UKtech50 2019 winner: Demis Hassabis, co-founder and CEO, DeepMind
Computer Weekly profiles the achievements and journey of the founder of one of the world’s pioneering artificial intelligence companies, after his recognition as the most influential person in UK technology for 2019
Just as the queen is the most important piece in chess, with its ability to move anywhere across the board, DeepMind founder and chief executive Demis Hassabis is at the centre of an emerging world order where artificial intelligence (AI) will impact every job, industry and country.
A world-class chess player himself, Hassabis is the winner of the 10th edition of the UKtech50, Computer Weekly’s annual list of the most influential people in UK technology. A researcher, neuroscientist, video game designer and entrepreneur, Hassabis founded machine learning AI startup DeepMind in 2010, alongside his friends Shane Legg and Mustafa Suleyman. The startup was acquired by Google in 2014 for about £400m.
A child prodigy, Hassabis was taught how to play chess at the age of four, and by age 12 he was a chess master, representing England in tournaments around the world.
“If you have a quiet, reflective personality like I had when I was young, you can’t help but think and introspect about what it is about your mind that allows you to come up with these moves in such a complex game as chess,” he said.
Hassabis started programming on a ZX Spectrum 48K computer at the age of eight, his first achievement at coding being an application that could play chess. This was what led one of the most respected entrepreneurs in the world onto the beginning of his path towards AI.
Shortly after graduating in computer science at the University of Cambridge in the late 1990s, Hassabis went to work at games developer Lionhead Studios, where he was the lead AI programmer for the game Black & White.
That brief stint was followed by his first experience as an entrepreneur, as the founder of his own games venture, Elixir Studios. The company released a number of games with various levels of success during its lifespan of about seven years, before selling its intellectual property and folding in early 2005.
A focus on neuroscience
There are only two subjects worth studying, according to Hassabis: physics and neuroscience. The latter was the subject he embraced in his PhD from University College London (UCL), as Hassabis returned to the academic world following the end of Elixir.
“While physics is all about explaining the external world, including the entire universe, neuroscience and psychology is, conversely, all about explaining what’s inside our internal world,” Hassabis told the Google Zeitgeist event.
“One of the things I’m excited about with artificial intelligence [is that] I think it will help us understand our minds better”
Demis Hassabis, DeepMind
“When I thought about this more, I came to the conclusion that the mind was more important, because, obviously, that’s the way we interpret the external world out there,” he added.
“And that’s one of the things that I’m excited about with artificial intelligence, as I think ultimately it will help us understand our minds better,” said Hassabis, citing one of his greatest scientific heroes, US theoretical physicist Richard Feynman, who defended the view that, to understand something, one has to be able to recreate it.
The years between the end of the 2000s and start of the 2010s were academically intense for Hassabis, who also earned a Henry Wellcome postdoctoral research fellowship to the Gatsby Charitable Foundation Computational Neuroscience Unit of UCL.
He focused on the field of autobiographical memory and amnesia, and produced a number of highly influential papers on related subjects. Along with recognition in the academic world, where his work is seen as achieving scientific breakthroughs, Hassabis contributed to bringing themes related to the functioning of the brain to the mainstream.
Advancing artificial general intelligence
The academic work in UCL led to a collaboration between Hassabis, Legg and Suleyman towards what would become DeepMind, in 2010. The company was born under the premise that neuroscience, combined with AI and machine learning, as well as other emerging technologies, could lead to powerful algorithms learning to the concept of artificial general intelligence (AGI).
Setting up DeepMind was also an opportunity for Hassabis, who earned a CBE for services to sciences and technology in 2018, to work once again with his friend and partner at Elixir, David Silver, who came on board to contribute with his games expertise.
The company gained prominence by training learning algorithms to master games, having notoriously achieved automated game-playing at a “superhuman level”, by using the raw pixels on the screen of Atari games as inputs.
