• 0 Posts
  • 6 Comments
Joined 2 years ago
cake
Cake day: July 2nd, 2023

help-circle
rss



  • Sure, it’s hard to say whether a computer program can “know” anything or what that even means. But the paper isn’t arguing that. It assumes very little about how how LLMs actually work, and it defines “hallucination” as “not giving the right answer” with no option for the machine to answer “I don’t know”. Then the proof follows basically from the fact that the LLM-or-whatever can’t know everything.

    The result is not very surprising, and saying that it means hallucination is inevitable is an oversell. It’s possible that hallucinations, or at least wrong answers, are inevitable for different reasons though.


  • I think the point might be reasonably condensed to:

    • Africa is big and diverse, and its internal geographic barriers (particularly the Sahara) are more significant than the ones dividing it from Europe and from southwest Asia.
    • Some parts of Africa have thousands of years of written or otherwise well-documented history, and each part has seen several waves of significant change, including colonization from other areas of Africa (e.g. by Egypt or Mali), from Europe (e.g. by Rome), and from southwest Asia (e.g. by the Umayyads); and colonization of other areas (e.g. of the Iberian peninsula by Morocco).
    • For some parts of Africa, the latest round of European colonization is arguably less significant than previous changes.
    • Thus, for serious discussions of history, “pre-colonial Africa” is not a useful division to make: you won’t be able to say anything meaningful without more precisely specifying the time and region (e.g. “medieval west Africa”).
    • This isn’t fixed by changing to “pre-European Africa”.
    • Both “pre-colonial Africa” and “pre-European Africa” additionally suck because, instead of using a more relevant division, you are using a less-relevant Eurocentric term.