Since we haven't "cured cancer" yet, its unlikely that AI is going to generate tokens in the "correct sequence" that leads directly to a cure. Simply because AI cannot be trained on solutions to unsolved problems, by definition, it's unprovable that the solution is ever going to be produced during inference.
However, given enough time, a monkey and a typewriter can theoretically reproduce the works of Shakespeare.
Maybe cancer researchers can spend more of their human-capital/creative-energy on unsolved problems now that AI can write their grant proposals and do their taxes for them.
Since we haven't "cured cancer" yet, its unlikely that AI is going to generate tokens in the "correct sequence" that leads directly to a cure. Simply because AI cannot be trained on solutions to unsolved problems, by definition, it's unprovable that the solution is ever going to be produced during inference.
However, given enough time, a monkey and a typewriter can theoretically reproduce the works of Shakespeare.
Maybe cancer researchers can spend more of their human-capital/creative-energy on unsolved problems now that AI can write their grant proposals and do their taxes for them.