Artificial Intelligence | 1 min read

Messing around with OpenAI Gym

So while I was looking around for interesting Python-based AI projects I came across OpenAI Gym, backed by mister Elon Musk himself. This application aims to provide the ultimate sandbox environment for AI-research, but at the same succeeds in providing the ultimate geek toy for AI-researchers. Let me show you why.

OpenAI Gym

First of all it might be useful to explain what OpenAI Gym actually does: OpenAI Gym aims to provide an easy environment to develop and test reinforcement learning algorithms. To be clear, OpenAI Gym doesn’t power any algorithms itself, leaving it up to more specialised packages like TensorFlow or Theano.

OpenAI Gym environments

So what makes this the ultimate geek toy for AI-researchers? Well, this is because of the many environments OpenAI Gym provides, one of them being the ‘atari’ environment. That’s right, you can test the performance of your reinforcement learning algorithms on a variety of different atari games and what’s more, you can automatically upload the performance of your algorithms and compare them to other people’s approaches.

OpenAI Pacman
Learning performance

To save myself a few weeks of research and computing time, I decided to download one of the top performing algorithms and see how it performs. The only input the algorithm gets is the raw RGB values of the screen pixels and the current score. Additionally, the algorithm is also made aware of all the possible actions within the game thanks to the OpenAI Gym framework.

On the right you can see that the final algorithm is the result of about eighty thousand iterations. The result, seen on the left, is quite impressive: notice for example how miss pacman waits in the top-right corner until a ghost arrives, which she promptly consumes using the power pellet.

Conclusion

To conclude, OpenAI Gym is a cool platform for anybody involved with reinforcement learning algorithms. The platform will allow you to test your algorithms in a variety of different environments without having to go through the hassle of making the right inputs available to your algorithm. It has been proven many times before that open-sourcing encourages innovation, and we can’t wait to see what kind of influence OpenAI will have in the future!

Related stories