[tremon]

I sincerely think that LLMs will at least make R&D cheaper.

How, exactly? Because to me it's just as likely that LLMs will hallucinate alternative solutions based on their flawed world model which will send numerous unfortunate researchers on wild goose chases that turn out to be exactly that. And I expect the volume of impossible-yet-probable-sounding solutions will dwarf the actual costs saved by using an Automated Induction system.

[bbor]

Long story short: every engineer and scientist in the country will suddenly get 100 interns. Will they make perfect work? Hell no. But current levels of Intuitive Computing LLM tech can be extended to much higher levels of autonomous/agential behavior, and that’s all you need to bring computers from a tool to a partner.

Your criticism of an all-in-one induction system / Reasoning Engine is well founded, no disagreement from me. I just think that they’ll be able to help in myriad, smaller ways. Finding synergies, analyzing data, designing and employing frameworks/simulations/tools specific to the researcher’s work, and just generally being a bank of knowledge that can be easily browsed through complex linguistic filters.

IMO :) I am an optimist. Maybe it turns out chatgpt is the best we get, in which case I’m very very pessimistic about our chances of meaningfully solving climate change, rocket-launched-lunar-dust or no. So… I have a “fingers-crossed” based leap of faith in my reasoning somewhere