With the back drop of AI in one month having solved one millennium problem and provided partial solutions to two more.

As the headline says, would a frontier model today with an unlimited token budget be able to on its own just based on a prompt giving it the direction solve the protein folding problem that was solved by deep mind with the help of AI and awarded the Nobel prize?

The hard version of this is that it correctly figures out the mechanism and correctly predicts forms with no experimental verification.

The softer version is that it gives experimental directions for people to verify and feedback but would lead the research effort in solving it. Or in other words, has AI advanced enough to essentially reverse the roles, solve it with the help of humans?

What does the hive mind think?