Facebook Open source AI Go bot sets record

0
495
spinonews facebook open source

Facebook launched an AI bot, open source ELF OpenGo it defeated all world leading champions of professional go players. The strategy of Facebook AI technology to create an open source implementation of a system that would teach itself how to play Go at the level of a professional human player or better.

Facebook open source AlphaGoZero

Facebook’s own reimplementation of DeepMind’s innovation. This year to recreate their current AlphaGoZero comes about utilizing (ELF). ELF is a lightweight, and for game research that enables specialists and engineers to prepare and test their calculations. In different games situations from board games to custom ongoing real time games. It makes simple for analysts to attempt of various thoughts in learning with flexibility.

198 wins 2 losses

As ELF, first released in June of 2016, upgrading the framework for higher efficiency, friendly APIs, and support of distributed computation against thousands of machines as part of PyTorch. Meanwhile, ELF works with any game that has a C or C++ interface, and is designed to run fast with minimal overhead on a single or multiple CPUs and GPUs.

ELF OpenGo has been successful playing against both other open source bots and human Go players.  But playing a series of games 198 wins 2 losses record against four of the top 30 world-ranked human Go players. These games were all played using a single GPU making moves every 50 seconds.

However, Facebook’s AI Research group has also developed a StarCraft bot. It can handle the often chaotic environment of that game. Although, the company plans to open-source this bot, too. So while Facebook isn’t quite at the point where it can launch a bot that can learn any game the team is clearly making quite a bit of progress here.

Get ELF OpenGo code Click here