Hugh Hancock, your friendly neighbourhood crafter of tales about supernatural get-rich-quick schemes gone horribly wrong, back with another bit of musing on what the Chatbot Future holds... See also Part 1 - Sexbots and Part 2 - Magical Beasts
In "Accelerando", Charlie posited the idea of a swarm of legal robots, creating a neverending stream of companies which exchange ownership so fast they can't be tracked.
It's rather clear to me that the same thing is about to happen to social media. And possibly politics.
What makes me so sure?
Microsoft's Tay Chatbot. Oh, and the state of the art in Customer Relationship Management software.
Turing Test 2: Is The Bot Distinguishable From An Asshole?
Microsoft unleashed its conversational bot on Twitter, and 4chan's /pol/ unleashed their opinions - or possibly their sense of humour - on it in turn. Hours later, it was a racist asshole.
But that's not the interesting bit.
The interesting and worrying part of the entire test was that it became a plausible, creative racist asshole. A lot of the worst things that Tay is quoted as saying were the result of users abusing the "repeat" function, but not all. It came out with racist statements entirely off its own bat. It even made things that look disturbingly like jokes.
Add a bit of DeepMind-style regret-based learning to the entire process - optimising toward replies or retweets, say - and you have a bot that on first glance, and possibly second through fourth glance, is indistinguishable from a real, human shitposter.
A lot of ink has been spilled worrying about what this says about the Internet. But that's the wrong thing to worry about.
The right thing to worry about is what the Internet is going to look like after more than one Tay is unleashed on it.
More than a hundred. More than a thousand.
