99.9% sure this is a joke. But in case anyone isn't joking or is blindsided by it...
Artificial intelligence is a huge ass field of research that involves many many different types of math and algorithms, and Large Language Models are the modern invention that people have a big problem with.
They are massive neural nets, and even neural nets alone aren't problematic. But this form of AI is used to do millions of calculations which are essentially lots and lots of linear algebra to calculate just the NEXT word in the response, going off statistics, likelihood, different neural nets being activated that were trained to be "experts" at different things, and use tons of cpu and tons of ram in the process because they are billions of parameters wide, meaning billions of calculations just to tell you "lol yeah bro you're totally cool and right".
And due to their design there's a focus on some extra randomization to give them creativity, make them say less likely things that still work, but can also lead to non deterministic behavior and hallucinations and drive engineers mad.
Lots of AI research is really cool. Shit, even LLM research is cool on its own, when separated from the reality of the situation. But the billionaires are fucking you because they think that LLM can replace you as a worker, and are greedy fucks who think they're right because the LLM stroked their massive ego.
Then there's totally unrelated ai, like AI used in game design for path finding or just interesting game agents making good decisions. These do not consume nearly as much computation as LLM. If it runs on your home computer, even a decent gaming computer, it doesn't scale close to what they're doing to render slop videos at scale.
The AI for a zombie to choose where to go is massively less calculations than just rendering the blocky ass zombie with lighting, which is magnitudes less than what chatgpt does to summarize an email, and I'd argue that's one of its more rational uses.