Comments by "" (@diadetediotedio6918) on "" video.

  1. "no one claim "reasoning" LLMs can think like human brain" No, quite some people do it, actually even the CEOs overhype this bs. Sam Altman was literally ambiguously quoting someone calling o3 a genius AI (in the human sense of genius), and they don't call those AI's "reasoning models" for no reason. Also, better, I'll quote OpenAI site directly on o1: "We trained these models to spend more time thinking through problems before they respond, much like a person would." So, nobody is saying that?
    3
  2.  @existenceisillusion6528  Using tools is cheating as the problem is easily bruteforceable. And it should not be a problem for 10k more context window, which those models have.
    2
  3. ​ @existenceisillusion6528  Using code tools is cheating, which defeats the purpose of the experiment. It is trivial to brute force this problem. The questions are if those models are able to at least try "consciently" solve the problem using the right algorithm without losing themselves, and I don't think the problems require "exceeding the context window", tower of hanoi for 10 combinations, for example, only requires ~1023 moves, the contex window of those models is way beyond 10k tokens. If they are using so many tokens to solve this problem, even with the correct solution already stated out, it IS a problem, a HUGE problem.
    1