Comments by "clray123" (@clray123) on "Lex Clips" channel.

  1.  @SafetyLucas  House cats and artists tend to disagree. The "work hard to earn something (social status/heavenly rewards)" is nothing universally human, but just a particular worldview that was popularized by Calvinists in the US, leading to great economic advances and military domination of the remaining world, thus becoming the accepted modern ideology. Being happy while (relatively) unproductive is an option which is frowned upon nowadays simply because it's not "politically correct" (being envious of other people and hypercompetitive is, on the other hand, quite politically correct). But pursuing individual happiness also has a long philosophical tradition (e.g. epicureism, asceticism [I bet those Christian saints also indulged in shrooms], or even hedonism). I never took any drugs nor do I intend to (because I don't wish to mess up my head and I'm pretty content with what my body produces for itself), but my view is that in times of surplus being the unproductive happy fool may be even the more reasonable choice (e.g. causes less long-term environmental damage than the opposite).
    1
  2. At 9:44 it is clear that Wolfram has not heard of glitch tokens... that's how clueless he is about LLMs.
    1
  3. The only thing he demonstrates is that mathematicians have a very peculiar notion of "existence". More specifically, they are quick to declare something "exists" when the find an algorithm with which some interesting tricks can be done. Same goes for their understanding of "infinities" (which really just means something like "an arbitrary number of code executions" in algorithmic terms).
    1
  4.  @katehamilton7240  why would they make agi impossible? is your own brain impossible or contradictory to math theorems?
    1
  5. What are you talking about, he usually looks like he's stoned or drunk to me, speaking very slowly and having difficulty finding simple words.
    1
  6. I thought reptiles would not sweat?
    1
  7. When you "imagine a scenario" for AI to "realize it", you have to define a clear target function for the AI to learn. It turns out that defining such optimization targets is just as hard, if not harder, than actually sitting down on your ass and implementing it directly.
    1
  8. Lex is the sort of guy I can see blowing some public funds on travel to some pseudoscientific "conference" to smoke a bong there.
    1
  9. It's not random, but the authors of the paper did a really shitty job explaining their design choices.
    1