Atari classics used to train artificial intelligence platform
1 min read
views
Nearly 60 classic Atari games will be incorporated into the brand new OpenAI Gym, a scenario generator for training the OpenAI artificial intelligence platform.
«Space Invaders», «Pac-Man», «Qbert», «Pitfall!» and «Pong», among other titles, make up the list of "scenarios" that the platform can use to learn and enhance its reinforcement learning system.
This section of the platform (namedArcade Learning Environment) uses Stella - an emulator of the Atari 2600 console - to evaluate whether said artificial intelligence system can learn and achieve a perfect game.
In this case - and as with the other more conventional challenges - the objective will be to get the highest score from the AI. In addition, OpenAI will keep a leaderboard open from which those with the best results can be compared.
It is worth remembering that in December 2015, Tesla CEO Elon Musk announced the launch of OpenAI, a non-profit organization dedicated to the study and development of artificial intelligence with the promise of sharing research with the rest of the world.
The so-called OpenAI Gym aims to simulate scenarios in which this platform can learn and enhance its reinforcement learning system; which is in fact the key to DeepMind, Google's artificial intelligence engine that recently achieved the incredible feat of beating one of the world's best human Go players.
Although in beta, OpenAI Gym is now available for testing in Python and is expected to be updated soon to work with other programming languages.
With information from The Other Side and Xataka
Posted by
Please comment politely. Spam comments will be moderated. Thank you.