Tesla Motors’ Elon Musk has been trying to redefine transportation on earth as well as in space since a long time. And in 2015, Musk founded OpenAI, a nonprofit dedicated towards furthering A.I. research.
Since its launch two years ago, one of the most prominent achievements of Elon Musk’s artificial intelligence company is the OpenAI Gym that they launched in April last year. The Gym lets developers train their AI systems on games and challenges.
Earlier this week, Musk’s Open AI took to Twitter to announce the launch of Roboschool, an update to the last year launched OpenAI Gym that could help developers in preparing better robots for the real world.
According to OpenAI’s website, their ultimate aim is to build safe Artificial general intelligence (AGI), and ensure that the AGI’s benefits are as widely and evenly distributed as possible. The organisation believes that though AI technologies will be hugely impactful in the short term, but their impact will be outstripped by that of the first AGIs. In layman terms, they’re working on building a machine that has the full cognitive and emotional capabilities of a human, something which is almshouse similar to what see in science fiction movies.
It is important to note here, that most of the A.I. employed in the real world today only focuses on small, specific jobs, but AGI’s goal is to be all-encompassing. According to the organisation, the latest improvements to their software bundle can be seen as an crucial next step towards realising their goal.
In addition to the fun components such as building software to beat the high score on some old games, Open AI Gym can be put to use to understand how the various simulations of robots will perform under different circumstances. Some people are also known to have developed a method to take their simulations into the realm of physical robots navigating the real world, which allows them to make use of their software today in actual machines.
According to experts, the launch of Roboschool will make the simulated environments much more realistic and applicable to actual robots. The Roboschool comes along with an upgrade to Musk’s AI organisation’s physics engine, which makes it much easier for more variables and constraints to be introduced. For instance, the tracks that simulated robots had to walk along, are now available in three dimensions, and the robots have to rise up to task of balancing in spite of more forces, such as the torque being generated by robot’s own legs.
OpenAI Gym integrated Roboschool is currently providing 12 different environments for controlling robots. Through this, it is providing a way to train multiple agent at once in one environment. Tasks researchers can leverage include multiplayer Pong, having two or more legs on the ground,walking and providing humanoid benefits.