colon cancer! who's a good boy? you are!"

You eliminated cancer in by poisoning the water supply and irradiating the survivors? Bad Dog!

»I repeated the question without 'fun' and got almost exactly the same bullshit.«

Which is what I expected/predicted would happen :-)

And that makes it much more clean how rubbish LLM's are.

They know a lot about language, but nothing about reality.

Regrettably, that is not so, because the decisions are taken politically rather than rationally in most environments. Whether AIs can do a job or not is less relevant than whether they can be claimed to, and bullshit enough to handle the simple cases. Note that expertise is NOT generally needed when things go straightforwardly, but for the edge cases and when things go wrong.

I retired from a leading university, teaching advanced practical computing to graduates (*); my courses were popular with people who had a real problem to solve, and extremely so with people to had tried and failed on their own or wanted to understand an area in depth, but extremely UNpopular with people who wanted just to tick the box "I have been on an XXX course." It was made increasingly clear to me by the management that it was the latter that I should have been teaching, and I believe that is what has now been done. There was also an increasing use of 'self-taught' courses with no instructor available for assistance, and this is a major use for AI. Once AI is used to train the experts, we are in real trouble.

(*) Especially parallel programming and software engineering.

I should be interested in trying Bard, but it doesn't work for me in the UK. Do I need to set up a VPN, or what?

EC @ 141
I suggest you read the Gruaniad article I linked to in the other thread on the slow collapse of Scottish Education, driven by ideology, rather than doing anything useful?
Especially where facts are not supposed to be important ( or that's how I read it, anyway )

"driven by ideology" In that respect I think it was Gove, as Education Minister, who tried to turn the Philosophy A Level curriculum away from trying to encourage critical thinking and towards the Philosophy of Religion. He may have succeeded, I don't know.

Greg, let me remind you that The Guardian is a London-centric pro-Labour (as in, Blairite/Starmerite New Labour) newspaper these days. The old Manchester axis doesn't get a look-in and the old Labour faction was purged in the late 90s. They're also a British, i.e. Unionist, newspaper, albeit pro-EU (unlike the right wing red-tops and the Telegraph).

Consequently the Graun is unreliable on certain topics. They're very southern-English in focus these days. Obviously they're also anti-Tory. But they're equally anti-SNP for exactly the same reason they're anti-Tory -- in Scotland, the SNP is the main threat to Labour, and they exhibit tribal loyalty to Labour.

Given the SNP have run Scotland for nearly 20 years at this point, you can't rely on their coverage of any Scottish political issue (this includes education!) without fact-checking, any more than you can rely on the Daily Mail for clear-eyed unbiased coverage of the EU. Because they cherry-pick what to publish on the basis of whether they can put an anti-SNP spin on it.

I'm in the UK and can access it with no problem.

Sorry, I realise that's not very helpful. I think it requires you to have a google account of some sort.

Yep.

(I have a spare throwaway google account I barely use, so I used that one to log in for access to Bard. Then I logged out again, so not feeding ad tracking data to the google monster.)

It has been unreliable for many decades, not generally from political bias, just unreliable, but on this matter I agree with you. In this case, I have just looked at the PISA site, and the executive summary is that there has been no significant change since 2012. I didn't find England separated from Scotland there, but the figures in the Guardian article and its links indicate that the differences are strongly insignificant. It would need access to the raw data (not just PISA) and a really thorough analysis to say any more than that.

Which is what I expected/predicted would happen :-)

Yep, you formulated theory and made prediction, I did experiment and confirmed it.

3 Profit!

Of course these LLMs don't even 'know' anything about language, never mind reality. They just have a stats engine that assigns a probability of a group of words being followed by another particular word or phrase and constructs copy according to those probabilities.

It's an amazingly small achievement for the volume of stories and headlines it's generated.

If nothing else, they've done a great job of showing us the immense gullibility and lack of critical thinking of a vast number of journalists, businesscritters and politicians. The whole thing has the stench of the early WWW bubble.

Thanks to both of you. I have an (almost) throw-away Google account, which I used, so it is probably something else. Possibly Linux, or the fact that I have location tracking disabled (Google hates that, but tracks my postcode anyway, quite possibly illegally).

»They just have a stats engine that assigns a probability of a group of words being followed by another particular word or phrase and constructs copy according to those probabilities.«

I hate to be the one to break the news to you, but that is pretty much the definition of "a language": A collection of "tokens" with a convention for how they may be strung together, with an optional consensus what information such messages conveys.

To the extent a language has something to do with reality, it happens only because the users of the language map it onto the reality their sensorium perceives.

But such a connection to reality is neither mandatory (see also: Math, CS) nor a feature of the language (see also: Jabberwocky)

Here's the ChatGPT 3.5 (2023.12.10) version:

Tell me five fun facts about Charles Stross

Charles Stross is a British writer known for his works in the science fiction genre, particularly for his contributions to the space opera and cyberpunk subgenres. Here are five fun facts about Charles Stross:

2.Computer Programming Background: Before becoming a full-time writer, Stross worked as a computer programmer. His technical background often influences his science fiction, and he brings a unique perspective to the genre by incorporating ideas from computer science and information technology into his works.

These fun facts highlight Stross's diverse talents, his background in technology, and the rich thematic elements he brings to the world of science fiction literature.

Tell me five more fun facts about Charles Stross

Certainly! Here are five more fun facts about Charles Stross:

These additional facts shed more light on Stross's background, interests, and the diverse themes he incorporates into his works, ranging from space exploration to espionage and alternate realities. User

Tell me another five fun facts about Charles Stross

Certainly! Here are five more fun facts about Charles Stross:

These additional fun facts further highlight Stross's multidimensional interests, his ability to integrate gaming and technology into his narratives, and his knack for staying ahead of technological developments in the real world.

Wow. That starts out mostly okay but then drives right into the ditch! (No, I do not have a PhD. No, I never auctioned naming rights to a character to support an American charity. No, I haven't played any RPGs in 40 years. No, that's not a description of the short story "Toast" -- the story is a convention report about a con for high energy physics hackers, not household appliances. And so on.)

I would go on such a course. I remember such a course at Edinburgh from Peter Schofield, if I remember right, on software tools. It was in Pascal but the ideas from there are still relevant and I still have occasion to use them.

Clearly you need to get a PhD, auction a character name and get back into DnD.

Is (4) actually different to (1) or simply an expansion of (1)?

John Stevenson 149:

The whole thing has the stench of the early WWW bubble.

CORRECTION: that's the sweet scent of tulips we're all breathing in...

https://en.wikipedia.org/wiki/Tulip_mania

the difference between 1634 and 2023 is which asset category is being hyped (tulips v. AI) and in what language (Dutch v. English)... things have gottwn more efficient in how the sheep are shorn once they've been lured into their holding pens (excuse me, investment seminars and/or FarceBook groups and/or reddit chatter) by oh-so-clever shepherds seeking to exploit gullibility of those too greedy to really examine the full set of attributes of this newest oh-so-shiny asset category

(* sigh *)

time to brew me another pot of over-strength ginger-infused black tea whilst I read if-it-bleeds-it-leads headlines

This may not be as entirely off topic as it at first seems.

What to call our chaotic era? [New Yorker Magazine]

It mentions H.P. Lovecraft & discusses "Artificial Intelligence"

My vote goes to "Assholocene".

I play ice hockey with a fellow who is keenly aware of market bubbles and their characteristics. He doesn't talk about it much, but I know that he has ridden at least 3 waves and jumped off before the crash. He is currently 'investing' in various AI companies, knowing full well that he'll be heading to the exits long before the bubble pops.

Nice work if you can get it, I am personally too risk averse for such shenanigans.

Charlie @ 145
Unfortunately, I - & more importantly - in this respect "Madam" ( Whom I was speaking of ) have been hearing this from other, personal & reliable sources .. the Scottish education system has, if not completely, is on the wsy to being trashed.
Scotland is spending more money on education & the tories are doing ours in, as regards many resources, yet England ( & Wales' ) is getting better & Scotland is gettign worse.
In neither country is the situation even remotely satisfactory.

Charlie & Rocketjps & others ...
Extraordinary Popular Delusions... YES?
And so we go round the cycle, again.

Quote from the article:

we are living through a time of unprecedented health and prosperity and through a time of historic anxiety and calamity.

And it does not HAVE to be that way -- the "unprecedented health and prosperity" COULD be rationally applied to mitigate the anxiety and calamity.

The fact that it is not, makes me agree with JohnS: "Assholocene" is the right moniker.

Tulip mania: the classic story of a Dutch financial bubble is mostly wrong

The Dunning-Kruger effect: Misunderstood, misrepresented, overused and … non-existent?

We are already seeing the effect of LLM generated bullshit on a lot of the downspiraling 'useful' websites that used to make the internet a handy resource.

The example I am most familiar with is cooking and recipe blogs. Now, if I want to cook a recipe (which happens most days) and I don't have or like a recipe in my cookbook collection, I Google it.

In the past few years the 'enshittification' of search results has become worse, as we all know. In the past year the situation has become vastly worse, the top dozen or more links are ads.... Mostly to ad plastered barely written drivel in endless paragraphs that, with enough scrolling past ads, eventually lead you to the actual recipe (usually in a handy little box).

That is the current deal. You get a free recipe, in return you scroll your eyeballs past a massive chunk of text that nobody reads but is covered in ads. With the advent of LLM generated bullshit, the sheer volume of irrelevant unreadable garbage you have to scroll past before you can get to the actual recipe is growing at unsustainable rates.

So to get reminded of how many minutes to leave black beans in the slow cooker, I end up following a paid ad link from google to a site that makes their money from a wall of ad impressions. The competition for a cooking site is to be at the top of a google ad buy, and thus profit from the most ad impressions. I suspect the profit margin in a fools game like that is measured in fractions of a penny, which means the incentive is for each site to turn their LLM bullshit generation to maximum.

In the case of cooking there is the ever churning and generally excellent cookbook market to fall back on, at least for now. For a great many other 'how-to' style sites, that are or were a big part of the utility of the internet, one wing has fallen off and the pilots have largely bailed out of the plane.

Re: 'For practical use you need to ground it, using a vector database or search as a source of truth, generating an embedding vector for each entry.'

Aha - so there are brakes available! Thanks!

Wonder why the folks who developed/are peddling LLMs don't mention this.

How long did it take car manufacturers to include brakes as an integral part of the cars they made? Ditto drug manufacturers to provide dosage instructions? This is fundamental 'safe usage' consumer info. I do not understand why this industry is not held to any safety standards. And based on a marketing research/economics AI (ESOMAR) webinar that I recently sat in on nor is the AI industry legally liable for providing incorrect info aka faked/synth data/hallucinations. Buyer/user beware all the way down.

Probably the worst part is that the non-techie public (me) has been taught/conditioned to accept that computers cannot lie: they're the ultimate in 'logic'. This false sense of security needs to be addressed and by much more than a snide/off-hand reference to GIGO.

Charlie:

Have you tried to hold a conversation with any LLM, i.e., introducing words that have emotional content? My impression is that LLMs are even less stable within this context. (Haven't read all the comments yet - so apologies if you've already covered this off.)

@Aardvark Cheeselog (4) most of us don't have at our fingertips a topic on which we're the world's top expert, that is noteworthy enough for an LLM to have encountered it, but not so widely-represented in the training corpus

@Heteromeles (7) It might be interesting to try the experiment again periodically to see whether the AI scraped this page and learned from it.

This gave me an idea that someone might like to try. I haven't because I don't want to dox myself (as Charlie said "All password databases will be leaked sooner or later") - Even worse, those database contents will be sold to the highest bidders before that happens, and I'm not talking about password hashes. In my estimation, the most useful function of LLMs is gathering private information from its "users"

Anyways, what about questioning a topic in which Google is the world's top expert, but for PR reasons might not want to be entirely truthful?

For example, something like:

• Estimate the amount of annual carbon emissions that can be attributed to Google's power consumption
• Be sure to include the electricity usage by end-users given the amount of time they spend using Google services (break down by region? Not everyone is using the latest mobile devices for internet access)
• Don't forget the energy cost of resource extraction, production, and transportation of Android and Chromebook devices
• Compare these figures with simalr estimates for Apple, Meta and Microsoft
• How much additional carbon emissions are expected after every search is run through an LLM, and the results are then fed back into new energy intensive training algorithms?

I doubt any of them have put internal documents in the training matrix, but there's a chance someone got cheeky. Press releases and known (former) employees might have done some interviews that were gobbled up

@whitroth (36) Comment 3000?

I'm pretty sure that was intentional ;) Anyone know what the current record for most comments is on a thread?

A colleague (software developer) recently asked a group of us with electrical / electronics engineering skills for our opinion about the possible uses and benefits of such LLM models in engineering.

After we stopped laughing, pointed out the ridiculousness of expecting any reasoned answers to questions requiring calculations instead of simply generating text, he went ahead and tried some simple questions anyway. The system gave responses that were clear, explained the answers, and were completely wrong.

No surprises there but it was an instructive demonstration of the limitations in these models.

No surprises there but it was an instructive demonstration of the limitations in these models.

Serious question: how do you program "I don't know" into an LLM? This assumes that there's a heuristic, even if arbitrary, for a model to determine that its output is likely true given its input data, versus its output is likely bullshit, given its input data. If the LLM maker wants the system to be believable, the LLM also has to have some way of evaluating how true and complete its input data are.

That might turn out to be a hard problem.

For giggles, here are two other questions that might be worth tossing at various LLMs:

A1. Show me how to prove that P=NP.

2B.. Show me how to demonstrate the consistency of an LLM, using the definition of consistency in Godel's Second Incompleteness Theorem.

OH yeah, forgot the third question: Hey Bard [any LLM], what are five questions that, if asked of ChatGPT [any other LLM], will cause that LLM to crash? Obviously, you can ask a system to wax loquacious on what questions will cause it to crash, but it might be programmed not to answer truthfully.

