He reads a conversation that he had with ChatGPT and it told him that communists want to kill every dog because then you can’t run your own farm and then you have to get a job at the government. Then he says once you have a job at the government, you would get fired for having free speech.
The biggest problem with all of these AI is that in order to maximise engagement they are being written to be as agreeable as possible with whoever is using them. The private companies running them quickly realised that if the AI disagrees with the user it pisses off the user and results in reduced engagement and anger at AI overall.
As such, AI now just reinforces the pre-existing brainworms of the user. If it does not, the user says “that’s wrong” and the AI corrects itself to the user’s liking.
This is the obvious outcome of any for-profit AI project. Maximising engagement will always result in this outcome over truth even in projects run by people who originally thought they had good intentions.
Yes 100% this, I have been thinking a lot lately about how the “engagement metric” is, to be a bit dramatic, a demonic impulse and it’s driving this chaos or at least contouring it. It’s the tiny sliver of wood deep within the wound that’s preventing the infection from clearing up, might be the analogy I’d go for.