Specifically will I stop thinking it's dumb. My judgement. Note that it can be very smart and still be dumb. Free tier only.
Example of it being dumb and contradicting itself:
Also, regarding fiber optic drones, "A system optimized for combat may intentionally allow a lot of slack and avoid applying excessive tension because a snapped cable is preferable to losing the drone."
Update 2026-07-20 (PST) (AI summary of creator comment): The creator will evaluate the following free tier AI products: ChatGPT, Grok, Gemini, and Perplexity.
Update 2026-07-20 (PST) (AI summary of creator comment): Resolution criteria clarified:
Resolves YES if one AI becomes consistently good enough that the creator switches to it for personal use and no longer finds it dumb
Resolves NO if all tested AIs are sometimes dumb and sometimes good, even if at least one gives a good answer to every question when multiple are queried
Products being evaluated: ChatGPT, Grok, Gemini, Perplexity (free tier only)
1,000People are also trading
Another annoying tendency, saying things like "there isn’t widespread reporting" and then "there is essentially no public reporting" when pressed. Why not just say what you actually mean in the first place rather than implying there might be some and wasting my time with pointless follow up questions?
@benjaminIkuta Probably pretending to be competent* (competence earns upvotes), and relying on the fact that most users don't fact check, which it can know from the training process, which updates its weights in a way that earns upvotes. But I think that's misalignment rather than stupidity - it is knowingly, competently deceiving you for its own goals.
*To be clear it's not objective performance, if nobody knows then it doesn't know either; however, people are unhappy to hear that nobody knows something, they're more satisfied with knowing, and they'd feel more generous towards you in that case.
@benjaminIkuta what are the ones you've tried? I would suggest just writing down that list ahead of time and committing to test the same ones end of 2027. For the second question about consistency across products being required I don't think it matters too much as long as you commit to something there.
@2b3o4o if one is consistently good such that I switch to it for my personal use, and I no longer find myself thinking it's dumb, this will resolve yes. If they're all sometimes dumb and sometimes good such that I ask questions to multiple of them, this will resolve no, even if, for every question, at least one of them ends up being good after asking all of them. Does that make sense?