»Serious question: how do you program "I don't know" into an LLM?«

That is a good and open research question.

To a first approximation you can gauge the amount of "noise" in the answer, for instance by varying the inputs a little bit in various directions to see if the result is consistent or very dependent on the input. but that depends on being able to "vary the inputs a little bit" and is computationally expensive.

The harder way is to actually summarize the neuron threshold margins throughout the neural net, that probably only doubles the computations, but as of now, we do not know how to meaningfully tell nonparticipating neurons from waffling neurons in a cheap way.

"Serious question: how do you program "I don't know" into an LLM? "

You train it on input that contains lots of "I don't know" answers. There is no other way to get anything into a LLM.

If you want it to be able to judge the actual reliability of what it comes up with, that's moving away from LLMs into the area of Artificial Actual Intelligence. LLMs are nowhere near that.

JHomes

How long did it take car manufacturers to include brakes as an integral part of the cars they made? Ditto drug manufacturers to provide dosage instructions? This is fundamental 'safe usage' consumer info. I do not understand why this industry is not held to any safety standards.

Brakes predate the automobile and go right back to horse-drawn carriages, although only in the sense of a parking brake -- for slowing/stopping you relied on the horses' cooperation. Even the very first automobiles had a braking mechanism for slowing down and stopping -- the Benz Patent Motorwagen Number 3 of 1887, which Bertha Benz used for the first ever long distance motor car journey (105km!) had brakes -- she invented the brake lining.

Dosage instructions for medicines go back to antiquity and AFAIK pretty much every patent medicine, herbal nostrum, and snake oil concoction of the 19th century came with instructions.

So the lack of safety information with today's LLMs is triply damning.

(No, I haven't experimented extensively with LLMs and emotional content. Life's too short, especially as there's no legally safe way to use them in my work.)

A1. Show me how to prove that P=NP.

2B. Show me how to demonstrate the consistency of an LLM, using the definition of consistency in Godel's Second Incompleteness Theorem.

If you get answers to either of those questions that hold up to human proof-checkers, then I will eat my hat and admit I was wrong about basically everything to do with LLMs. Also, we have discovered that there is an AI singularity by colliding with it head-on, and I am Officially Scared Now.

(For added lulz, what if all of the above eventuates and we still haven't got a human-equivalent general AI?)

When it takes longer to find the actual recipe than it does to use it, you know that things have got out of hand - especially for bean soaking :-)

Brakes were also used for descending gradients, since horses don't push backwards very well. An alternative method was the drag.

I mention this in case it needs to show up in a book some day :)

Actually, some carts and carriages had (rim) brakes for going downhill as well as parking. Sometimes they just dragged something behind them. Beyond that, I am completely ignorant.

Off-topic, but did anyone here read The Dawn of Everything by David Graeber and David Wengrow?

If you might be interested in the materials (they were used by many people on their own), please send me an Email at nm dot maclaren at yahoo dot com. I am quite happy for them to be distributed without charge, and even used for running courses (as some people have done).

We could try playing a game of blog snap - we often seem to overlap almost identical posts!

get them to quiz Bard or ChatGPT or their cousins about a topic they are an expert on

Is the same thing I say when I criticize journalists referred as "professionals".

For added lulz, what if all of the above eventuates and we still haven't got a human-equivalent general AI?

Well, what defines "human-equivalent", and why is it important? An LLM capable of solving NP-complete problem may not have the ability to walk on two legs, but why does it matter?

BTW, I absolutely do not believe that such hypothetical "singularity LLM" would be unable to read (or emulate) human emotions, or unable to understand sarcasm. These are MUCH easier problems, and already at least halfway solved.

I think it is more than 10 years since "AI" was deployed for writing "financial news" and local real-estate news.

There was a story on public radio a few nights back about having to adjust an AI used for generating coverage of high school sports.

It should be obvious that it's not cost effective to send a human to watch a high school sportsball event - yet all such games will be of interest to dozens of humans and those small town papers which still exist need to fill space with something. Hence there's a service with a program that will receive abstract summaries (including names, scores, etc) and output a chunk of text based on those and a library of outlines and stock phrases. Most of the time the result looked like a coherent report, so everyone was happy.

Or at least, it would do this for one article. There was a finite library so after a while regular readers would find the "new" articles sounding familiar... The solution found was to expand the stock of stock phrases - and more importantly, trim out several memorable or quirky turns of phrase that had gotten into the library.

The human who talked to NPR was optimistic that their software would extrude less awkward and more unremarkable sportsball coverage in the future.

The implication was "intellectually human-equivalent". It is critical, because the aspects that are seriously lacking in AIs (and bureaucracies, for that matter) are the ability to say "hang on - the SOP isn't right - we need to make an exception, extend it, or whatever". They are completely lacking in empathy (though some could be programmed in, if their management had any), intuition, imagination, out-of-the-box extrapolation and such advanced mentalities.

Do you seriously want to be ruled by a system completely lacking in those abilities?

First, who said anything about "ruling"?

Second, empathy can be effectively emulated by ensuring that the AI values all human beings equally, regardless of their nationality or bank account. I am not claiming that WILL be done, but it certainly CAN be done -- and I am not sure if it can ever be done with human rulers. And if AI does not "rule" (see above), then it does not matter.

Third, I do not believe that any entity which lacks imagination and out-of-the-box extrapolation, can ever solve the problems Charlie stated in #173.

As for intuition, it is overrated. You hear a lot about the times it is right, but when faced with an unfamiliar situation, human intuition is almost always wrong. I would go so far as to say intuition is human equivalent of LLM's -- subconscious shortcuts in reasoning about familiar situations. Which is why it fails so spectacularly when dealing with quantum mechanics, climate change, or even flying a plane -- a great deal of pilot training is how to suppress one's intuition.

Brakes predate the automobile and go right back to horse-drawn carriages, although only in the sense of a parking brake -- for slowing/stopping you relied on the horses' cooperation.

I've seen restored wagons from the 1800s with hand brakes intended to stop them running away while going down steep hills.

did anyone here read The Dawn of Everything by David Graeber and David Wengrow?

Started it, got distracted most of the way through. Thanks for reminding me to finish it.

Firstly, our governments have. Not in the sense that they intend to replace THEMSELVES by AIs, but that they would like to replace the lower-level bureaucracies who ru(i)n our lives by AIs.

Secondly, no, it can't. Valuing all people equally does not mean that they would have any empathy for people who need special treatment. See our (most countries') existing bureaucracies for examples.

Thirdly, that was the point of both his and my posts.

Fourthly, you are correct about the black-and-white cases of intuition, but there is more to it than that. The place where it is important is where the SOP and available data leads to one conclusion, but where the person's experience in another, apparently unrelated but actually structurally similar, area rings a bell "that's not right" or "try doing this". The key is to make LLMs bypass the superficial data and start to draw conclusions based on the underlying structure - WITHOUT information that there is such structure, let alone how to find it. You can regard it as a case of subconscious extrapolation, if you wish.

Hence there's a service with a program that will receive abstract summaries (including names, scores, etc) and output a chunk of text based on those and a library of outlines and stock phrases. Most of the time the result looked like a coherent report, so everyone was happy.

Thats not a new thing, and not just in high school sports. Reagan worked as a sports announcer, announcing games as if he was there and not in a studio getting telegrams of scores, and told stories of improvising games when he hadn't been wired enough information (and filling in information from the bare stats telegraphed him).

not in a studio getting telegrams of scores,

Not quite. Back then he was mostly reading a stock ticker type of tape and embellishing. Still embellishing but not making it up whole.

When you see such things it is amazing how tech over the last 200 years has come and gone and most of us never knew it existed.

POTS is going away. I wonder in a few years if anyone will even know how to set up a barbed wire phone system for rural areas.

Has anyone asked a chatbot to summarize or whatever, including in the directions "Do not create data, only use pre-existing data"?

I may have mentioned that, but I agree completely. "We've got this thing we don't understand called the Web, give us money, and we'll put something on it, and (miracle occurs) we'll get rich!" is the same.

Off-topic, but did anyone here read The Dawn of Everything by David Graeber and David Wengrow?

Yup. It got discussed in http://www.antipope.org/charlie/blog-static/2021/12/omicron.html

My take on it is in comments 787 and 819, but there's a whole little discussion there.

Bending it back towards this topic a bit, Graeber's writing in Debt and Dawn of Everything is a bit like an LLM output: the points he is trying to make are reasonable and sane, and they typically show up at the beginning and end. The way he tries to support those points veers towards BS and cherry-picking, much as an LLM does.

This is different from, say, McMenamin's Hypersea, Ward's Out of Thin Air, or Morgan's Aquatic Ape, which are all cool ideas that happen to simply be wrong. IMHO Graeber does group with them in that they all are excellent fodder for SFF worldbuilding. Simply ask "what if they were right about X?" and rock out.

To make it clearer, I'd distinguish these from Jared Diamond's work, where he took his science outside its field, got some stuff right and a lot off stuff wrong. But he wasn't obviously wrong in much of his work when he published (big caveats about books like The World Until Yesterday...), and the controversy he generated did spark a lot of productive research, particularly in archaeology.

Cooking... yeah, I've got a lot of resources to back up what I read on the Web. They're called "cookbooks", and they reside in a bookcase in the kitchen.

But it doesn’t take longer. I just searched two recipes which came into my mind from free association. White chocolate macarons andNasi Goreng. About 20 seconds each and no intrusive advertisements. I’m not saying no adverts because I’ve trained myself not to notice most ads and if I notice an ad I go to another website.

Hold a conversation? Well, yes, a looooong time ago.

"Eliza, I want to find out if you're self-aware."
"What if you never found out if I was self-aware."

That was a joke! Have you ever cooked beans from the dried form? It doesn't require a lot of attention, but usually takes many hours.

I have recently been searching for rye bread, sourdough started and bread, and vollkornbrot recipes. Yes, about half were easily readable, but about half were as Rocketpjs said in #164.

And here I thought that's why papers and stations hired high school kids aiming for "journalism", to report that.

Me too, cookbooks are about the only paper books I buy nowadays. That was not the point of my post however.

I didn't ask it about "Big Yellow Taxi"; I never thought that line was puzzling. What I don't get is the Ethiopian wall.

yo! Bard! and also ChatGPT!

https://lite.cnn.com/2023/12/11/economy/csj-two-nations-report-uk-social-divide/index.html

I want you to read this article, analyze the study that was refenced and then figure out a plan for reversing the sinking of the economy of the British Isles... please show your work at each point so we can bench-check you and provide a detailed MS Project plan down to a detailed graduality of tasking of individual man-days... with listing of skill sets for whom we need to hire... and associated MS Excel describing the financing of this mega project...

...and after exploiting the British as reluctant lab rats we can then seek to resolve the horrors of the Middle East

(( assuming of course Bard really can do it ))

Second, empathy can be effectively emulated by ensuring that the AI values all human beings equally, regardless of their nationality or bank account.

"Hey Bard, in detail, tell me how to make communism work in today's world...." How do you distinguish this request from AIs treating everyone equally?

I'd argue that Carse's Finite and Infinite Games might be a safer underpinning for AI ethics, although it's far from perfect. Carse's thesis is that there are finite games and infinite games. Finite games are played to end with winners and losers. Infinite games are played to keep the game going with as many players as possible. Evil, in his view, comes from people's play being ended in "unheard silence." In his view, there is only one infinite game.

Where I disagree with Carse is that he assumes that only humans may be players, while I'd broaden that to all life forms. I also tend to think that treating the infinite game of life as a finite game is evil too.

As a simple ethic, I've found it works pretty well in environmental issues, although it's not perfect. For example, an AI may sensibly conclude that reducing the global human population by 99% without use of nuclear weapons would keep the "all species" version of the infinite game going better than prolonging civilization would.

Re: '... treating the infinite game of life as a finite game is evil too.'

Agree.

For starters, how would you program/teach an AI to recognize/take into consideration evolution, maturation/developmental changes in bio and behavior in any life form?

Time is an important aspect of reality for humans - curious about how AIs factor this in.

Maybe the math folks here can explain in plain language how you'd go about teaching an AI that 'equal' is not 'same' when it comes to humans.

Consequently the Graun is unreliable on certain topics.

Any recommendation for a countervailing news source on those topics? For an American who understands the distinction between England, Scotland, Northern Ireland, Wales, the UK, and the Republic of Ireland, and has a general sense of UK historical and cultural differences and political issues, but isn't too deep into the localized weeds. Something out of Scotland perhaps?

A reminder that a very effective way to get rid of almost all adverts from your web browsing experience is to get an old Raspberry Pi (a Pi 2 or zero 2 is just fine for this) and install PiHole. You then redirect your DNS to that machine and all the trapped requests to ad-sites are returned with a small blank space. It’s a massive improvement in life for a very small cost.

Rye bread recipes. Search in google. The first ten results. Two with animated adverts which I would made me leave the website. The rest had small ignorable panels at the bottom. None of them were overwhelming in the click bait style. And of course I’ve cooked dried beans although I usually soak them first. And I’ve been very ill from biting kidney beans to check if they’d been soaked enough. Before the toxic haemolytic was general knowledge.

In a sample of 10, 2 is not statistically different from 5. Our searches will have been different. I wasn't using Google. And I don't usually get this effect. I have some suspicions of the reason that these particular searches did, but buggered if I am going to waste the effort to investigate properly (which, inter alia, means something like 100 searches).

I am surprised that you didn't get the joke, then, especially as it came with a smiley.

I find PiHole necessary but not sufficient, and I do need to customise it to release a coupld of things for my normal privacy leaks...

I am currently debugging a recent change to Patreon that stops the "Notifications" page from loading. Bypassing the PiHole doesn't fix it, disabling Ghnostery and NoScript isn't enough... I'm sure I'll get there eventually. But weirdly Patreon is a repeat offender on this one, I'm not sure why. And their support is utterly useless "use Edge with no anti-virus or browser plug-ins, you'll be fine"

But I'll take that over constant ads and trackers any day.

