Comments by "Chrysippus" (@4.0.4) on "David Shapiro" channel.

  1. It's a problem when most of the "AI safety" community is fear-mongering, hyperbolic and trying to find the sci-fi novel in the data. And usually the best "solutions" this bunch comes up with is "more government regulation".
    87
  2.  @sahilx4954  which will we get first, GTA-6 or GPT-6?
    3
  3. There have been AI winters in the past, so it's perfectly possible SOTA doesn't go past Transformers (any opinion to the contrary sounds like a bunch of mamba jamba to me 😉), but even then, there's a lot of tech that's possible today and just not implemented. Regarding robots, biggest issue is battery. Animatronics are glued to one spot and power consumption is a footnote.
    1
  4. "Hey let's do the golden gate bridge thing but for making the model more politically correct" Open models are the future.
    1
  5.  @MokeAnit  Fair enough. 1) Educate people about what AI actually is, so they don't get savior complex and become a Blake Lemoine when a bunch of linear algebra says it's conscious. 2) Normalize the idea that anybody right now can make a deepfake out of you and there's nothing we can do about it. If people falsely believe government can "solve this", they will gladly pass as many draconian legislations as possible in the fruitless attempt to "solve the problem", which will never be the real goal, and will never be done. 3) Force big companies (OpenAI, Anthropic, etc) to open their research, their data, etc, to minimize the potential for centralized abuse of power. 4) Empower open research, local AI and "freedom of alignment" - no company should decide what vague idea of ethics the AI should follow but the user. 5) In general, put AI and all its responsibilities in the hands of human beings, not CEOs, politicians and "trust and safety" councils. 6) Dispell fantasies from science fiction that got everything wrong but are still staining the collective psyche with anthropomorphizing nonsense.
    1