I have an idea how to test whether AI can be a good scientist:
Train on all published scientific knowledge and observations up to certain point, before a breakthrough occurred. Then see if your AI can generate the breakthrough on its own.
For example, prior to 1900 quantum theory did not exist. Given what we knew then, could AI reproduce the ideas of Planck, Einstein, Bohr etc?
If not, then AI will never be useful for generating scientific theory.
I don’t think this is the main point of the paper. They’re not claiming that AI is capable of scientific breakthroughs. Rather, they argue that AI excels at summarising vast amounts of existing scientific knowledge.
Or just have the AI generate new specific experimental setups and parameters that we can try and be like "oh yeah, we just made a room temperature superconductor".
Honestly given what we know about physics, the AI should be able to simulate physics within itself or deduce certain things we've missed.
Train on all published scientific knowledge and observations up to certain point, before a breakthrough occurred. Then see if your AI can generate the breakthrough on its own.
For example, prior to 1900 quantum theory did not exist. Given what we knew then, could AI reproduce the ideas of Planck, Einstein, Bohr etc?
If not, then AI will never be useful for generating scientific theory.