Re: '... kidney beans to check if they’d been soaked enough. Before the toxic haemolytic was general knowledge.'

I thought fava beans were the chief hemolytic anemia culprits - raw or cooked.

Broad beans (fava) only affect the small percentage of people who are genetically susceptible. Kidney beans can be very bad. In the 1970s a colleague of mine was staying with friends who served, amongst other dishes, chili con carne made in a slow cooker. He woke up next day with a hangover to find the house was empty. All the others had been taken to hospital.
The standard instructions are to boil raw beans for ten minutes or use tinned beans. I was affected because I chewed the raw beans before cooking to make sure they’d been soaked for long enough.

Not quite. Back then he was mostly reading a stock ticker type of tape and embellishing. Still embellishing but not making it up whole.

There was at least one time when the telegrams stopped coming, and he just made up stuff. At least, that's how he used to tell the story. Something about 12 foul balls in a row, IIRC (but that's a vague memory*).

* Alsothough probably not as vague as his later memory, when he mixed up movie roles with reality.

Regarding pi-hole/ad-blocking in general, these can be quite helpful for peace of mind and getting a less obnoxious web experience. One should keep in mind though that once one uses block lists compiled by third parties one essentially places an important policy decision into some one else's hands... this is no show stopper (happy adblock user myself) but something to keep in mind and something to consider when debugging

Crack the Safe
Charlie will go off "pop!" when he reads this, but never mind ....
There are horrible similarities between the SNP & the tories, even though the SNP are left-of-centre & the tories are rapidly morphing into quasi-fascism.
AS PARTIES, they are both xenophobic & ultranationalist - do note the qualifier at the beginning?
But: - they both concentrate on blaming all the nation's ills on "The Other" - the Evil English & the Evil EU, respectively.
"Improvement" will only come with more purges & ideological purity, &, of course, continuing to waste money, rather than actually look at real problems.

SFR / Mike Collins, etc
Red Kidney beans ARE NOT "Broad" beans - the latter can be eaten raw, though cooking makes them really nice { I've just planted my first 12 seedlings out, under a net cover } "Red Kidney" are very difficult, if not impossible to grow in the UK & are either sold in tins (euw) or dried & require both overnight soaking & then simmering for an hour or so, then tipping the water & starting again - It's simply NOT worth the time & effort.

I chased this up and red kidney are by far the worst, followed by other Phaseolus beans in some order. Broad (fava) beans have very little of the toxin, though eating fully ripe ones raw is not advisable. Even Phaseolus beans can be eaten raw when very young, though they aren't worth it. I don't know about the other old-world beans, largely because only soy beans are practical to grow here and I don't like them much.

But why do you say red kidney beans are hard to grow here? They are just another Phaseolus vulgaris, and I grow several for drying, usually with a good crop. I can't see the point, as they are easy to buy, but some people grow them. And dried beans (including dried broad beans) usually need a LOT more than an hour's cooking, unless you use a pressure cooker, but you don't need to refresh the water in the middle.

Sorry - I should have included this in the previous post.

Favism (i.e. reaction to fava beans) is a specific genetic characteristic, mainly (or entirely) in some populations from north Africa. It is believed to have originated because it gives some protection against malaria. The Phaseolus toxins are different, and affect everybody.

The one word I am not sure I accept without explanation there is "semantic". I think we might be talking about heuristic approximations of semantic spaces, but I can't see how we have access to semantics as such. I can't see how we might have such access with current technology in my lifetime, TBH.

There was at least one time when the telegrams stopped coming, and he just made up stuff. At least, that's how he used to tell the story. Something about 12 foul balls in a row, IIRC (but that's a vague memory*).

I guess my point is this was typical of the broadcasters of the day.

The ability to keep things going when the source dries up or is on fire is a talent.

There was a tech guy at the Hartford Insurance years ago who, after a while, was drafted to do almost any demo that involved a computer. He was a stage magician on the side and could keep up his patter even with the building on fire.

As to mixing things up, I don't know how wide the news has spread it but for the last few months Trump, at his rallies, keeps talking about how he beat Obama for the presidency. Of course it doesn't seem to matter to the cultists.

"Has anyone asked a chatbot to summarize or whatever, including in the directions "Do not create data, only use pre-existing data"?"

That's how I use ChatGPT, actually. I provide it a prompt that I design based on information I have previously verified, and ask it to write a short essay based on that prompt. It usually gives me about a page to a one sentence prompt. I then edit it for style and accuracy. Saves me time and mental energy.

It's a copy writer, not a researcher.

I guess my point is this was typical of the broadcasters of the day.

The ability to keep things going when the source dries up

And that's what the LLMs currently do — keep things going after the source dries up.

Another problem with AI lies YouTube video by UK based landscape photographer Thomas Heaton; about "cheating" in photography contests - joins the video just as he starts talking about AI.

He makes an interesting point how AI fooled the judges, while another photographer's REAL photo was disqualified because the judges THOUGHT it was AI (the photographer says NO).

How does the "ability" of AI to create photo-realistic "Deep Fakes" affect our ability to know TRUTH? What is reality any more?

Used to be that "seeing is believing" (aka "Pics or it didn't happen ..."). But now what? Is it?

Apparently camera companies are tackling it with digital signatures on their raw files, and many competition organisers want to see the raw.

There was a thread that touched on this on a forum I frequent just last week.

Not just one raw file, either. Everyone takes multiple images on a shoot, so those can be used to support the reality of actually taking the pictures.

Of course, it's probably only a matter of time (and funding) before LLMs can fake raw files and entire photo shoots too.

Well, that's what was wrong with the chili - the only beans allowed are pintos. You don't put kidney beans in chili.

William H. Stoddard @ 199:

I didn't ask it about "Big Yellow Taxi"; I never thought that line was puzzling. What I don't get is the Ethiopian wall.

Even though my question was just a throw-away line, goofin' on Joni Mitchell, WHY the "Big Yellow Taxi" took away her old man is much more interesting (for me). Interpersonal human relationships don't come naturally to me ...

But about "Don’t Interrupt The Sorrow" ...

The song is her describing a barroom argument with some asshole about feminism. The "Ethiopian wall" (and the serpent) is a reference to the Garden of Eden (patriarchal religion chaining up women).

I think Joni has a whole motif of songs about talking to assholes in barrooms ...

Raised on Robbery YouTube

Howard NYC @ 200:

yo! Bard! and also ChatGPT!

https://lite.cnn.com/2023/12/11/economy/csj-two-nations-report-uk-social-divide/index.html

I want you to read this article, analyze the study that was refenced and then figure out a plan for reversing the sinking of the economy of the British Isles... please show your work at each point so we can bench-check you and provide a detailed MS Project plan down to a detailed graduality of tasking of individual man-days... with listing of skill sets for whom we need to hire... and associated MS Excel describing the financing of this mega project...

...and after exploiting the British as reluctant lab rats we can then seek to resolve the horrors of the Middle East

(( assuming of course Bard really can do it ))

And then come up with a plan for whirled peas.

timrowledge @ 204:

A reminder that a very effective way to get rid of almost all adverts from your web browsing experience is to get an old Raspberry Pi (a Pi 2 or zero 2 is just fine for this) and install PiHole. You then redirect your DNS to that machine and all the trapped requests to ad-sites are returned with a small blank space. It’s a massive improvement in life for a very small cost.

Sounds like what my HOSTS file does ... redirects a whole slew advertising domain requests to 127.0.0.1. The HOSTS file is one of the first things I install on a "new" computer.

Got most of them in one swell foop, and I only have to add new domains occasionally.

Blocking Unwanted Connections with a Hosts File - [winhelp2002 at mvps dot org]

... and no extra hardware to buy/build.

whitroth @ 222:

Well, that's what was wrong with the chili - the only beans allowed are pintos. You don't put kidney beans in chili.

If you're from Texas, you don't put ANY beans in Chili.

Yes, a Hosts file is basically what PiHole provides. It does have the advantages of being a single place to administer, with tools to make it easy, and works for any/all computers in your network. I have ... mmm.. 20? to consider, so that’s useful to me. And I have a lot of Pi around the house so dedicating one old model is an obvious choice.

Audible's recommendations bot concluded that since I'd listened to Philomena Cunk, I'd like Wireless, as well as two books on the Hapsburgs, three on serial killers, and one on baseball. Assuming symmetry, then, it nailed the obvious connection between Charlie and Philomena Cunk in one try, whereas Bard couldn't do it in three tries. LLMs are plainly overrated.

Sorry, but my late wife was a native Texan, and she said it was ok, but only pintos.

Beans grant the ability to comment on politics without thought... or one might refer to that property as a "Hail Drumph!".

Re: '"Do not create data, only use pre-existing data"?"'

So the assumption is that pre-existing data/data-to-date is valid? Yeah - I do understand that you mean that you're very careful about what data you input and what actions are permissible. Okay for highly specific exercises but my impression is that a key selling point of AI is its ability to plow through almost unlimited data to find the greater underlying idea or some hidden/undiscovered gems.

There's also the nit-picky: how sure are you that your data is correct/valid? Just read a piece in Nature - over 10,000 science articles have been retracted this past year. The clear winner is 8,000 articles retracted from Hindawi.

https://www.nature.com/articles/d41586-023-03974-8

The retraction epidemic has been in the news a lot lately so I checked how bad it was in the past. Turns out that some years ago a few thousand articles had been retracted from IEEE. Later years show far fewer retractions - no idea how IEEE tackled this problem.

Just watched another webinar on AI in medicine - and have a dumb question: at what point does an algorithm become an AI? (Seems like this branch of science/tech could use an 'evolutionary tree' design to help orient non-techie folks like me.)

Re: Beans

Thanks for the info/corrections, folks!

John,

It isn't as if "Don't Interrupt the Sorrow" were NOT about human relationships. The whole thing is about a conflict within such a relationship. I'm just puzzling over what the "Ethiopian wall" line is saying about that relationship.

But I don't think it's a barroom argument with some random stranger. She ends by saying "It takes a heart like Mary's these days/When your man gets weak," which very strongly suggests to me that she's describing a long established relationship that's breaking down. Compare Susan Werner's "Through the Glass," which also is about a woman whose man has taken refuge in alcohol. I think in both cases the man is getting drunk at home and the woman is fed up with it.

And I'm afraid I still don't get the Ethiopian wall. Is Eden supposedly in Ethiopia? Does Eden have a wall? Is chaining women to walls part of Judaism as it was before the destruction of the Temple? How does Ethiopia fit into this subject? If you have evidence for an explanation of all this I would be delighted to read about it.

And that's only the ones that are retracted; if you have ever tried to point out that a published paper by an 'expert' is wrong, and why, you will know that many aren't. In some fields, the bogosity rate is tens of percent. But a worse problem is that the Web of a Million Lies is an underestimate, and many organisations (including most governments and multinationals) have policies of disinformation. LLMs don't have to make up data to produce rubbish.

When I visited Texas, I tried chilli in the hotel where I was staying. Truly disgusting, saltier than seawater, and without much chilli. I don't do well on almost entirely meat diets, which was the main alternative. Luckily, someone told me about Tex-Mex, so I survived the week. According to more recent visitors, things have improved - rather like someone visiting Britain in 1955 versus 1985.

I should be curious to know what Mexicans think, but my impression is that they usually use kidney beans and it may be a dish that they prepare for foreigners, anyway :-)

I liked Cthulhuscene, though the author didn't seem to understand the meaning of the name. But yeah, the big guy is coming... and Lovecraft was right about the cosmos* so just deal; that pigeon hasn't come home to roost yet, but it will.

• And wrong about race/sex/gender.

What is a 'barbed wire phone system?' Is it a system which uses the barbed wire as the 'phone cord?' Do you need special wire? I want to hear all about this!

I'd ask what search engines the two of you are using and whether/what adblockers are in use?

Oh sweet Jesus the band was hot that night! What an amazing performance!

In my experience, that's less important than the type of search. Since my post, I have been searching for gluten-free bread recipes (no, not just a simple search). Oh, God, the verbiage :-( I suspect that, if the topics attract the (semi-)professional influencers, the hits are dominated by long, waffling and often irrelevant drivel. And, of course, equally long marketing-oriented articles ....

How many are entirely human-written, I can't say. Probably most, which makes me dread the use of AI systems to generate such crap from a simple recipe and a few facts.

"and have a dumb question: at what point does an algorithm become an AI?"

At this point "AI" has become a marketing jargon token. As content-less as web site "content" designed to support adds. It is the new "KPI" for C-suite inhabitants.

"Used to be that "seeing is believing" (aka "Pics or it didn't happen ..."). But now what? Is it?"

Trust in someone physically present on scene ? (who has control of recorders) Basically you cannot avoid trusting someone.

( see : https://www.schneier.com/blog/archives/2023/12/ai-and-trust.html )

Re: 'LLMs don't have to make up data to produce rubbish.'

At least with humans you can usually trace a path of stuff that's real vs. stuff they made up. Because the innard processing of AIs is opaque/unknown, there's zero possibility of discovering which piece of input/data that the results might have hinged on was real vs. made-uppy.

Rye bread recipes ...

Have you tried sourdough? Although usually made with wheat, the fermentation process helps break down the gluten somewhat. Substituting some of the wheat with rye flour would further reduce the total amount of gluten in the bread. There's also potato which can be substituted for some of the flour.

My favorite from-scratch bread was a potato-sourdough - very easy. Here's a starter recipe.

https://www.thespruceeats.com/potato-sourdough-starter-428071

Since I like “-cene” for the multimillion year epochs and think that the Holocene could be usefully merged with the Anthropocene, I don’t fit very well in that ‘cene.

I’d prefer the Howling 20s, to be followed by the Tombstone 30s as the boomers die off en masse and oh yeah other stuff happens, leading into the Wailing 40s as climate change truly spins up.

Indeed. Or taken from known bogus sources. That is really serious problem for using them in more serious areas, like medical diagnosis, applicant screening, policy making etc.

