Last month OpenAI, the non-profit organization focused on artificial intelligence research, released the public beta of its OpenAI Gym (H/T WIRED). The Gym is a toolkit for reinforcement learning algorithms. These algorithms govern decision making and motor control, and the OpenAI Gym has been used by the company in support of its own research on reinforcement learning.
The public release of Gym is consistent with OpenAI’s intent to be an open source operation. A goal of that focus is to try and ensure that the research and knowledge about AI remain accessible to those without the deep pockets of a Google or a Facebook. It certainly doesn’t hurt that some of the current challenges in the field would benefit from having a lot of different people working on the problems and generating their own datasets for comparison. In support of that goal, those who download and use the OpenAI Gym are encouraged to upload their results and try and reproduce the results of others.
Presumably one of the metrics of success would be use of the Gym and growth of a community of researchers and research data. Not quite a month after release, I suspect it’s too early to evaluate where things are.