i feel like there are people who structurally don't have a good enough grasp at logic to talk about AI. take this for example
>>105992275
>And currently, AIs are training on self themselves so they're getting even further ahead.
like i intuitively know that can only degrade the results, but for some reason, some people think it's good? like it should be obvious that if even 1% of all the stuff AI produces is a completely incorrect hallucination, and also it trains on itself or other AIs with a non-0 hallucination rate, then on a long enough timeline they will all regurgitate nearly 100% hallucinations