Comments by "Nicolae Crefelean" (@kneekoo) on "Lunduke v. ChatGPT - "Cease Making Defamatory Statements"" video.

  1. It looks like ChatGPT is trained (1) to make stuff up, (2) through an ideological bias. So even when it has information contradicting what it generates, it will ignore that and prioritize the bias. The huge problem of those two training criteria is that it also makes it highly unreliable.
    38
  2. First comment and like, apparently. :)
    5