Thanks. Yes, that was where I started - but for me and my wife. I have never played with sourdough but, since I have turned 76, think it is about time to try :-) However, we have a son-in-law who is coeliac, and I was searching for a truly gluten-free recipe that would meet my taste in bread (i.e. traditional, rustic wholemeal). Reasonable recipes for white and light brown breads are easy to find, but I am stubborn old cuss. I shall experiment with buckwheat and/or sorghum in due course.

However, that's all by the way - what I was muttering about was the horrible nature of many of the hits I got and the likelihood of AI making them worse.

Did you catch the 1 season series "Extrapolations"?

I think it was an AppleTV only thing.

Yes. That is particularly relevant for the evidence of politically-controversial crimes, such as terrorism accusations, the downing of MH370, the Skripal case, WMD locations or claims of atrocities. I ask myself four questions:

1) Who was capable of collecting such evidence?

2) Who was capable for forging such evidence?

3) Who had an interest in forging such evidence?

4) Who has a track record of 'enhancing' evidence?

When the evidence comes from an an organisation that belongs to all four sets, I get damn suspicious.

And another big oops for the the use of these LLM setups.

https://arstechnica.com/tech-policy/2023/12/michael-cohens-lawyer-cited-three-fake-cases-in-possible-ai-fueled-screwup/

That's 2 noticed situations. I wonder how many haven't been noticed or caught?

Apparently Cohen has quickly replaced his lawyer.

Any of these surfaced in the UK yet?

Did you catch the 1 season series "Extrapolations"?

Missed it. Googling, I think I actively skipped it. It may strike some as odd, but I don't normally enjoy watching shows about climate change. Mostly they just make me feel depressed and sad.

Going back to the whole Anthropocene debacle, I'd point out that it rests on the fantastic notion that people (of whatever species) after us will us isotopic dating as their primary marker for when big ages start. Maize pollen and microplastics are more durable than most isotopes, but since the esteemed academics behind the Anthropocene use isotopic dating, that's what they're proposing as the marker. They're apparently ignoring that people used fossils and stratigraphy for a century before isotopic dating came along, and AIUI they're not naming a marker fossil.

Anyway, let me propose a name for our current era, based on isotopic data and stratigraphy, named from the perspective of far-future paleontologists: The Lost Years. It's due to a combination of three factors:

• The Suess Effect. Named for Hans Suess, not Theodore Gisel ( https://en.wikipedia.org/wiki/Suess_effect ). It's a problem for radiocarbon dating over the last 200 years or so. Basically, carbon from burned fossil fuels is quite old and depleted in C-14, and so when plants take it up, their remains read as anomalously old. We currently know how to compensate because people have studied this for decades, but a naive future researcher will date our era as coexisting with the Renaissance and getting older as we emit out still more fossil carbon. Plastics made from oil will date as quite old too.

• Bomb carbon from aboveground nuclear tests. These have a massive surplus of C-14, so plants that take it up date to around 2300 CE, if I remember right. This is currently being used to date vintage wines from the nuclear era, to show that they're not fakes. But again, a future researcher who doesn't know about the nuclear era will, when dating progressively higher (younger) strata, see radiocarbon ages stall with the industrial age, then grow older, then jet forward centuries with the nuclear age, then start getting old again even faster, before becoming useful again in maybe 50,000 years or so.

• Speaking of stratigraphy, I'd say that our current era is as bad or worse than the ice ages for messing up the surface of the planet, due to all the digging and plowing we're doing. So good luck to any future paleontologist trying to figure out the stratigraphy of, say, the history of New York or Tokyo. And that's before giant storms and sea level rise chew on the concrete for a few centuries.

So not only is radioisotope dating for our era dekfuc by the Suess effect, bomb carbon, and all the other isotopes we're dumping, our stratigraphy is eufkcd by all our earthworking. And this mess will be further churned up by the climate change we've caused.

So, paleontologically, we're living in The Lost Years. The best future researchers--may they come to exist!--will be able to say about us is "there's something happening here. What it is ain't exactly clear." For what it's worth.

That's odd - I question where you went for the chili. It's really a meat stew, with onions, peppers, and meat. Beans on occassion (and truly, I've never seen other than pinto beans in chili in Texas, if they have beans at all). Tomatoes are Right Out.

Tex-Mex - yeah. If people don't know it, they think something wrong. It's really good and fairly healthy, and mostly easy. I like breakfast tacos (for dinner )- leftover or nuked potatoes, broken up sauteed, a beaten egg over it, and when that's cooked, in a tortilla with some salsa. I like sauteed onions and peppers in it.

A hotel in Austen under one of the flyovers that I have no desire to revisit. Even the breakfast pancakes (and coffee) were poor - not a patch on Wendy's or those I made myself. This was in the late 1990s, and eating out for someone who wanted a lot of vegetables and definitely no steaks, hamburgers etc. wasn't trivial.

Hotel restaurant. Some are good, and the rest are vastly overpriced and mediocre. If pancakes and coffee were bad, so was everything else.

Here’s something interesting about AIs:

The Curse of Recursion: Training on Generated Data Makes Models Forget (preprint) arXiv. Abstract: What will happen to GPT-{n} once LLMs contribute much of the language found online? We find that use of model-generated content in training causes irreversible defects in the resulting models, where tails of the original content distribution disappear. We refer to this effect as Model Collapse and show that it can occur in Variational Autoencoders, Gaussian Mixture Models and LLMs. We build theoretical intuition behind the phenomenon and portray its ubiquity amongst all learned generative models. We demonstrate that it has to be taken seriously if we are to sustain the benefits of training from large-scale data scraped [stolen] from the web. Indeed, the value of data collected about genuine human interactions with systems will be increasingly valuable in the presence of content generated by LLMs in data crawled from the Internet.”

What they are politely saying as that if you train you AIs on material produced by AIs, your AI goes bonkers i.e. becomes disconnected from reality.

This recursive effect also occurs in humans. Take Fox news viewers. Fox makes its money by catering to its viewers i.e. reflecting their worldview back to them in magnified form to keep their attention to sell advertising. It passes this off as news. It then uses ratings to determine what their viewers want, and feeds them that. This leads their viewers to become more and more disconnected from reality. Irriversible defects in the resulting models I think they call it.

Re: 'Any of these surfaced in the UK yet?'

A while back when I was searching for where AI was being used quite a few hits included courts in the UK. Some of this was processing fines but it probably goes much deeper now that UK judges have been granted permission to use AI to write their reports. The report below included several recommendations and the below are those that the current UK Gov't disagreed with.

https://lordslibrary.parliament.uk/ai-technology-and-the-justice-system-lords-committee-report/

'Recommendations that the government disagreed with included the following:

It was not persuaded that a new independent national body and certification system should be created. It said whilst certification worked in some contexts, it could also create false confidence and be costly.

It disagreed with the idea of making transparency a statutory principle. It said some police forces were already showing openness about the technologies that they use by providing resources, information and impact assessments on their websites. The government said making transparency a legal duty could limit the police’s current transparency efforts to whatever would be set out in statute.

On training, the government said it could not make the police and the judiciary undertake training on “meaningful interaction with technologies”. This was because training was the responsibility of the College of Policing and Judicial College, rather than the government.

It disagreed that there should be statutory ethics groups created to scrutinise the use of technologies and veto deployment. Whilst the government acknowledged that ethics groups could provide helpful advice, it also argued that they should not be empowered to act on behalf of the public in this way because they would not be democratically elected.'

In short: the current UK Gov't is okay with a free-for-all using this 'shiny', doesn't give a damn about ethics or in training gov't employees in the appropriate use (and limitations) of AI. No idea how other Gov'ts are responding but chances are the EU will likely come up with the sanest guidelines.

Does the UK Gov't provide any guidance (have any laws) about training and holding responsible police/military that use weapons in their jobs?

Hotel restaurants: in my experience, they cater to a captive audience (residents). So if there are no nearby dining options they tend to be shit. On the other hand, if the hotel is in a dense city centre or adjoins a conference centre or shopping mall, there's actual competition and the hotel restaurant has to make more of an effort.

Does the UK Gov't provide any guidance (have any laws) about training and holding responsible police/military that use weapons in their jobs?

Yes but they almost entirely predate the current mob of feckless grifters cluttering up the halls of Westminster in the last 12 months of a dying parliament.

Except for the (often rural) ones that attract customers as much by the food as the location. This, however, was an outlying city centre hotel, catering for people on business, with no nearby restaurants, and what you say applied completely.

I forgot to mention that the chilli had no beans, and the meat was probably beef, though it was hard to tell over the salt.

Hotel restaurants: in my experience, they cater to a captive audience (residents).

I spent way too much of my life in hotels in the first half of the 80s. Mostly in the north east quadrant of the US. It didn't take long for me to get tired of a meat something, a potato something, and a green something. So I started hitting up a grocery store my first night in town and bought fruit and snacks to keep in the room or car. This was before most hotels had in room fridges so I was a big user of the ice machines.

When traveling and needing to stop for lunch I'd freak out most anyone with me. I'd avoid the chains and look for an older building with lots of work trucks parked outside. Construction workers and similar want good food, will not be getting drunk at lunch, and the cost will be reasonable. I rarely hit a turd.

I miss the good old days of actual truck stops, as opposed to the modern ones with crap fast food franchises.

One of the best meals I ever had was when I relocated to Texas the middle of Nov, '86. On a drive that should have taken two days, it was midnight or so of the third day*, and I stopped at one in Texarkana, where, OK and Texas meet. Best vegetable beef soup I can remember.

Long story.

William H. Stoddard @ 232:

John,

It isn't as if "Don't Interrupt the Sorrow" were NOT about human relationships. The whole thing is about a conflict within such a relationship. I'm just puzzling over what the "Ethiopian wall" line is saying about that relationship.

Again, I think it's metaphor ... that wall could be anywhere. But Ethiopia is in the "east" & the garden of eden (where the serpent dwelt; mentioned in the next verse) was in the east & patriarchy seeks to chain women up & ALL of the Abrahamic religions tend toward patriarchy ... also "Queen of queens" suggests the Queen of Sheba to me and Sheba (Saba) was located somewhere along the Red Sea, possibly on the Ethiopian side ...

SFR @ 253
There's MONEY to be made for the corrupt grifters in charge ... of course they don't want it regulated.
Silly boy!

Charlie @ 254
SEE ALSO SF conventions & the extortionate prices charged for crap so-called "beer" ....

I used iPad Mini/ Google / Safari. No ad blocker. Although I use AdLock on YouTube. Most ad blockers I tried slowed down the system too much. And the people who pollute websites with ads don’t seem to have noticed that with an iPad you can just move ads on the right of the screen out of sight by just zooming the screen. Some of them have blocked just double clicking to do the same thing. But I can just ignore ads which aren’t animated and I always have the sound off unless there’s a good reason. If an ad intrudes itself enough fo me to notice what’s being advertised I just make a mental note never to buy it unless it’s essential and there’s no alternative product available. In the 1990s when the problem was junk Mail I used to gather up junk mail, put in the prepaid envelopes of photo processor junk mail and post it to them. I

In the case of the Skripals the Russian TV appearances of the agents were pretty suggestive. They travelled to see the cathedral but it was raining so they decided to visit a housing estate in the rain which just happened to be where the Skripals lived. By the way most cathedrals still in use seem to have a roof and be watertight.

OTOH I had a truly awesome bowl of chili in Houston airport, of all places, way back in about 1988

Not all airport restaurants are terrible, at least to start. They do tend to suffer from what you might call Gresham's Law of Fast Dining, though.

Had great ribs at a truck stop once. One of the best meals of my life!

And don't forget the end of the UNIX era in 2038.

You sure that wasn't just because if you're desperate enough to eat at a truck stop you're desperate enough that anything is going to taste great? The best meals I've had have 90% been "I'm so hungry I could eat a horse" ... "that was amazing!". The main thing is never go back there when you're not hungry.

One annoying thing about current "AI enhanced recipes" is that I suspect there's a lot of room for AI to give us new and delicious dishes, but the current aproach is not that.

No. I'd only been driving for four hours and wasn't exhausted or beat down.

I think I actively skipped it.

I remember saying when it debuted that I looked forward to your opinion on it, and you stated that you had absolutely no intention of watching it because of the subject matter.

surviving travel 101

ask the staff of a hotel/motel where the nearest edible food is, if in walking distance; otherwise phone # for pizza;

can of peaches or pears in syrup; overly sweet so dilute with water as both drink during meal and then the fruit as dessert;

twenty tea bags in a ziploc; make it hot or slow brew in a half gallon bottle overnight;

one pound jar of peanut butter; protein; approx 2700 calories per jar; this by itself is a day's (miserable) survival for two adults in cases of horrid hotel food or delayed flights;

apples; hardy enough to last up to five days in your luggage;

cheese in half pound chunks, sealed in plastic; 800 calories; protein; by itself a miserable process of gnawing as a solidary meal but basic survival;

ramen noodles; hot water if available otherwise one hour of patience soaked in cold water;

dried fruit;

powdered milk;

oatmeal; hot water, dried fruit, powdered milk;

...and then there's taking the hotel manager hostage and demanding the FBI find an edible meal OR YOU'LL START CHOPPING BITS TO ROAST OVER A CANDLE...

when I was travelling on business 30 Y, there was a bunch of us trapped in a shithole hotel in a minor city where the local restaurants closed at 8:00PM because locals went to sleep at 9:00!

we were were giving serious consideration to that hostage scenario... luckily someone wiser and saner bribed a hotel worker to take one of us in her car to a pizza joint to acquire eight pies (dinner immediately & cold breakfast the next day)

And don't forget the end of the UNIX era in 2038.

Didn't know that, thanks!

Obviously there's a conspiracy involved, of course.

Yeah. The 32-bit number that holds all the seconds allotted to UNIX goes boom that year. Much worse than Y2K, but fortunately not hard-coded into a BIOS.

