Flippanarchy

Correct, "AI" does not reason like animals do. And that reasoning that we humans do is not what I mean when I say "reasoning" in regards to "AI". I neither need nor want truly intelligent machines. But what "AI" can do is stochastically recall relevant information and stochastically string it together to spit out useful programs. Despite the inherent randomness and indeterminacy of these systems, a good background in probability theory, specifically the theories of statistical learning and empirical processes, allows us to analyze machine learning systems, although the theory needs to be brought up to speed with the development of large language models.

In the meantime, we know that they can solve some practical, technical problems. (Emphasis on technical; I have no interest in "AI"-generated images, sounds, culture, etc. I am artistically, historically, philosophically, culturally interested in the goings-on of sentient beings, not nonliving things like rocks and computer algorithms.) In particular, I use it as a LaTeX assistant. (LaTeX is a document typesetting programming language aimed at mathematics documents. It is practically like Word for math people, but you write it like a program, since LaTeX is also a Turing-complete programming language). I already knew LaTeX well enough to do my homework in it before the LLM craze, but still, "AI" has saved me countless hours of unenlightening LaTeX debugging. It is not much of a stretch to imagine that other programming languages may someday be "understood" by the "AI", whatever that means mathematically. And of course, there have been some discoveries in medicine that have been enabled by "AI", although medicine is not my research area.

It is the cardinal sin of the statisticians who invented the field that "artificial intelligence" uses and abuses concepts from psychology, philosophy, and pedagogy as names for their mathematical concepts. For the time being, I believe it is fair to use the words (like "reasoning", "learning", "memory", etc.) with the caveat that in this context, each of these words mean to encode an ill-defined concept applied specifically to machine learning algorithms. Or to use computer science terminology: we overload the meaning of some of these words specifically for "AI" systems. But in the future, we really need to unfuck the vocabulary of this field...and take the math and the hardware back from the capitalist scum who are using this useful math to destroy the world.