[ababaian]

It's not OT at all. It depends on what you mean exactly by LLM, but we use them all the time. ESMFold2 was an LLM [https://www.science.org/doi/10.1126/science.ade2574] and was instrumental in advancing deep protein fold prediction in metagenomic space. Likewise AlphaFold2 it's direct application in creating FoldSeek for ultra-deep homology search. Both of these actually have radically improved our capacity to say Obelisks are _REALLY_ not like anything that's known.

Besides that I encourage all students to use ChatGPT for research, coding, copy editing, etc... I haven't encountered an LLM that can deal with difficult domain problems like we're facing, but I'd welcome the help. I'm for using all tools available, my main criticism with AI/LLM in general is the poor way in which uncertainty is reported.