It doesn’t bode well when the co-founder of Hugging Face, the preeminent open-source AI platform, starts a conversation with “current AI systems are unlikely to make the scientific discoveries some leading labs are hoping for.”
“In science, asking the question is the hard part, it’s not finding the answer,” Wolf said. “Once the question is asked, often the answer is quite obvious, but the tough part is really asking the question, and models are very bad at asking great questions.” […] “Models are just trying to predict the most likely thing,” Wolf explained. “But in almost all big cases of discovery or art, it’s not really the most likely art piece you want to see, but it’s the most interesting one.”
And even more damning:
He argues that what we have instead are models that behave like “yes-men on servers”—endlessly agreeable, but unlikely to challenge assumptions or rethink foundational ideas.
