@fdchud@jsuarez We don't care about the drone, or if the problem has been solved or not. It's the fact this solution can be arbitrarily solved in seconds using reinforcement learning. That's insane.
The rubiks cube has also been solved, but you can still use that problem to display progress.
@Bharath70424554@ylecun@mmbronstein@anshulkundaje DeepMind did not solve the NP-hard problem mathematically but instead, they bypassed it by shifting from a physics-based calculation to a data-driven pattern-matching prediction
@Bharath70424554@ylecun@mmbronstein@anshulkundaje They can have workarounds taht are exciting. For example the prediction of a protein's 3D structure from its amino acid sequence was and theoretically remains to be an NP-hard problem.
@bcherny and @ClaudeDevs I was just using claude code and it went completely insane and couldnt end its own stream, quite funny dialogue but a real token burner if I didnt stop him! https://t.co/JNgrCli6MX