Google's DeepMind releases Lab AI platform

December 6, 2016 // 11:37 a.m.

Tags: #ai #alphabet #artificial-intelligence #deep-learning #deepmind #deepmind-lab #google #machine-intelligence #mi #openai #openai-universe

Google has followed the OpenAI initiative in launching a learning platform for artificial intelligence, created by AI-centric subsidiary DeepMind.

Google's DeepMind, acquired back in 2014 in a deal valued at £400 million, has been leading the company's considerable efforts in machine learning and artificial intelligence. Its most famous success came with the development of the world's first champion-level Go AI, which beat the world's greatest human player at a notoriously challenging game boasting significantly more possible moves than chess. DeepMind has also been behind a partnership with the NHS to analyse eye-scan data for early signs of macular degeneration more quickly and more accurately than human experts.

Now, Google's DeepMind is releasing its Lab platform under a permissive licence for others to use. Similar in concept to OpenAI's Universe platform, DeepMind Lab takes the form of a three-dimensional game-like platform with a first-person viewpoint in which artificially intelligent agents are able to move, interact, and learn.

'Artificial general intelligence research in DeepMind Lab emphasises navigation, memory, 3D vision from a first person viewpoint, motor control, planning, strategy, time, and fully autonomous agents that must learn for themselves what tasks to perform by exploring their environment,' the company explained in its release announcement. 'All these factors make learning difficult. Each are considered frontier research questions in their own right. Putting them all together in one platform, as we have, represents a significant new challenge for the field.'

DeepMind believes that teaching AI in a 'realistic' game-like environment may also be useful for the development of true general-purpose intelligence. 'As well as facilitating agent evaluation, there are compelling reasons to think that it may be fundamentally easier to develop intelligence in a 3D world, observed from a first-person viewpoint, like DeepMind Lab,' the company claimed. 'After all, the only known examples of general-purpose intelligence in the natural world arose from a combination of evolution, development, and learning, grounded in physics and the sensory apparatus of animals. It is possible that a large fraction of animal and human intelligence is a direct consequence of the richness of our environment, and unlikely to arise without it. Consider the alternative: if you or I had grown up in a world that looked like Space Invaders or Pac-Man, it doesn’t seem likely we would have achieved much general intelligence,' DeepMind's spokespeople added, in an apparent dig at the approach chosen by OpenAI of using off-the-shelf game software.

The company has published a paper on DeepMind Lab (PDF warning), while the software itself can be found on the official GitHub repository.
Discuss this in the forums