Comment by godelski

  > What scares me is the rampant inaccuracy

What scares me is the massive incentivization to manipulate the results.

With AI ads you get all the power from big data aggregation, the trust/framing of an authoritative voice, and cheap personalization that specifically optimizes for what convinces you. It's too powerful. Even if it only works a small percentage of the time we're interacting with these things so frequently that a small percent is a large number. They're already feeding user profiles into these machines and there's explicit talk about having the LLMs optimize ad campaigns. It's already dystopian if it's ads to get you to spend your money, but people seem to dismiss that. Do we not care that this is also being used in the same way to convince you to believe certain things? To join certain political organizations?

Yeah, these things help me write more lines of code faster (if we include all the lines from our design docs) but I don't like the idea of pointing a supercomputer at my brain and someone else using it to try to manipulate me. That's not a game I'll win. It's not a game you'll win either.