So, does Google own DeepMind? In fact, it did, but the company did not want to disclose the details of the deal. As it turns out, DeepMind is an AI company that makes money selling its technologies to Alphabet, the company’s parent. Last year, Alphabet raked in about $182.5 billion in revenue.
DeepMind, meanwhile, lost around $680 million per year. In 2019, Alphabet wrote off DeepMind’s debt of about PS560 million and trimmed its losses by nearly PS2 billion. However, by 2021, DeepMind’s revenue is set to increase by 8%, making it the biggest AI firm in the world.
While Google is not explicitly confirming or disowning DeepMind, the company is denying that it owns it. While Google may have lost two Ethical AI researchers in the past year, the company is insisting that DeepMind retain larger autonomy. The company also denies any intention of disseminating its research to the public. If DeepMind is indeed owned by Google, it will be the company’s most important asset.
In addition to its commercial use, DeepMind also develops algorithms to help keep its data centres cool. The company claims that its algorithms have reduced Google’s energy bills by 40%, but it has already been using algorithms to optimise its data centres for years before DeepMind was formed. Although DeepMind is based in London, Alphabet pays DeepMind handsomely for its services. In 2017, DeepMind bill Alphabet companies for PS54 million, and spent PS200 million on staff.
The company has also pushed DeepMind to commercialise its work. In their first year after the acquisition, DeepMind reported no revenue. The company’s WaveNet paper, for example, is a groundbreaking approach to speech generation that is being deployed in Google Assistant. Other DeepMind projects include developing an AI system for data centers that can reduce energy bills and cooling costs. With all these efforts, DeepMind has been the focus of headlines.
While DeepMind claims to be independent, the ATRC ethics board is a big concern. DeepMind’s CEO David Feinberg said the company’s work on health care is “inextricably linked” to its other projects. It also plans to expand DeepMind Health’s services to other countries. While this may sound scary, it’s not surprising that Google owns the AI company. It has made major strides in the field of artificial intelligence in the last year. However, privacy campaigners argue that Google has broken its promises about keeping data private.
The company’s AI department is working with the U.S. Department of Veterans Affairs. This project is working with 700,000 historical medical records to determine if machine learning can identify risk factors for patient deterioration and improve treatment. Patients who are prone to patient deterioration account for 11 percent of all deaths in hospitals. DeepMind has also collaborated with Moorfields Eye Hospital to develop an algorithm to detect sight-threatening diseases.
The AlphaGo program created by DeepMind beat world champion Lee Sedol in the Go game by using supervised learning, reinforcement learning, and machine learning. The AlphaGo AI also beat 60 rule-based systems at Google and is being used for fraud and spam detection. With AlphaGo, Google has proven that AI can mimic human intuition and critical thinking. It was developed using a technique known as deep reinforcement learning. However, the company still maintains full ownership of the company.
In the demo, Hassabis leaves some crucial details out. The AI program cannot respond to tiny changes in the environment. To be effective, it would take thousands of reinforcement learning rounds to develop a program with high diagnostic and mechanical intelligence. Further, it is difficult to release a program that is perfected in a virtual space. In such a case, Hassabis has questioned whether Google is the right company to release the technology.
Hassabis’s AI team has reportedly been studying the human mind for the past decade. Its goal is to develop a general-purpose AI that could handle any task. Hassabis says this AI team will change the world. However, the results are not yet conclusive. But it is certainly promising for the future of artificial intelligence. It will help the company improve the quality of life for humanity.
While the company has not yet released the details of its AlphaCode program, DeepMind has announced that it will use this AI program to tackle competitive programming. This game requires a long problem description, which competitors then use to write a program that answers it. In addition to these programs, DeepMind would also analyze anonymised eye scans to find early signs of blindness. These programs seem to be quite a departure from their previous Alpha programs.