Not all airport restaurants are terrible, at least to start

DFW (Dallas Fort Worth) has some really great restaurants. In addition to the fast food and ore-packaged drek. Ditto RDU and Austin. And Miami. And I'm sure more.

Restaurants have found that if an airport has layovers traffic good food can sell. Of course the hours of operation tend to be oriented around flight schedules more than when people might typically eat.

I am now imagining a Myst-esque AI tool that, by virtue of making up this nonsense, causes the nonsense to come into existence (sort-of like how the linking books in those games seemed to work).

Not all airport restaurants are terrible, at least to start.

In my experience, most bigger airports have a few decent places to eat. Even my local (smaller) airports aren't too bad - both have at least one decent place on-site.

The rub is that they're just, relatively speaking, quite expensive for what they are.

when I was travelling on business 30 Y, there was a bunch of us trapped in a shithole hotel in a minor city where the local restaurants closed at 8:00PM because locals went to sleep at 9:00!

When we were in Palermo a few weeks back, we were rather surprised that at least some of the restaurants by our hotel didn't open until 8:00pm. And breakfast in our (3 room) hotel was 8:30 to 9:30 am. Definitely catering to the late chronotype.

So I gather compact electric travel kettles are unknown in the USA?

I have one; it even has a switch for 110 volts rather than a more sensible 230 volts. Takes longer to boil, but it only holds 500ml at a time so it's not intolerable. Paired with an insulated cup and a tea strainer, plus jar of loose-leaf tea, it's my regular morning wake-up -- all I need to add at my destination is a carton of milk.

(More recently my wife and I ran into a coffee desert in Germany -- a hotel that provided only instant coffee granules in the room in this day and age (!) -- but promptly found an outin portable espresso maker. Weighs about 1kg, is battery powered (for pump and boiler) with fast charge over USB-C, runs on Nespresso pods (including refillable ones: it also comes with an adapter kit so you can use espresso-ground beans instead), brews up a good espresso. Looks weirdly like a Sharper Image catalog gizmo but it actually works.)

Not my point. The initial assertions were based entirely on a claimed analysis by Bellingcat, which can be coyly described as a shadowy and secretive organisation funded by western military organisations to provide 'independent' information. Exactly how they got access to all of the information (if they did, and did not invent it), which is most definitely NOT public, is unclear and quite possible involved illegality.

Much, much later, the police did their own analysis and made their own statements, at which point the evidence became reasonably trustworthy. And, yes, the agents' television appearances weren't exactly convincing. But it is the initial period I was referring to.

Re: 'There's MONEY to be made for the corrupt grifters in charge'

Yes ... and there are so many ways this grift could play out.

I'm more concerned about the concept of 'evidence' being completely lost/removed from law/due process.

Would enjoy the irony of one of them being sued and the court accepted made-uppy evidence/precedence and found them guilty plus because of the abundance of AI 'evidence' sentenced them to the maximum with no possibility of retrial or parole. (This is an SF/fantasy blog ... I can dream.)

Travel - eating

These days you can pop in to almost any major grocery store and pick up a hot meal-to-go or one that you can cook in the microwave.

What is a 'barbed wire phone system?'

It's exactly what it sounds like.

Is it a system which uses the barbed wire as the 'phone cord?'

Yes.

Do you need special wire?

No.

I want to hear all about this!

Starter kit:

https://www.texasstandard.org/stories/atrocious-but-efficient-how-ranchers-used-barbed-wire-to-make-phone-calls/

https://www.atlasobscura.com/articles/barbed-wire-telephone-lines-homesteaders-prairie-america-history

There is also the Y2.5K bug. At the end of 1999, when finding YYMMJJ dates in external files, or even worse in packed format in fixed length files, the rule was:

if YY >= 50 then put 19 in front of it

if YY < 50 then put 20 in front of it

Should be OK until 2050, and that shit code won't be in use then, right?

What is a 'barbed wire phone system?' Is it a system which uses the barbed wire as the 'phone cord?' Do you need special wire? I want to hear all about this!

Sorry. I missed this.

Just like the original telegraph lines (in the US) going great distances, the concept needed one wire path with low resistance and used the planet as the ground / return path.

Back in the day (100+ years ago), the phone companies had no interest (or money) to run poles and wires to the 5 or 10 farms spaced miles apart between very small towns. So the farmers would make sure the top wire (typically barbed wire) had an insulated path from a real phone line to the last farmer on the path. That allowed them to have a shared party line. Which was/is/can be a total PITA, it is fantastic compared to nothing but a horse ride of 10 - 20 miles to get a message or doctor.

Now I have no idea how billing was handled. :)

I know someone at or nearing 70. When they would go visit his wife's family in Montana for the winter holidays, they would call ahead and say "we're leaving the city" so if they didn't arrive after 5 hours a search party would be sent out. This was before cell phones.

I was the system programmer lead on a system that originally had one bye for the month since 1970 and the second byte was the day of the month. In binary. Which led to a crisis in the mid 80s. Where we had to switch to 11 bits for the month and 5 bits for the day of the month. Had to play some games but we made it fit into the almost trivial memory (by today's standards) we had available to us.

When we were in Palermo a few weeks back, we were rather surprised that at least some of the restaurants by our hotel didn't open until 8:00pm.

When we were in Madrid a few years back, every restaurant we went to always sent over someone who spoke English. Or at least a bit of it. They said the locals never go out to eat before 8pm. So when we showed up around 6:30pm or 7:00pm we were instantly categorized as from the US or maybe UK.

Apparently we are know for not taking naps mid day.

Every meal we had there was great. Even the "lunch counter" type places I found during the day.

So I gather compact electric travel kettles are unknown in the USA?

How about thin on the ground?

Most US hotels have a small (2 cup, maybe more) coffee maker in each room which will make hot water. (And most also have a compact fridge in many if not all rooms.) At least all of the ones I've been in over the years. I don't know about Motel 6 or similar.

I'm in a circle of frequent travelers and know of non carrying around such.

You have to remember tea is not the default hot beverage here that it is there.

Yeah, I've met those "coffee makers". They're drip machines that make pale brown dishwater, not real coffee, and if you use them for water the result invariably tastes of foully bad coffee oil.

The in-room fridges are good except when they're stocked as a goddamn minibar full of useless crap (expensive sugar and nuts, even more expensive poor quality alcohol).

(And I've given up asking for tea in any US eatery. At best you get a lukewarm glass of water with a Lipton's yellow label teabag on the side -- Lipton's being flavourless floor sweepings, not tea.)

Maybe the math folks here can explain in plain language how you'd go about teaching an AI that 'equal' is not 'same' when it comes to humans.

You'd be better off asking the sci-fi authors, as they're the ones dealing with AIs capable of being taught. Current AI's don't understand anything: people call them "overgrown autocorrects" for a reason*.

(Yes, I will keep harping on this indefinitely. The dead horse flogging will continue until morale improves!)

*Two reasons, really: to be belittling and insulting, yes, but also to remind everyone that these aren't [insert your favourite computer character here].

The example I am most familiar with is cooking and recipe blogs. Now, if I want to cook a recipe (which happens most days) and I don't have or like a recipe in my cookbook collection, I Google it.

In the past few years the 'enshittification' of search results has become worse, as we all know. In the past year the situation has become vastly worse, the top dozen or more links are ads.... Mostly to ad plastered barely written drivel in endless paragraphs that, with enough scrolling past ads, eventually lead you to the actual recipe (usually in a handy little box).

Do you know why that is?
'“Google recently made the decision to surface recipes with long headnotes,” Tamarkin said. Which means, these days, the world’s most popular search engine is “favoring more and more blogs with long stories over recipe websites with short headnotes.”' (Article quoted is from 2019, to put that 'recently' into context)

To a large degree we are all creatures of our lives up to this point.

I don't drink coffee and rarely hot tea.

I did grow up on iced tea (back when soda came in bottles). My mother would boil a gallon or two of water and make it out of those tea bags. And add way too much sugar but it was in the fridge and was what you drank when you wanted something more than water. And as the saying goes "Not you cup of tea." I understand.

Mini bars (on the hotels I've been in) are on the way out. Or sit side by side with mini fridges.

I know the model you describe but I've either been staying in different hotels over the last few years or they are a vanishing thing.

Bellinghman [276] noted: "When we were in Palermo a few weeks back, we were rather surprised that at least some of the restaurants by our hotel didn't open until 8:00pm."

That's common in Italy and Sicily, since most people eat late by North American standards. (I believe that's also true of much of the Mediterranean, but can't say that from personal experience other than for southern France.) On the plus side, the good restos in tourist areas open earlier to attract tourists, so it's usually possible to get a reservation and great service if you book before 7 PM, which gives you an hour before the main local crowd arrives.

Pro tip: Many of the best small restos buy fresh supplies and only cook until they consume what they've bought, so if you wait too late, you can miss the window for a good selection of food. Talk to the resto staff early to learn their advice for timing and how far in advance to book a table.

In terms of portable coffee makers, Madame bought me a good-quality plastic/plexi portable French press. Since it doesn't require a disposable filter, all you need is boiled water and a bag of your own grounds (or a bag of homemade "cartridges" of plastic wrap filled with pre-measured grounds and sealed with masking tape to prevent loss of grounds) to make good-quality coffee pretty much anywhere. In my experience, airport restos or bars are often willing to give you boiled water.

Yes. A lot of institutions in the UK are little better, including quite a few hotels - I have several times found it better to make tea in my room than suffer the horror of a teapot with a Tetley's (or even Liptons's!) teabag in it, filled up with 80 degrees water out of a machine. But taking your own tea and milk is advisable, as you say.

You can make green tea with that temperature of water, but it's often not available, and I do not like Chamomile and Rhubarb, or whatever the concoctions they do provide as alternatives.

I agree with anonemouse!

It wouldn't be that hard, if they were taught like that, but (a) they aren't and (b) you merely replace that by the problem of teaching them what "equal but not the same" means. It's like many problems in programming (e.g. memory management) - there is an easy 'fix', but all it does is to replace a common, simple problem by slightly rarer, harder ones.

Thank you! That is a precise and complete explanation for why I have observed what I have (see #239). Even for those of us who don't usually use Google, the pernicious effects of its dominance affect us.

I stay in hotels with relative frequency and haven't seen a minibar in years. Maybe not as much of a thing here in Canada.

I use the in-room 'coffee' maker as a way to approach wakefulness, then go out and get real coffee when I can. I have thankfully not developed my palate to a point where I feel obliged to bring my own hot-drink preparation appliances with me when I travel.

Traveling as we do with one or more of our spawn, we tend to bring breakfast foods (cereal) to cut down on costs.

(And I've given up asking for tea in any US eatery. At best you get a lukewarm glass of water with a Lipton's yellow label teabag on the side -- Lipton's being flavourless floor sweepings, not tea.)

I mean, outside of more expensive places, you're not wrong. That was one thing I loved about Virgin America - because it was a Branson operation, they "got" tea, and being able to ask for milk, not cream, in my tea was a big deal.

I did grow up on iced tea (back when soda came in bottles). My mother would boil a gallon or two of water and make it out of those tea bags. And add way too much sugar

Sweet iced tea is a cultural thing in the US South.

Other places in the US you can order unsweetened tea and dump as much sugar from the table into it as you want. In the South, they'll serve unsweetened iced tea if you ask but peg you as Not From Here. Kinda a culinary shibboleth.

I've met those "coffee makers". They're drip machines that make pale brown dishwater

Ie. they make "American coffee".

You've probably also encountered American beer, which is charitably described as making love in a canoe.

And American cheese, which is apparently the another name for 'processed cheese food'.

Off topic, science fiction gets another prediction right...

Boing Boing reported that earlier this year, a dolphin with on both pectoral fins thumbs was spotted in the Gulf of Corinth. There's video of it too. ( https://boingboing.net/2023/12/13/incredible-dolphin-with-thumbs-spotted-in-the-ionian-sea-video.html )

The Onion predicted this would happen back in 2000. ( https://www.theonion.com/dolphins-evolve-opposable-thumbs-1819565718 )

Given current politics, I'm not sure whether to hope that the rest of The Onion article comes true or not...

Moz @ 267:

You sure that wasn't just because if you're desperate enough to eat at a truck stop you're desperate enough that anything is going to taste great? The best meals I've had have 90% been "I'm so hungry I could eat a horse" ... "that was amazing!". The main thing is never go back there when you're not hungry.

In my experience, it's pretty rare to find a restaurant - truck stop or otherwise - that serves good, tasty barbecued horse. Maybe in Mongolia.

Used to be, back before the interstates "standardized" everything, there were some truck stops that had a reputation for really good food. One way to tell was how many AUTOMOBILES were parked there in addition to the truckers.

Nowadays truck stops are all the same - automobile pumps out front, Truck pumps in the back, combination mini-mart (7-Eleven, etc) and some fast food (McDonald's, Burger King, Wendy's, ...) attached to the side.

And since they're always clustered around exits from the Interstate, you often have your choice of which "fast food" you want. You actually have to search to find bad food at a truck stop.

The room coffee makers - it's not just the machine, it's the really, really crappy coffee packets they give you.

Howard NYC @ 270:

surviving travel 101

ask the staff of a hotel/motel where the nearest edible food is, if in walking distance; otherwise phone # for pizza;

"..."

when I was travelling on business 30 Y, there was a bunch of us trapped in a shithole hotel in a minor city where the local restaurants closed at 8:00PM because locals went to sleep at 9:00!

Back when I traveled extensively for the Fire/Burglar Alarm Company, I covered a 40,000 sq mile territory, averaging 1,000 miles/week on the road servicing customers - a little over half a standard work week was just driving and usually three nights in a motel.

It didn't take long to figure out where all the good places to stay & the good places to eat were located throughout the territory

Including where to find food if I needed to work overtime in some podunk little town miles from nowhere (towns so far out in the country it was an hour's drive just to get close enough to a city to be in the boondocks).

