Ask HN: Is it possible to backtest math/scinse breakthroughs with AI models?
Posted 2 hours ago by ran_cohen
2 points
By creating a synthetic dataset of everything we know with a cutoff of say 1900 and see if it actually can come up with the same (or more interestingly, different) theories.