|
|
|
|
|
by lacker
1520 days ago
|
|
You can generate arbitrary problems easily enough The problem with generating arbitrary math problems is that you can generate an infinite number of boring, pointless problems. For example, prove that there is a solution to the system of equations x + 2y + 135798230 > 2x + y + 123532 and x - y < 234. Training an AI system how to solve these problems doesn't do anything. I think we are in a stage for mathematics similar to where solving Go was in the 80's. For Go back then we didn't even have the right logical structure of the algorithm, we hadn't invented Monte Carlo tree search yet. Once a good underlying structure was found, and we added on AI heuristics, then Go was solvable by AIs. For mathematics I think we also need multiple innovations to get there from here. |
|