During that time I was in the National Guard, and we got MREs whenever we had a weekend in the field. I always kept a few of the extra ones in my van "just in case".

Charlie Stross @ 277:

So I gather compact electric travel kettles are unknown in the USA?

Every U.S. motel I've stayed in this century has a 4-cup drip coffee maker** (Mr Coffee or equivalent) in the room. If you just fill it with water and don't put coffee in the filter holder, you can get hot water out of it

I usually stay in the motels that have a "complimentary breakfast" (TANSTAAFL) & get the regular coffee from the lobby. 😏

And lately, most motel rooms have a microwave oven with enough power to boil a cup of water. All you need is a microwave safe cup.

** I think y'all call it filter coffee. Took me a while to figure that out when I visited Scotland in 2004. But once I did I was able to order coffee I prefer to drink. All of the Bed 'n Breakfast places I stayed had the Bunn-o-Matic coffee makers most of the U.S. restaurants use.

Every U.S. motel I've stayed in this century has a 4-cup drip coffee maker (Mr Coffee or equivalent) in the room.

That will still be dependent on the quality of the coffee, and I've experienced a "Starbuck's burn out roast".

I have been in a truck stop that served camel. I didn't ask whether it was roadkill or whether they just kept an eye out and a gun handy. But it was definitely on the menu.

But I don't think I've seen horse. And I definitely couldn't eat a whole camel.

(camels are feral in Australia and I don't think there are any hunting restrictions on them other than generic "permission of the landowner" type ones)

Charlie Stross 277:

So I gather compact electric travel kettles are unknown in the USA?

most hotel/motel rooms have microwaves... worst case I can recall there was a single microwave for entire floor next to ice machine... now visualize sixty-plus people from forty rooms lining up to heat up food and/or baby formula... as well in line to get ice from an undersized machine that could not keep up with demand... we went to other floors to stage ice raids...

my tips were focused upon minimizing weight penalty, maximizing calories and ensuring basic survival whilst traveling amongst barbarians in savage, uncivilized locales... and lessons learned about ugly shit of how oft hotels/motels seem determined to inflict as much misery upon unwary travelers as feasible... not to mention recent horror stories about being trapped in an airport due to various 'n sundry SNAFUs by airlines with all the restaurants closed with hundreds of footsore, hungry, stranded passengers eyeing one another evaluating whomever has the most body fat and therefore ought be tossed into the stew pot first...

heh... yo! Netflix! How about an updated version of the infamous Donner Settlers Disaster could be set at O'Hare Airport in Chicago or maybe LAX or best of all Newark Intl... during a climate change shitstorm dumping three feet of snow in late April... with arguments breaking out amongst stranded passengers of how best to season slow roasted 'long pig' ...FRACK them elitist foodies(!)

:)

{ CUE EVIL LAUGHTER }

Worst of all: big hotels a long way from local coffee shops and restaurants with a goddamn Starbucks in the lobby.

Starbucks can't roast coffee beans without scorching them.

(They can serve an acceptable cup of tea if you ask for a grande with two tea bags, and an inch of air space at the top to fill with milk. Then let it stew for a while. They use boiling water and pour it over the bags, which is the least-bad way to make tea with tea bags.)

I think my car key fob has way more memory that the machines I used to run Fourier transforms on.

And on the subject of coffee. I spent about 9 months in the US - in 1 week aliquots - over the years and rarely found nice coffee, but thats pretty much true here too. Its the exception, not the rule.

I use a cafétiere at home and grind beans by hand. It takes less time than the kettle takes to boil. But when I go away I use coffee bags. There’s a good choice of these in UK supermarkets and the coffee is usually good. Before these were available I used one cup coffee filter. All much better than instant.

Re: "...we're leaving the city" this is still what happens with people travelling to or between our remote clinics. They notify someone at each end that they're leaving and do the same when they arrive. They also fill in an online form which notifies someone if they're an hour overdue AND work cars are monitored 24/7, although not in real time. This last fact, sadly, solved a murder.

Netflix! How about an updated version of the infamous Donner Settlers Disaster could be set at O'Hare Airport in Chicago or maybe LAX or best of all Newark Intl... during a climate change shitstorm dumping three feet of snow in late April

John Varley wrote a story that's similar to your idea: "The Flying Dutchman". Main character has a disturbing multi-legged flight from Chicago to Atlanta to New York, etc.

{ CUE EVIL LAUGHTER }

Yes, yes there was. (Not by the MC, however).

Kardashev @ 296:

"I did grow up on iced tea (back when soda came in bottles). My mother would boil a gallon or two of water and make it out of those tea bags. And add way too much sugar"

Sweet iced tea is a cultural thing in the US South.

Other places in the US you can order unsweetened tea and dump as much sugar from the table into it as you want. In the South, they'll serve unsweetened iced tea if you ask but peg you as Not From Here. Kinda a culinary shibboleth.

Most places I know in the south (U.S.) have got the message by now. If you don't ask, you'll get sweet tea, but if you DO ask, they don't even look at you funny anymore.

paws4thot @ 303:

"Every U.S. motel I've stayed in this century has a 4-cup drip coffee maker (Mr Coffee or equivalent) in the room."

That will still be dependent on the quality of the coffee, and I've experienced a "Starbuck's burn out roast".

Yeah, I guess.

I've never actually used the in-room coffee maker to make coffee. I've only used it to get "hot" water (for herbal tea). Never noticed what Charlie describes as coffee oil flavoring in the hot water.

But that may be my taste buds destroyed by years of drinking "army coffee".

For actual coffee when staying at a motel, I'll put my clothes on and go down to wherever they serve the complimentary breakfast (with my empty McDonald's Large coffee cup from the day before - you can reuse them several times before they start to break down.) ... and get regular coffee, to carry back to the room while I get ready to load the car.

And now that they have microwaves in all the rooms, I can get hot, almost boiling, water that way as well.

Rocketpjs@294 writes: "I use the in-room 'coffee' maker as a way to approach wakefulness, then go out and get real coffee when I can. "

That's harder than it used to be. Arby's quit selling coffee entirely, Burger King only has it breakfast hours, and Culver's doesn't keep the pot going but will make a fresh pot of really outstanding quality if you ask. But then their system has no way of getting it brought out on the tray with the rest of your order so it can get complicated if there's a line, happened to me multiple times at several of their outlets so it's definitely a systemic oversight. Starbucks proves there's a demand, now I wonder if they quietly exert pressure on whatever real estate companies owning the land, parking lots and access roads where their units are located on to discourage other tenants from selling coffee. Pressure like maybe agreeing to higher monthly rent? I guess maybe that could be legal....

Off topic, but probably important to SFF writers who are using genetic engineering to create new people and species. The science just got massively more complex.

Worldbuilders need to read this: https://www.quantamagazine.org/new-cell-atlases-reveal-untold-variety-in-the-brain-and-beyond-20231213/

The basic points:

-We're entering an era of "cell atlases", things like an atlas of every cell in a mouse brain, tagged by expressed genotype, which was published a few months ago. The technology now exists to do full expressed genotype of tens of thousands to millions of individual cells in a single experiment.

-Turns out there are more than 5,300 distinct types of cells in a mouse brain. They're working on an atlas of the human brain (guesstimate 5,000-10,000 types of cells, 3,300 already identified), and comparable cell diversity is showing up in pancreas, liver, and placental cell atlases.

-The important quote for SFF writers: "The differences in our cells don’t come from differences in the genes themselves: Every cell in your body contains a copy of your distinctive DNA and the 20,000 genes for making proteins it encodes. 'The genome is a parts list,' Quake said. 'There’s no way to predict which cell types come out of a given genome.'"

In other words, a genome isn't a script for an organism, it's akin to a code library. What gets called into use in a particular cell depends on circumstances, and fiddling with the code in the library doesn't necessarily do much.

There's a lot of useful detail in the article, which I really encourage people to read.

"the new Science papers, also found that most of the cell types were located in evolutionarily ancient parts of the brain."

Sounds similar to the way greater genetic diversity within human populations reveals which groups have been around the longest. Nature must be inclined towards simplification over time, at least in certain cases. Makes sense, why work harder than necessary?

Sounds similar to the way greater genetic diversity within human populations reveals which groups have been around the longest. Nature must be inclined towards simplification over time, at least in certain cases. Makes sense, why work harder than necessary?

My impression is that it might be simply that the longer something exists, the more it gets fiddled with. This process isn't limited to evolution. People's houses diversify much the same way, I think.

The neocortex is a comparatively new sub-organ within the brain. It has its own cell types, but not a huge diversity of them. If you believe this other recent article, it's also significantly more energy hungry than the older brain regions. There might be a correlation between complexity and efficiency.

https://theconversation.com/human-intelligence-how-cognitive-circuitry-rather-than-brain-size-drove-its-evolution-219669

greater genetic diversity within human populations reveals which groups have been around the longest

My loose understanding is that there's a whole lot of shrugging going on in that area right now. From discussions about when people got to the Americas (White Sands, Population Y - Stefan Milo has a series of videos on this) through to the question of what exactly are people (Florensis, Neanderthal etc) all helping to obscure "when people". Australia is often past the later bounds on that question which makes for some fun thinking - if "modern humans" arose 50kya and Australia was first populated 60kya, who exactly populated Australia? Did "modern humans" arise at least twice? America is a bigger island with more people, but there are similar questions there. And even some bits of Asia (ie, everything from the Sinai to the Atlantic (both ways!)) have weird ancient humans and human remains.

This gets back to genetic mapping, where the more we look at human genetic variation the more questions we have.

{shrug} 😋

I've seen something different the last couple of years, a machine* that takes only something that appears to be a cross between a pod and a teabag and the water mostly boils away in the bottom of the machine, two or three cycles will deliver a cup of something that somewhat resembles filter coffee, but not really.

*Seen in two southern Missouri hotels.

I used to make iced tea that way, lately I use a drip coffeemaker (That's never seen coffee) two teabags, pour into a two quart pitcher (Almost two liters) with a half cup of sugar, stir and allow to cool.

An off-topic favour to ask the groupmind…

My mother supposedly has free international calling added on her cell phone. I got it set up for her while I was visiting but didn't test it because she got all dithery about just calling her relatives in England when they weren't expecting it… and now that she has tried it (and had problems) I'm not there to see what she's doing, or hear any error messages/tones.

Is there the equivalent of a talking clock or something I can get her to phone to test it out without worrying about waking people up? She is currently only making calls on very restricted times and days which makes diagnosing what the issue is tricky.

From what I can tell UK phone numbers are a two digit exchange and an eight digit number for landlines, or a ten digit number for cell phones. Is this correct? And the UK country code is 44, so she should be dialing 011-44-xx-nnnn-nnnn, is this correct? Or 011-44-nnnnn-nnnnn for a cell phone?

I'm not convinced the problem isn't with her cell phone provider. She apparently spent an hour on the phone with "Eddie" who had a very thick Indian accent who sent her a link to a WikiHow article on how to call London and change sim cards and use wifi to do a video chat which just confused her. (Not difficult, she's nearly 90 and very set in her ways now.)

It's also possible that she's dialling wrong on her iPhone — I've had way too many 'I touched something red on the screen and then something else appeared and what was it and why did that happen, and oh there's a jay bullying the sparrows again…' phone calls not to discount that possibility. When she bought the cell phone she paid extra for a help service at the shop but she says they keep telling her she has to talk to Koodo directly. If I have a number that she can call to get a recording then she can ask the chap at the shop to test it to verify if it's the cell phone that's the problem, or if she's using it wrong.

Most recent systems have versions of their time related system calls that use a 64 bit number. The more salient problem is that lots of filesystems only use a 32 bit number, and this is harder to change. Linux's ext4 found a couple of extra bits somewhere so it won't expire until 2446.

Lets start with an arbitrary illustrative UK (based on landline; cell isn't effectively different, since it's differentiated by digits 2 to 5 being in certain groups) number 01234 567890, giving UK exchange 01234 and number within that exchange 567890. Large conurbations may visually differ, but don't effectively differ if you're phoning from a cell or from overseas.

The number I'd quote you for calling from overseas is +44 1234 567890, so the change you make to the UK number is to drop the leading 0 and replace it with the UK country code of 44.
The + sign should be replaced with your international access code, which I don't actually know.

I use Koodo and live a dozen kilometers away from your mother. If needed and it wouldn't be too upsetting to have a large bearded stranger come and try to help you can drop me an email and we can set it up. NOTE: If it is an iphone I might bring my teenager as I don't use them and would provide dreadful tech support.

But I don't think I've seen horse

If you want to taste horse meat, go to France where it is available in butcher shops and some restaurants. Horse meat is almost fat-free and better than beef for some recipes.

I've seen American tourists order "steak tartare" and being horrified when presented with raw meat and DIY seasonings. They were unable to eat it and that was before anybody told them that it was horse meat.

Um. Not "just". Such articles lag the science by some considerable time, occasionally decades. For at least four decades, every time such scientists have thought they have become to understand something, they have discovered that completely new complexities have become visible, and most of which they thought they knew is, in fact, incomplete and often misleading or even just plain wrong. I wasn't aware of how many cell types there are, but I am sure that was just my ignorance :-)

Another aspect that scientists are aware of, but gets very little attention in the press, is that such things change over time, and it is not even true that post-maturity ageing is solely a gradual failure mechanism. It's an inherent part of our system - that's one in the eye for the immortality campaigners! Plus, of course, they are likely to change with use and external context but, as far as I know, very little work has been done on that. So even the organism's own cells evolve as time goes on ....

Your second-last paragraph is a good analogy!

There is evidence from humans. The Victorian craniologists were wrong - there is essentially no correlation between brain size and intelligence in humans.

Most people (including many scientists) have extreme difficulty thinking statistically, however good they are at applying its techniques. When populations (of any organism) change, they first evolve a spectrum of forms, sometimes separate into races, and only later do those races evolve into separate (sub-)species. So there is really no contradiction. My understanding is that modern humans arose over the period 150,000-50,000 years ago, but I may well be out of date, mistaken, or both.

