Scholars: AI isn't "hallucinating" – it's bullshitting
psypost.org
psypost.org
During the last lessons when people where generally pretty ok with processing normal text we started to work with markov chains. These can be very interesting for a kind of generation where you still "feel" the source material present in the branching paths of the output. I am aware that LLMs are different from markov chains, but I can't help to think about markov chains and their way of just giving you likely texts, not necessarily correct ones.