Some of Google’s most cutting-edge AI developments are coming out of DeepMind, and the company has picked its first research base outside of the U.K.: Edmonton.
The new research lab will have a close affiliation with the University of Alberta’s Alberta Machine Intelligence Institute, and will be led by three of its professors – Richard Sutton, Michael Bowling and Patrick Pilarski. DeepMind will also be sending seven of its researchers to Edmonton to contribute to the research being done there. U.K.-based DeepMind was acquired by Google in 2014.
Seeing Edmonton selected as an AI centre might be a bit unexpected for those that have been following the sector in Canada and have seen Montreal and Toronto more frequently pegged as hubs by research institutes, incubators and national companies alike. But Demis Hassabis, co-founder and CEO of DeepMind, pointed out in a blog post announcing the expansion that nearly a dozen University of Alberta graduates have gone on to work for DeepMind, and said the company has previously sponsored the university’s machine learning lab to provide additional funding for PhD students.
One of those graduates, David Silver, was a key developer in one of DeepMind’s most noteworthy accomplishments so far: AlphaGo. The software platform was developed to compete with human players in the Chinese strategy game Go – which has an average of 200 possible moves in a given turn and a nearly endless number of board configurations in total – as a way to test and prove the reasoning capabilities of artificial intelligence. AlphaGo was recognized with the Innovation Grand Prix in Cannes in 2016, and beat the world’s top ranked player earlier this year.
The University of Alberta’s Alberta Machine Intelligence Institute is also one of the recipients of funding from the federal government’s Pan-Canadian Artificial Intelligence Strategy, and has also received a 15-year investment of more than $40 million from the Alberta government.
The connection between DeepMind and the University of Alberta extends to their approaches to AI as well. Sutton is considered a pioneer of AI research and a world leader in the area of reinforcement learning – an area of machine learning inspired by behavioural psychology. Unlike other forms of “teaching” a machine that provide “correct” outputs or actions that it is taught to recognize, reinforcement learning instead allows the machine to determine its ideal behaviour by maximizing the reward in a given context. Reinforcement learning is also a major part of DeepMind’s own approach to AI, which has been more directly inspired by the way human intelligence works and developed neural networks and systems aimed at understanding things like context. Sutton was also an early advisor to DeepMind when it first launched in 2010.
Photo courtesy John Ulan/UAlberta
