Could someone make an LLM that is only trained up to the day before a patent was filed, then ask it to solve the problem at hand.
Hopefully it would come up with the patented idea and thus 'prove' it is obvious and thus not able to be patented. Then you could make different vintage LLMs and basically spam them at trolls to invalidate the patents.
I've had approximately the same thought, to different but complementary ends "More generally (not covered in this question) perhaps this could also be a fun way to interrogate the tech tree, e.g., what could have been discovered given the data at a given cutoff, how early or late certain advancements came, etc." https://manifold.markets/MikeLinksvayer/llm-trained-on-data-...
Hopefully it would come up with the patented idea and thus 'prove' it is obvious and thus not able to be patented. Then you could make different vintage LLMs and basically spam them at trolls to invalidate the patents.
...just a thought from a lurker