Elon Musk Launches AI Gym To Train Robots
29 April 2016, 9:57 pm EDT By Katherine Derla Tech Times
A gym for artificial intelligence? Yes, that’s right. We wonder how much a monthly membership would cost. Elon Musk launches the OpenAI Gym Beta to help train robots. ( Bill Pugliano | Getty Images )
Business magnate and Tesla CEO Elon Musk has opened a "gym" where developers can effectively train their individually built artificial intelligence (AI) systems. The gym is under Musk's OpenAI, a non-profit AI research company.
The newly opened OpenAI Gym Beta was released on April 27. It provides AI developers with a toolkit that can help them advance and analyze reinforcement learning (RL) algorithms.
"It consists of a growing suite of environments (from simulated robots to Atari games), and a site for comparing and reproducing results. OpenAI Gym is compatible with algorithms written in any framework, such as TensorFlow and Theano," wrote Greg Brockman and John Schulman in the company blog.
The "environments" provide situations where the AI's skills can be tested. For instance, the AI can be put in an environment where it needs to complete several tasks or play games.
These environments currently include classic board games such as Go, where AI can play on 9x9 and 19x19 boards. It also includes 59 Atari games such as Pac-Man, Space Invaders and Asteroids.
The OpenAI Gym Beta also includes environments where the AI can control a 2D or 3D robot simulation or complete computations using multi-digit numbers.
AI developers can upload their test results to the OpenAI Gym system, but the company decided against creating a conventional leaderboard.
"What matters for research isn't your score (it's possible to over fit or hand-craft solutions to particular tasks), but instead the generality of your technique," added Brockman and Schulman.
They also said the environments are currently written in Python but their team will eventually make them easier to use in any language.
They added that the Open AI Gym is actually used by their company as a tool to help them advance their own RL research. They hope that making it available to a wider community will benefit more AI developers.
OpenAI is hoping to get feedback for the beta version of their AI gym so the company can make adjustments and make it a better research tool. If you're an AI developer, perhaps it's time to hit this gym.
Business magnate and Tesla CEO Elon Musk has opened a "gym" where developers can effectively train their individually built artificial intelligence (AI) systems. The gym is under Musk's OpenAI, a non-profit AI research company.
The newly opened OpenAI Gym Beta was released on April 27. It provides AI developers with a toolkit that can help them advance and analyze reinforcement learning (RL) algorithms.
"It consists of a growing suite of environments (from simulated robots to Atari games), and a site for comparing and reproducing results. OpenAI Gym is compatible with algorithms written in any framework, such as TensorFlow and Theano," wrote Greg Brockman and John Schulman in the company blog.
The "environments" provide situations where the AI's skills can be tested. For instance, the AI can be put in an environment where it needs to complete several tasks or play games.
These environments currently include classic board games such as Go, where AI can play on 9x9 and 19x19 boards. It also includes 59 Atari games such as Pac-Man, Space Invaders and Asteroids.
The OpenAI Gym Beta also includes environments where the AI can control a 2D or 3D robot simulation or complete computations using multi-digit numbers.
AI developers can upload their test results to the OpenAI Gym system, but the company decided against creating a conventional leaderboard.
"What matters for research isn't your score (it's possible to over fit or hand-craft solutions to particular tasks), but instead the generality of your technique," added Brockman and Schulman.
They also said the environments are currently written in Python but their team will eventually make them easier to use in any language.
They added that the Open AI Gym is actually used by their company as a tool to help them advance their own RL research. They hope that making it available to a wider community will benefit more AI developers.
OpenAI is hoping to get feedback for the beta version of their AI gym so the company can make adjustments and make it a better research tool. If you're an AI developer, perhaps it's time to hit this gym.