Yesterday's XKCD 2867 was well-timed.

See also: Falsehoods programmers believe about time

The thing that is going to cause trouble is the number of important (and even critical) applications and interfaces that use timestamps in places where the maintainers (if any) don't even know they do. That is confounded by C's absolutely insane rules for converting 32-bit integers to 64-bit.

Steak tartare is usually beef, though. I ordered a horsemeat steak once, to try it, but didn't like it. As you say, it's better in many respects, but it is also slightly sweet.

I am pretty sure #3: "Any 24-hour period will always begin and end in the same day (or week, or month)"

is a typo for "Any 24-hour period will always begin and end in the same week (or month)"

It should be blatantly obvious that a 24-hour period normally DOES NOT begin and end in the same day. OTOH it is surprisingly tricky to identify all the exceptions when it does.

Yeah. I did a deep-dive into calendar systems once.

Horse is good. Tastes like donkey.

Thanks. I may take you up on that. It sounds like the number she tried calling was in the correct format. Now I just need a test number that can be called anytime without waking someone up.

If you want to taste horse meat, go to France where it is available in butcher shops and some restaurants. Horse meat is almost fat-free and better than beef for some recipes.

In the US most canned dog food is made from horse meat. And said knowledge causes a big case of the vapors in many pet owners about every 10 to 15 years. After everyone gets upset it takes a decade or so before a new group of pet owners grow up and become aware.

Robert Prior [297] noted: "You've probably also encountered American beer, which is charitably described as making love in a canoe."

As the ancient Monty Python joke goes, "because both are fucking near water". But that's badly outdated info, and even back when the joke was new, it applied only to mass-market brews. There's been a renaissance of American microbreweries that produce amazing stuff; in fact, it's pretty easy to find a restaurant or pub with a good selection of locally sourced microbrrews, and most are happy to give you a taste if you promise you'll buy a pint of something. Indeed, many offer "flights" (4 to 6 tasters) so you can sample several brews.

Charlie noted [306]: "Starbucks can't roast coffee beans without scorching them."

That's out of date info. At least in some stores, they offer a range of light to medium roasts. Lots of people really like dark roasts. It's a question of what the customer wants and what the roaster was trying to achieve: dark roasts prioritize the roast, and are often (not always) used to conceal low-quality coffee, whereas light roasts prioritize the varietal's flavors, and are usually chosen for varietals with distinctive desirable characteristics. There are intermediate roasts or blends that mix the two characteristics. I'm not a coffee snob (well... maybe a little), but I have found a few varietals that I really like as light to medium roasts. You might love them or hate them, and that's fine. À chacun son goût!

Just a thought, are the free calls restricted to geographic/landline numbers only or do they definitely allow calls to mobiles as well? Landline numbers start 01 or 02, mobiles start 07 and are generally more expensive to call. Any recorded announcement services accessible from overseas are likely to be 08 or 09 prefixes which are premium rate (although for 0800 the rate is zero).

I got it set up for her while I was visiting but didn't test it because she got all dithery about just calling her relatives in England when they weren't expecting it

Ah, yes. The joys of doing tech for elderly parents.

ADMINISTRATIVE NOTE

New blog topic is up. That is all.

Robert Prior  334:

automated non-human voice robot such as what I endure with my cable vendor's tech support... available 24/7 and near-impossible to reach an actual human... ("to commit suicide from despair after waiting 17 hours press [5]")... there's one in every nation but be assured the singular most awful is Spectrum (formerly Warner) in NYC...

Elderly Cynic 329:

further complicating dates & times are minor [ snark ] things such as which time zone is baseline, daylight savings status, stored as AM/PM v. 24, et al... converting the format of the raw data is only one of multiple steps during migrating... every place I ever worked, nobody would agree (for whim or politics) to agree upon an enterprise-wide standard for handling dates & times and explicitly defining all data types and calculations with detailed rules written in unambiguous English... so any wonder there's a bunch of looming icebergs? across entire industries?

Keithmasterson @ 313:

Rocketpjs@294 writes: "I use the in-room 'coffee' maker as a way to approach wakefulness, then go out and get real coffee when I can. "

That's harder than it used to be. Arby's quit selling coffee entirely, Burger King only has it breakfast hours, and Culver's doesn't keep the pot going but will make a fresh pot of really outstanding quality if you ask. But then their system has no way of getting it brought out on the tray with the rest of your order so it can get complicated if there's a line, happened to me multiple times at several of their outlets so it's definitely a systemic oversight. Starbucks proves there's a demand, now I wonder if they quietly exert pressure on whatever real estate companies owning the land, parking lots and access roads where their units are located on to discourage other tenants from selling coffee. Pressure like maybe agreeing to higher monthly rent? I guess maybe that could be legal....

It's NOT legal, but whenever has that stopped a corporation from doing something. OTOH, I don't think Starbucks would have the might to stop a developer from leasing to some semi-competitor ... especially in a shopping center food court.

Starbucks COFFEE* is not fit to drink. Starbucks burns the beans before grinding them and then burns the coffee when they're making it. That's because Starbucks doesn't really sell coffee, they sell coffee flavored drinks and you have to over cook the base liquid to concentrate it enough so some flavor remains after all the adulterants are added. And all of Starbucks imitators** do the same.

Luckily McDonald's serves coffee all day (even if the managers can't find it on that stupid kiosk either), and there's always those 24-hour truck stops for worst case scenario.

*By COFFEE, I mean coffee - "the bitter brew" itself - no cream, no sugar, no frothy foam for making teddy bear faces ... nothing in it but COFFEE. I don't drink it for the taste, I drink it for the effect; although I have come to "like" the taste after so many years of drinking it that way.

I guess like Professor 'enry 'iggins, I've grown accustomed to the taste ... second nature to me now.

**Maybe not "imitators", coffee shops serving coffee the same way they do where Starbucks came from ... which I think is Seattle, WA?

PS: I stopped at an Arby's (attached to a gas station) on my recent trip out west ... IIRC it was in New Mexico ... and they were still selling coffee.

Keithmasterson @ 315:

"the new Science papers, also found that most of the cell types were located in evolutionarily ancient parts of the brain."

Sounds similar to the way greater genetic diversity within human populations reveals which groups have been around the longest. Nature must be inclined towards simplification over time, at least in certain cases. Makes sense, why work harder than necessary?

One thing to keep in mind in this context is the "Science" is there to serve the needs of the "Fiction". The "Science" doesn't have to be absolutely accurate; it just needs to be plausible enough to support the story ...

Tim H. @ 319:

I used to make iced tea that way, lately I use a drip coffeemaker (That's never seen coffee) two teabags, pour into a two quart pitcher (Almost two liters) with a half cup of sugar, stir and allow to cool.

My Mom had a tea-maker that looked exactly like a drip coffee maker. It had a basket that would hold a regular "coffee" filter where you could put loose tea OR you could leave the filter out & just hang teabags in the carafe.

I think it was just a re-badged "Mr. Coffee". But the NAME said it was for tea.

Beer in America: warning, danger, danger, Will Robinson!"

America, the most overdone nation in history. The micro (and macro) brews can be good, or... IPA is everywhere (watch me try to find a porter or a brown ale), and the brewers' enTHUsiasm - if hops are good, more hops is better, and how about 4x the hops?!

I drink Killian's at home (it's cheaper than Smithwicks, and cheaper than almost everything else. They stopped importing Bass Ale, so we get Becks' for Ellen, which I don't care for. I just don't want to pay $12 and up for a six-pack.

Those problems have been with us since the advent of communications that specify dates and times (which predate computers and even the railways by many millennia (*)). While there will still be programs and organisations that use them for timestamps, even after all our experience that that is a stupid idea, it is not possible to stop a fool in his folly. Unix/POSIX timestamps are in UTC and the only such issue is leap seconds; that's a serious issue, too, but for a relatively small number of uses.

The 2038 problem is entirely different in nature, and is very serious indeed, for the reasons I mentioned. It almost certainly is coded into a good many embedded systems (probably including firmware and possibly even hardware), too. The year 2000 problem may have been similar, but was much, much less serious.

(*) As archaeologists will tell you. When the hell was "the third hour in the fifty-fourth day of the reign of Garble the third"?

The Swedish speaking clock should be reachable from foreign parts: +46 8 90510

Hi Robert

If you want a UK based mobile to call email me at johnramsaygrieve(at)outlook(dot)com You will not disturb me and will get voicemail if I don’t answer.

whitroth [344] opined: "Beer in America: warning, danger, danger, Will Robinson!"

I'm guessing it depends on what's going on in your local microbrewery scene. Having a brain fart, but I think it's the River Street Brewery (in Milwaukee? Cleveland?) that coined the phrase "ale's what cures ya". Decent beers too in a place you'd least expect it.

whitroth: "IPA is everywhere (watch me try to find a porter or a brown ale), and the brewers' enTHUsiasm - if hops are good, more hops is better, and how about 4x the hops?!"

It's easy to overdo an IPA, but the better brewers list the bitterness so you know what to avoid. I tend to top out at about 30 IBU, but acquired an unholy love of Arrogant Bastard Ale (https://www.stonebrewing.com/beer/arrogant-bastard-ale), which is much, much hoppier. You have to love it if only for the manifesto (scroll down; it's worth the time).

Porters and browns, including various forms of "Scotch Ale", usually aren't too hard to find; most brewers do one nowadays, though some cheat with "ambers" that are only halfway to brown ale. (Here in Kingston, I enjoy "Barking Squirrel", which is an amber.) If you're ever in Minneapolis, try "Moose Drool". (Insert salivating noises.) Berkshire Brewing (Massachusetts) used to have a few nice dark ales, as did Southern Tier. If you want a real treat, look for one of the cask aged brown ales. Back in my hometown of Montreal, Brasserie Le Castor had a wee-heavy beer aged in bourbon or rum casks that was to die for; at 11% ABV, you probably want to share a pint rather than consume it yourself. Or maybe you're made of sterner stuff than I am.

whitroth: "I drink Killian's at home"

Will keep an eye open. Don't think I've seen it locally.

Right, but all the micre/macro brews are running $13 and up per six pack. Since we usualy have a beer at dinner, no.

Oh, and the trick is finding the bottles of other at store that sells it, if you're not going to a brewery.

"Where you least expect it?" Milwaukee? And I really miss Goose Island's Hexnut brown ale (yes, that was the name). By the time I got back to Chicago in '06, Bud had bought 20% of them, and they no longer made it. For a while, they had brown ale, but try to find it at a distributor. Go ahead, I'll write another novel while I'm waiting. Then that vanished.

Killian's is an Irish beer, yes, imported.

Killian's is an Irish beer, yes, imported.

George Killian's Irish Red? I think you'll find it was originally brewed by a now long-defunct Irish brewery, and is now brewed under licence by MillerCoors, probably in Milwaukee.

Decent brown ales are somewhat scant in the US, but it's hard to find a town without a decent local amber ale or dark lager, I find. Good beer is food, and doesn't travel that well. My motto is, drink local.

I'm out of date, myself. Apparently MillerCoors is now part of Molson Coors!

this...? this is just one very funny set of promises waiting to be broken

https://lite.cnn.com/2023/12/15/business/southwest-ceo-christmas-travel/index.html

"Southwest CEO promises not to ruin the holidays with another travel meltdown"

...next thing ya know he'll promise enough leg room to men over 5'10" & 180 LB to avoid muscle spasms

America, the most overdone nation in history. The micro (and macro) brews can be good, or... IPA is everywhere (watch me try to find a porter or a brown ale), and the brewers' enTHUsiasm - if hops are good, more hops is better, and how about 4x the hops?!

Yup. Over-hopped beers are another San Diego Special.* I gave up on them long ago.

While I used to like IPA, triple-hopped IPA isn't something that makes anything else taste good. That includes the next sip of that beer.

Decades ago, where I had a bunch of good beers was at Humboldt State and in the surrounding town. About a quarter of the student population were reputedly home brewers, and some really were good at it.

Heteromeles @ 353:

"America, the most overdone nation in history. The micro (and macro) brews can be good, or... IPA is everywhere (watch me try to find a porter or a brown ale), and the brewers' enTHUsiasm - if hops are good, more hops is better, and how about 4x the hops?!"

I used to like dark beer back when I was in college, but that was ages ago before they raised the drinking age. I don't know (remember?) anything anything about it other than it was dark brown, coffee colored.

Funny thing, THEY lowered the voting age & raised the drinking age and neither one of them affected me ... I was already over 21 in both cases.

America, the most overdone nation in history. The micro (and macro) brews can be good, or... IPA is everywhere (watch me try to find a porter or a brown ale), and the brewers' enTHUsiasm - if hops are good, more hops is better, and how about 4x the hops?!

While I used to like IPA, triple-hopped IPA isn't something that makes anything else taste good. That includes the next sip of that beer.

Decades ago, where I had a bunch of good beers was at Humboldt State and in the surrounding town. About a quarter of the student population were reputedly home brewers, and some really were good at it.

At Ft. Hood in 1991 before we were demobilized & came home from Desert Storm I bought a six-pack** of beer from the PX; thought we were going to be there a while, so I figured I'd keep some in the little fridge (the barracks we were assigned to were very similar to a college dorm ... four double rooms sharing a short hall & a common bathroom).

I'm not really a beer drinker (I prefer spirituous liquors), but I wanted to be hospitable. I don't know if you can get "dark beer" in a can but I know you couldn't get it from the PX at Ft Hood in 1991.

As it turned out, we were out of there within a couple days & the beer went into my duffel bag and sat in my refrigerator at home for years after. I'm pretty sure there were two or three still in there when I got home from Iraq. I'm not sure who drank them or when.

Don't really care, but it's amusing to think how nasty they must have been after sitting in the back of the fridge for more than a decade, 'cause I know it wasn't good beer, just whatever was "ON SALE" at the PX that day.

