“OpenAI admits AI hallucinations are mathematically inevitable, not just engineering flaws – Computerworld”
> OpenAI’s own advanced reasoning models actually hallucinated more frequently than simpler systems.
An observation I made in my book, even in the first edition two years ago, was that research gave us reason to believe that hallucinations got worse with increasing model size
So, y'know... I literally told you so
Follow
@baldur the thing about more advanced models hallucinating more, I suppose it's kind of the same effect as when you fit a polynomial to a set of points? If you try to get fancy with a higher degree polynomial you can make it fit a larger number of points, but the curve will shoot off even more crazily in between.
