Hacker News new | past | comments | ask | show | jobs | submit login

Beating AlphaZero at Go, Chess, and Shogi, an mastering a suite of Atari video games that other AIs have failed to do efficiently. No explicit heads-up contests with a trained AlphaZero; but apparently hits an ELO threshold w/ fewer training cycles. Yowsa.



And specifically, enabling the application of reinforcement tree searching agents into domains with blackbox environments. This paper is not only about increased convergence performance, it is about enabling agents in real world scenarios where the definition of environment is not feasible. Maybe of arbitrary complexity. I do not believe in AGI yet, but that's one of the bigger steps in the right direction, one would assume.




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: