Access and Feeds

Artificial Intelligence (AI): DeepMind Leapfrogs from Playing Games to Optimizing UK’s Countrywide Power Grid

By Dick Weisinger

DeepMind is a technology from a UK-based technology startup that Google acquired in 2014.  DeepMind tries to combine machine learning and system neuroscience into general algorithms that can be used for learning.

Initially DeepMind was developed and prototyped simply to play games.  It was programmed to play simple Atari games. As it grew in sophistication it was applied to more advanced games.  In 2016 DeepMind beat a top-ranked player of Go.

Recent advances in DeepMind technology have allowed it to cache memory of the different types of games which it has come up against before.  Rather than starting from scratch when attempting to learn a new game, techniques learnt from previous games can adapted and applied.

James Kirkpatrick, Research Scientist at DeepMind, said that “our approach remembers old tasks by selectively slowing down learning on the weights important for those tasks,” the paper says.  explains that the algorithm selects what it learned to successfully play one game and keeps the most useful parts. We only allow them to change very slowly [between games]. That way there is room to learn the new task but the changes we’ve applied do not override what we’ve learned before”.

DeepMind algorithms are now being applied to general learning and decision-making scenarios.  DeepMind was applied to the problem of more efficiently using electricity to Google’s datacenters.  The results of the experiment were that DeepMind AI was able to save 15 percent on overall energy costs and more than 40 percent on the costs of cooling.  Google has proposed to UK, and it is being seriously considered, that the same techniques be applied to UK’s countrywide energy grid.

Digg This
Reddit This
Stumble Now!
Buzz This
Vote on DZone
Share on Facebook
Bookmark this on Delicious
Kick It on
Shout it
Share on LinkedIn
Bookmark this on Technorati
Post on Twitter
Google Buzz (aka. Google Reader)

Leave a Reply

Your email address will not be published. Required fields are marked *