One small point to mention, LLMs != AI. “LLMs”, or large language models, are the chatbots you see being promoted as “AI”. LLMs are under the umbrella of AI, but not all AI models are LLMs. This is very important. LLMs are essentially a more complicated version of autocomplete, except that it predicts the next sentence(s). LLMs are pretty bad at maths, but not all AI models are bad at it!

There are many applications where (non-LLM) AI models are being used in science, and some of them are genuinely pretty cool! Not just in math-aligned fields either, AI models can be used anywhere with a large quantity of data that needs to be processed. You also need to take into account that these models are usually only trained on relevant data for the specific study, making them far more focused and less contaminated than your average LLM, which is fed pretty much the entirety of the Internet (garbage in, garbage out).

LLMs owned by big corporations aren’t that great at most things and are horribly inefficient, but AI in the general sense could be an important tool (not a replacement, but a TOOL!) for science and discovery…

It doesn’t help that marketing corpo people say the AI keywords twice a sentence in literally everything.