How Google’s new AI plays video games like a human
What the IBM-made Deep Blue computer did with chess, Google Inc’s new DeepMind AI may soon do with video games. Google has developed a new AI capable of teaching itself how to learn and master video games in ways that could have big implications for both gaming and the future of AI in general.
Google acquired DeepMind Technologies over one year ago for $500 million, beating out a bid made by Facebook Inc, who had been “deep in negotiation talks” with DeepMind. The AI technology startup made important strides in machine learning, and a recent study published in Nature shows the progress the company has made since joining Google.
Titled “Human-level control through deep reinforcement learning,” the study examines an AI that was able to learn and master 49 different Atari 2600 video games. These include classic games such as Pong, Breakout, Space Invaders, and others.
While a computer playing against another computer might not sound revolutionary, the difference is the way Google’s AI learns. At the outset, the AI did not know anything about the games it played. Rather it learned them through experience, and from that experience it built the skills necessary to succeed at them. In other words, the AI learned the games the same way a human would—by playing them.
General-purpose AI is “many decades off”
DeepMind co-founder Demis Hassabis says that despite the advancements in AI, the technology still has a long way to go before it can truly think on its own.
“[The AI] is mastering and understanding the structure of these games, but we wouldn’t say yet it’s building conceptual knowledge or abstract knowledge,” Hassabis told Bloomberg. “The ultimate goal here is to build smart, general-purpose machines, but we’re many decades off from doing that.”
Hassabis says that the next step in DeepMind’s development of the AI would be to enable it to learn how to navigate three-dimensional games, which would involve spatial reasoning and objectives that are more complicated than Pong’s “bounce a square using a line” or Breakout’s “bounce another square using a line.”
DeepMind’s progress is impressive, but it will likely be some time before its AI is capable of deciding which Dragon Age NPC it wants its character to sleep with.
photo credit: littlelostrobot via photopin cc
Since you’re here …
… We’d like to tell you about our mission and how you can help us fulfill it. SiliconANGLE Media Inc.’s business model is based on the intrinsic value of the content, not advertising. Unlike many online publications, we don’t have a paywall or run banner advertising, because we want to keep our journalism open, without influence or the need to chase traffic.The journalism, reporting and commentary on SiliconANGLE — along with live, unscripted video from our Silicon Valley studio and globe-trotting video teams at theCUBE — take a lot of hard work, time and money. Keeping the quality high requires the support of sponsors who are aligned with our vision of ad-free journalism content.
If you like the reporting, video interviews and other ad-free content here, please take a moment to check out a sample of the video content supported by our sponsors, tweet your support, and keep coming back to SiliconANGLE.
