Main image of article A New 'Gym' for Building and Testing A.I.

If you’re interested in working with machine learning and artificial-intelligence algorithms—but unsure of how to start—check out the OpenAI Gym, now in beta. The premise behind OpenAI Gym is simple: it’s a toolkit for building reinforcement learning (RL) algorithms, which govern bots’ decision-making and motor-control capabilities. Reinforcement learning is a key element in A.I. development, as it allows software to deal with random, unpredictable environments; one “classic” problem involves balancing an untethered pole on a rolling cart: Screen Shot 2016-05-02 at 10.40.15 AM OpenAI is a non-profit “artificial intelligence research company” funded by some heavy hitters in the tech world, including Tesla CEO Elon Musk and venture capitalist Peter Thiel. Its altruistic goal is to develop open-source A.I. software that’s “friendly” to humanity. According to a blog posting accompanying the launch of OpenAI Gym, RL research is slowed by two factors: a need for better benchmarks, and a lack of standardization of environments used in publications. The latter issue “makes it difficult to reproduce published research and compare results from different papers.” Collections in the OpenAI Gym include board and Atari games, simulated 2D and 3D robots, and a variety of algorithms. If you’re playing around with those assets, you can upload your results to OpenAI Gym, although the company has declined to set up any sort of leaderboard to track progress. “What matters for research isn’t your score (it’s possible to overfit or hand-craft solutions to particular tasks), but instead the generality of your technique,” the blog posting added. “We’re starting out by maintaining a curated list of contributions that say something interesting about algorithmic capabilities.” It remains to be seen whether OpenAI can help govern the evolution of A.I. In the meantime, though, anyone interested in machine learning has a new resource to check out.