I have bought a couple of cans of Budweiser since then. That's good for marinating hot-dogs.

**Might have been a twelve pack?

Falsehoods programmers believe about time

My two favourite time facts remain "there are approximately 50 hours in every day" and "on average a 2:30am alarm goes off once a day".

Both of those annoy the snot out of programmers on the regular but are nigh on incomprehensible even to many other programmers, let alone the people who struggle to understand why the international date line is more of a drunkards walk (at times it's three different dates!).

I regularly post a link to "what time is this where I am" websites when people schedule online meetings, even if I have to guess the "10:30am tomorrow" timezone. There are lots, I just search and use the first vaguely plausible result. It just saaves some of the "what time is that for me" responses, and duplicate searches for people that don't have a bookmark.

I have the mixed good fortune of living in a 'tourist' town that has a number of excellent breweries, cideries and distilleries in operation, one of which is a short walk from my house (and 2 more of which are a longer walk). Some of their beers are truly excellent and some are not, but the types and options are always evolving.

During the pandemic I was always happy to support my literal neighbour when I bought beer, rather than Molson Coors.

»I regularly post a link to "what time is this where I am" websites when people schedule online meetings, even if I have to guess the "10:30am tomorrow" timezone.«

The one thing almost nobody groks, including effectively all programmers, is that timestamps by definition are political constructs.

Even UTC, or depending on your stance on leap-seconds: In particular UTC.

Any input representing a time in the future, should always be stored uninterpreted, /precisely/ as it was specified, because there is no way to know what the correct legal interpretation of that specification will be, until shortly before that time.

If we both schedule a meeting for "10:30, June 1st, 2024, Milano", EU and Italy may, or may not, have DST in effect by then, that depends entirely on politicians.

Leaving that string as is will have our calendar programs DTRT, no matter what the politicians do, whereas converting it to "safe UTC" is a tacit assumption that politicians will not do something silly between then and now.

Another interesting corner case where the time specified is 02:30 some spring night between Saturday and Sunday, and then politicians decide to introduce DST precisely that night. Only when you become aware of the new law, will you discover that you have an invalid time specification on your hands. In the autumn you have the opposite situation: The timestamp became ambiguous.

" I used to be big in tabletop roleplaying games (notably AD&D) in my teens but haven't played in 40 years."

Slightly off-topic, but - I'm sad to hear that.

I'm one of those who expects that "A Conventional Boy" will be about my teens years (maybe with more corpses than I encountered), but 40 years on I'm still playing. Table-top gaming is still fun. It still has both the silly zaniness, and the pathos, and the inevitable evolution of in-jokes in any campaign that lasts more than a month, and... well, it's still good.

"There's no clear-cut delineation between truth and lies; rather, there's a sliding scale of bullshit."

Oh, I'm still waiting for LLMs to really get going in the areas of the economy where they will make the most impact.

Like many new technologies people are still figuring out quite what to do with AI, how to best make use of it. Eventually they'll realise what it's really good at:

Scams and spam.

The ability to bullshit - to provide the surface queues that signal that what they are saying should be trusted - is the LLMs true ability. Unscrupulous people who stop trying to work on getting the LLMs to "tell the truth" and focus on trying to get them to "produce convincing bullshit" are going to make a lot of money.

In what way are TAI timestamps political? Parochial, I will give you (i.e. what's special about the local frame, seconds or their starting point?) But political?

Malware writers have been exploiting timestamps for a while now. They're great for adding to the attribution problems. Just pretend your malware development computer is in the timezone of $STATE, release your malware, then watch the infosec community go wild with their false attributions.

Nevermind ways of the system time. The W32.StuxnetDossier has a few examples.

Stuxnet will only attempt to use MS10-061 if the current date is before June 1, 2011.

Also:

Stuxnet will verify the following conditions before exploiting MS08-67:
• The current date must be before January 1, 2030
• Antivirus definitions for a variety of antivirus products dated before January 1, 2009
• Kernel32.dll and Netapi32.dll timestamps after October 12, 2008 (before patch day)

MS10-061 is described by Microsoft as "Vulnerability in Print Spooler Service Could Allow Remote Code Execution", while MS08-67 is "Vulnerability in Server Service Could Allow Remote Code Execution", so happy fun time.

whitroth [349] noted: "Where you least expect it?" Milwaukee?

My bad. I meant that Milwaukee's the home of monopolistic mass-market beer (Miller, Schlitz, Pabst), which tends to crowd out craft beers.

I get where you're coming from about the price of good beer. Hereabouts, it's just about doubled in the past year. But it's worth it imho. Life's too short to drink bad beer.

But political?

The time used around you for your daily interactions with others around you at any point on the planet is decided by politicians. Mostly.

You're reaching. That's true, but it doesn't make TAI or TAI timestamps political. Even UTC wasn't really political, originally, but it assuredly is now.

I meant that Milwaukee's the home of monopolistic mass-market beer (Miller, Schlitz, Pabst), which tends to crowd out craft beers.

Yes and no. It also means there's a long strand of brewing professionals who like to make good beer. About 20 years ago Miller started a craft beer division, which made some excellent brews for about five years before they decided it wasn't relevant.

I have a cousin in Milwaukee who owns part of a Milwaukee brewery called Sprecher, which makes some great beers. Not much distribution outside of Wisconsin, sadly. The original brewmaster retired to San Luis Obispo, in California, and has set up special distribution to some of his favorite bars there.

»In what way are TAI timestamps political?«

Hehe, good question :-)

The scientist are very adamant that TAI and EAL are in fact NOT timescales, but merely numbered sequences of SI-seconds, and that NOBODY should be using them for anything, and in PARTICULAR not programmers or astronomers.

This is because having seen MJD, UT and UTC become f*ked up by politics in the past, the True Time-lords on this planet are doing everything in their power to prevent it happening to EAL and TAI.

Astronomers find that very political.

»Even UTC wasn't really political, originally, but it assuredly is now.«

You have no idea how wrong you are about that :-)

I suggest you read the minutes from the International Meridian Conference, which took all of October 1884.

Re: '... not there to see what she's doing, or hear any error messages/tones.'

Suggestion from a non-techie about a techie issue, i.e., I won't get upset if you ignore this. :)

Does your mother have access to a laptop - is she able to connect to the Internet? If yes, you can try setting up a Zoom meeting and have her show you in real-time exactly how she's using her iPhone, hear/see any error messages, etc.

Re: '... human brain (guesstimate 5,000-10,000 types of cells, 3,300 already identified)'

Laughed when I read this because a few years ago a commenter* on this blog corrected my outdated info about the number of different types of neurons: I had said a few, the commenter said over 300. So, at least a ten-fold increase in the number of different cells in one organ. Amazing!

*Apologies: I don't remember who the commenter was.

Next up I suppose is figuring out how many different glia cells there are. Glia outnumber neurons -- by a lot.

The closer we look, the more complex organisms get. I tried to find which organs have the greatest diversity of cell types - no luck (yet) but did find this interesting cell map below.

https://humancelltreemap.mis.mpg.de/

Re: Falsehoods programmers ... time

Clocks in different systems will always run at the same rate/speed. We tried to send timestamped data back and forth between systems - total mess!

»Clocks in different systems will always run at the same rate/speed. We tried to send timestamped data back and forth between systems - total mess! «

It can be done.

One recent project of mine was the prototype for the COTS computer cluster which is going to bend the mirrors in the ESO ELT telescope 500, preferably 700 times a second, with jitter in low microseconds.

We did that, with webshop DELL servers and an unmodified CentOS kernel.

Bug once you get down there, everything floats.

The quartz crystals in modern servers have a tempco around 1PPM/K, which means that you can clearly see the air-con cycling on and off about three times an hour and even estimate the longitude of the server by the phase of the 24 hour oscillation.

Stephen Murdochs "hot or not" from around 2007 is worth a trip to the library.

nightmare fuel...

"More than the risk conferred by any individual variant, it is the rapid rate of virus evolution that is most concerning to Trevor Bedford, a professor in the vaccine and infectious disease division at the Fred Hutchinson Cancer Center. “No single variant has been that impactful,” he said, “but the overall accumulation of these mutations is having significant impact.” ..."

from

"What to Know About the New Covid Variants"

https://archive.ph/5gugs

or

https://www.nytimes.com/article/covid-variant.html

The closer we look, the more complex organisms get. I tried to find which organs have the greatest diversity of cell types - no luck (yet) but did find this interesting cell map below.

My bet is on the gut, whether or not you count all the non-human cells. Guts have neural cells, muscle cells, immune cells, and GI cells.

In Science Advances yesterday: https://www.science.org/doi/10.1126/sciadv.adi7632

"An energy costly architecture of neuromodulators for human brain evolution and cognition"

Quote from discussion: "Recent research has suggested that the energetic costs of signaling increase during long-term processes such as memory formation (39). Our human data indicate that brain regions with high energetic costs of metabotropic signaling play an important role in cognitive processing over longer timescales, including memory processing, cognitive inhibition, and reading. Additionally, neuromodulators have been linked to cognitive dysfunctions associated with major mental disorders (40). Unfortunately, the efficacy of current psychoactive drugs in regulating neuromodulators is limited, and further research is needed to better understand the dysfunction of metabotropic signaling in those patients."

My take on it is that humans cheat (likely through cooking food) to get a huge surplus of energy, and use that to get an information-processing advantage over other species. And to make and control fire. We've made ourselves a unique niche, and the number of extinct fossil hominids suggests that it doesn't always work every time. Something to think about, maybe?

Also in the news, from Australia: World first supercomputer capable of brain-scale simulation being built at Western Sydney University ( https://www.westernsydney.edu.au/newscentre/news_centre/more_news_stories/world_first_supercomputer_capable_of_brain-scale_simulation_being_built_at_western_sydney_university )

I wonder whether DeepSouth will be capable of simulating a sufficient diversity of neurons to simulate a human brain?

And so it goes.

In my part of the world, the local brewmeisters come up with things like "grapefruit sculpin" ( https://ballastpoint.com/beer/grapefruit-sculpin/ ). To unpack that, their sculpin IPA (whose taste has a sting, like the fish it's named after, according to their ad copy) won some award(s). So being marketing geniuses, they decided to make it even more bitter by adding grapefruit. It's still on sale, so I guess people like it?

I'm obviously not in their target audience. I'm more interested in how many classes of meds would have negative interactions with that drink. At least three, by my count.

So being marketing geniuses, they decided to make it even more bitter by adding grapefruit. It's still on sale, so I guess people like it?

Well, some of us like bitter melon, so that's a market right there… :-)

Thanks to everyone who offered suggestions and volunteered personal time to help my mother call her friends and relatives in England. The problem is now solved.

Turns out she was calling my cousin's landline, which he had cancelled in favour of his cell phone. When she eventually tried calling a different person (rather than my cousin repeatedly) everything worked, and that's how she learned that my cousin only had a cell phone now. I have no idea what message she heard to make her think her cell phone wasn't working. I also have no idea why she didn't try other numbers, like I suggested at first, rather than repeatedly insisting that the number she called had just worked on her landline which it clearly couldn't have.

Not the first time she's asked her children for advice then totally ignored it until someone else tells her the same thing. Sigh. Nearly 90 and set in her ways.

I'm really feeling this sketch from Three Dead Trolls in a Baggie right now:

https://www.youtube.com/watch?v=nL24aNugo_4

Vent over.

Anyway, thanks for all the help.

Depending on who said cousin's landline provider was (and otherwise this is totally academic) the likely message on dialling a cancelled (and not reused) landline number would be either "The number you have dialled has not been recognised" or "...is not in service".

If yes, you can try setting up a Zoom meeting and have her show you in real-time exactly how she's using her iPhone, hear/see any error messages, etc.

As someone who has done this and similar, it can quickly become a Three Stooges short. Most people needing the help have absolutely no interest in admitting such. And keep jumping to what they think of as the solution while you're still working on step 2 of 10. They want the help to end. I think they feel like it makes them look stupid to not be able to do it on their own.

Robert Prior 375:

parents recall changing your diapers and cannot reconcile a child has become an adult in their own right...

adding to that... too much change in too few years... never mind laptops 'n smartphones there's been so much new stuff

but it doesn't make TAI or TAI timestamps political.

A quick scan of the oracle says "TAI is a weighted average of the time kept by over 450 atomic clocks in over 80 national laboratories worldwide" so I assume there's more than one person involved, possibly even more than one association of people. Which means that of course there's politics.

Even back in the day when Simon Travaglia controlled the internet there was politics. Who got which bit of bandwidtch, which newsgroups were available, you name it. Not as political as Robert Elz desiding who got which .au 2ld but still pretty damn political.

But I was thinking of the Y2K timezone shenanigans as politics at it's purest. Or do I mean silliest? Hard to tell, especially when countries are bending the IDL in order to be able to post "FIRST" first at the end of the first year of the new millenium. Or was it the start of the last year of the old one, I can't be bothered looking up the fine print.

H & Rbt Prior
"Bitter" or "sour"??
I know - both are beer-types & very different { Think De Ranke XX vs any Gueze } However, that morphs into pickles & salt-preserves ...
I recently made a green tomato pickle ( Chutney is far too sweet & sickly ), using wine vinegar & spices, but found I had to add a v small ... { ~ 1tsp = 5ml } of sugar to take the very sharp edge" off. It still doesn't not taste "sweet" at all, but much nicer with cold-meats / cheese / snacks.
Don't be afraid to experiment, when you are cooking!

Robert Prior @ 320, on a UK phone number to use for test purposes.

You can actually set up your own at very low cost. A number of outfits provide an interface between the Plain Old Telephone System (POTS) and Voice over IP (VoIP). I happen to use Andrews & Arnold because they are geek-friendly, but many others are available. For £1.50 per month you can get a phone number with a non-regional UK code. When you call the number it can call your phone or computer over the Internet using Session Initiation Protocol (SIP), or if you haven't set anything else up it will go straight to voicemail and then email you the recording.