Comments by "Brent Jacobs" (@br3nto) on "Lex Clips" channel.

  1. I think to say we don’t know what is happening in the inner workings of AI models is underestimating what the AI experts understand. I think they could very precisely tell us what is happening. It’s not a big mystery this guy is making it out to be.
    3
  2. In the green potato example before 17:07, surely it’s response is still just predictive text… it’s not choosing to give those responses, rather it’s model is dictating the response. It’s a static deterministic response.
    2
  3. 1:18 who’s they?
    1
  4. Why does it matter how an AI or AGI learns? Is there something more valuable about learning similar to human vs learning via forced training techniques?
    1
  5. 1:58 this description of LLMs writing to each other using human language as the transport layer is right out of the foundation series by Isaac Asimov! Where they receive a political letter and it distills down to basically nothing! Wow he was one forward thinking cookie.
    1
  6. How do you that humans don’t just spew back the things they’ve absorbed? At what point is a human sentient?
    1
  7. “They can play chess. How can they do that without reasoning”. Chess is deterministic. It doesn’t need reasoning. If anything, an AI chess game would be boring because it doesn’t need to consider the moves of the other player, it just applies the winning strategy for the current board setup.
    1
  8. The problem with self driving cars is that they are measured at existing speed limits. If the speed limit was reduced by 20kph there would possibly be no issues with the current tech
    1
  9.  @topchoicerollzroyce1181  yeah dude, that’s why I’m saying what I’m saying.
    1
  10. I wonder if the “collective intelligence paling in comparison to the magnitude of the general super intelligence in the AI systems we build” would only be a temporary state. Right now they just make connections where we, individually may not have made connections yet… like how statistic and data analysis highlights links and trends we couldn’t see otherwise… what happens to AI intelligence once the AIs have finished making all the links? Can they make fundament new discoveries for anything beyond the scope of the data it has access to or will they only be able to make discoveries within the scope of data it has access to and thereby placing an upper limit on the order of magnitude their intelligences surpass the individual and collective intelligence of humans?
    1