Alphabet’s DeepMind hires US team to work on Google products

Bloomberg

DeepMind, the London-based
division of Alphabet Inc. that’s responsible for numerous recent breakthroughs in artificial intelligence, is hiring its first researcher in the US to boost collaboration across the Atlantic.
The “applied research scientist position” described in a job posting on DeepMind’s website will be located at Google’s headquarters in Mountain View, California. It would be the first of “a couple dozen” the company hires, a spokesman said, as DeepMind builds “a small team” in the US “to bridge the gap between Google and our team in London, helping us collaborate even more closely to bring our research breakthroughs to Google users around the world.” DeepMind has not previously had any researchers based outside of London. “The Applied team in Mountain View will be made up of a mixture of software engineers and research scientists who work together to solve real-world problems at Google-scale,” according to the job listing.
The new hires will have their own offices at Google’s headquarters. Google’s other artificial intelligence research division, Google Brain, which uses machine learning to improve several Google products, also has offices on that campus. Both vie for the highly competitive field of artificial intelligence experts.
DeepMind is divided into two divisions, one of which does pure computer science research and another, called its applied division, which tries to build real-world products and services based on artificial intelligence. About three-quarters of DeepMind’s staff of 350 work for the research division, while the rest work for the applied unit.
The applied division has partnered with the UK’s National Health Service on two pilot projects, including one that uses AI to read eye scans. But its primary customer has been Google itself.
The division used machine learning to figure out how to improve the efficiency of Google’s data centers around the world, resulting in a 40 percent reduction in the amount of energy needed to cool the facilities and a 15 percent improvement in overall power usage efficiency, which measures the power consumed by the data centers’ computers compared with the total energy the facility uses.
DeepMind co-founder Mustafa Suleyman, who heads the company’s applied division, has talked about using similar techniques to try to bring efficiency gains to national electricity grids. The company has said its AI technology has also been incorporated into unspecified improvements in how Google displays advertising and in the shopping experience in the Google Play store, which sells apps for Google’s Android operating system.
In the future, Google may seek to incorporate DeepMind’s recent breakthrough in computer-generated speech, using a technique in which the AI system correctly predicts how to form the next part of a sound wave, to Google products such as its personal digital assistant. The newly advertised job’s responsibilities include making reinforcement learning -— which uses rewards, such as points in a game, to teach an AI program how to perform a task — more relevant to solving real-world problems.