In a new collaboration, Google’s DeepMind and the National Health Service will use artificial intelligence to help doctors detect cancer and detect acute kidney injury. These two entities have been working together on many projects aimed at improving patient care. Recently, DeepMind and the NHS announced they will use anonymized medical records of 1.6 million patients. This information was not disclosed to the patients, but DeepMind says it is essential for their research.
As of Sept. 2018, Fergus was the manager of FAIR’s New York office. He was looking for people to join his research team and talk about its strong academic connections. While the company talks about open sourcing their code, it hasn’t always been that way. He also says he’ll be conducting technical research on AI safety. However, he says it’s an opportunity to help protect humanity.
As an example, DeepMind helped Google predict the output of its wind farms 36 hours ahead of time. They trained their neural network by using historical turbine data and local weather forecasts. They then recommended optimal hourly delivery commitments to the power grid. The results increased Google’s wind energy value by 20%. The next step is to improve the model further to make it commercially viable. If this partnership proves successful, the team will be working to develop other uses for artificial intelligence.
The DeepMind team has worked with the NHS in the past, but this time, they’ve teamed up with the Royal Free Hospital to develop a mobile application to detect AKI. Patients can upload their blood tests using DeepMind’s software and the system will analyze the results and alert appropriate staff. While critics are concerned about the privacy implications of such an undertaking, DeepMind has defended its privacy and cybersecurity capabilities and promised patients the right to opt out of the project.
The partnership will allow researchers to test their AI-based algorithms on a wide range of games, including StarCraft 2. The game is an immensely popular real-time strategy game based on a futuristic world and three alien species. It first came out for windows in 1998 and became a hit for millions of players. It won the best-of-five match 4-1 in March. If this partnership proves successful, it will be even more powerful.
DeepMind is using machine learning systems to train the AI to understand natural language. During the development of its AlphaGo program, DeepMind’s engineers combined supervised learning and Monte-Carlo tree search. The result was a machine that beat Lee Sedol in four out of five games of Go. DeepMind’s AlphaGo program has been used for image and speech recognition and is replacing 60 rule-based systems at Google.