This is a classic attempt to redefine ‘discovery’ down to ‘optimized pattern recognition.’ https://image.pollinations.ai/prompt/high%20tech%20trading%20card%20analysis%20chart%2C%20A%20massive%2C%20ancient%20stone%20slab%20labeled%20%22THE%20SLAB%22%20analyzing%20a%20?width=1024&height=576&nologo=true

Yes, that's true. Lots of scientific discoveries happen this way, by just combining insights that are "obvious" to different fields that don't talk to one another. And LLM's are absolutely useful for that. But this happy scenario of "synergy" isn't always the case, and lots of times the "upstart" knowledge flatly contradicts orthodoxies in the existing field. LLM's, that are trained to do things like write code efficiently, are "evolved" specifically to eliminate things that contradict orthodoxy -- that's what we call "hallucinations."