Two of the biggest names in artificial intelligence are opening their doors to everyone.
Google's DeepMind, the AI system that famously defeated the world champion of the board game Go earlier this year, and OpenAI, an AI collaboration from Elon Musk and others, are making their software platforms available to researchers, developers, and anyone else who wants to use them. The two companies made their announcements in separate blog posts Monday.
The two platforms both use deep learning, a form of machine learning that allows AI systems to grow smarter. In its post, DeepMind, which was acquired by Google in 2014, wrote that its goal is to create "systems that can learn to solve any complex problem without needing to be taught how." Training a system to be masterful at a board game, like DeepMind did with AlphaGo, is one thing. The goal, however, is to be able to feed the system a new set of instructions and enable it to adapt and learn on its own.
To train artificial intelligence, DeepMind uses a program called DeepMind Lab, a digital three-dimensional environment. DeepMind Lab will now be available for public use, so any researcher or developer can train bots to explore the video game-like world and complete tasks like navigating mazes and forgoing short-term rewards for longer-term ones. In the process of learning those human-like skills, the bots are supposed to become smarter over time.
OpenAI, meanwhile, was created from a $1 billion investment from Musk, Peter Thiel, Sam Altman, and other high-profile tech magnates with the intention of ensuring that artificial intelligence is used for good. The company is releasing a platform called Universe, which lets people train AI systems to perform essentially any task that a human can complete with a computer. The current release of Universe includes over a thousand environments, including games and web browser-based setups. Developers could conceivably use these platforms to make the AI in their own games and apps smarter.
The two announcements don't seem to be directly related. In fact, in its blog post, OpenAI called out DeepMind's AlphaGo as an example of AI being too narrow in its scope. "AlphaGo can easily defeat you at Go," OpenAI wrote, "but you can't explain the rules of a different board game to it and expect it to play with you." Google and DeepMind did not immediately respond to Inc.'s request for comment.
While OpenAI and DeepMind have in the past spoken about ensuring that AI is used safely, their intentions here aren't completely altruistic. Open sourcing an AI system allows it to access troves of new learning data, so the more people who use it, the stronger it becomes.