Comments by "Martinit0" (@Martinit0) on "World Science Festival" channel.

  1.  @anythingplanet2974  But AI alignment is not the biggest problem. The problem is (and has always been) alignment of humans with other humans. Unaligned (evil) human V.P. will advance AI no matter what regulations in the US or Europe mandate. You can attempt to curb it by limiting the means of production, i.e. computing hardware, not by limiting the distribution of information (which is too easily leaked and cannot be taken back).
    2
  2. Evil people are not deterred by regulation. Pretty sure actors in China and Russia are already working on it, if not even deploying. They are basically laughing their asses of about the West crippling themselves.
    2
  3. LLMs by themselves lack the feedback mechanisms to implement planning. So you need another component that basically guides the LLM
    1