“AlphaGo successes hinted the possibility for general AI, to be applied to a wide range of tasks and areas to perhaps find solutions to problems that we as human experts may not have considered”
Demis Hassabis, DeepMind
“I think AlphaGo successes hinted the possibility for general AI, to be applied to a wide range of tasks and areas to perhaps find solutions to problems that we as human experts may not have considered,” Hassabis said in an interview published by DeepMind.
The concept of AGI developed by the team led by Hassabis has, in recent years, accomplished a number of other breakthroughs in areas beyond training computers to play games. This has included advancing research on AI safety and the development of a partnership with London’s Moorfields Eye Hospital for the use of artificial intelligence to identify and treat degenerative eye conditions.
More recently, the company has been delving even deeper into investigating some of the most serious issues around the future of health, including work around learning to predict the 3D shapes of proteins, elements on which the human biological machinery is built.
The company has already made progress with AlphaFold, its algorithm that can predict protein structures, and has inspired other systems that can predict the 3D structures of proteins from their amino-acid sequences. By cracking what leads to anomalies on proteins, Hassabis hopes to shine a light on paths that may lead to the design of new proteins, thus fighting disorders such as Alzheimer’s and diabetes.
Ethical questions
Reflecting the growing ethical questions surrounding the introduction of AI technologies, DeepMind has also encountered controversy in its dealings with the NHS.
In 2016, the UK Information Commissioner’s Office (ICO) investigated a patient data-sharing deal that gave DeepMind access to the healthcare records of 1.6 million patients that pass through three hospitals in North London, which fall under the care of the Royal Free Hospital Trust. The ICO found that the arrangement failed to comply with UK data protection laws.
Inevitably, Hassabis will find himself pushing the social and legal boundaries of AI, as well as being a technological pioneer. But continuing his chess metaphor, he hopes that humanity will be able to discover “some beautiful new moves and beautiful new ideas that can then go to another level”.
“If you were to take us back to our hunter-gatherer days and say that one day we’re going to build Manhattan and then fly over from London to Manhattan on a 747 regularly, that would be mind boggling – and yet, humanity’s done that,” he said.
“I don’t think we stopped enough to think how amazing that is, because the human brain is incredibly adaptable and as soon as we do something, it becomes kind of boring and mundane, but when I take a transatlantic flight I always think about how we, with our monkey brains, managed to come up with such reliable technologies,” he added.
“And so, if you think about that, then if we now build something like AGI and enhance our own capabilities with this amazing tool, I feel like almost anything might be possible.”
Demis Hassabis: 'I am honoured to receive this award'
DeepMind CEO Demis Hassabis gave Computer Weekly the following statement in response to being selected as the UKtech50 winner for 2019:
I’m so honoured to receive this award in its tenth year. I think it’s a great reflection of the hard work everyone at DeepMind has been doing throughout 2019. We know we’re following in the footsteps of some amazing winners over the past decade. On behalf of everyone at DeepMind, I’d like to say a big thank you to the judges and everyone who voted.
It’s been a really exciting time for DeepMind, and we’ve seen a number of research breakthroughs. This time last year, our program AlphaZero taught itself from scratch how to master the games of chess, shogi (Japanese chess), and Go, beating a world-champion program in each case. There was also our work on Capture The Flag, where our team trained AI agents to cooperate with human players, and AlphaStar, which achieved unprecedented standards of play in the complex real-time strategy game StarCraft II.
I know many people wonder why we choose to use games like these for our research.
This all comes back to our mission as a science company - using games to test and expand the capabilities of our algorithms so we can apply them to real-world scientific problems. Take for instance our program, AlphaFold, which uses similar techniques to those found in AlphaZero and last year confounded expectations by winning an international protein folding tournament. Protein folding is a crucial challenge if researchers are to one day discover treatments for diseases believed to be caused by misfolded proteins, such as Alzheimer’s, Parkinson’s, Huntington’s and cystic fibrosis. Likewise, some of the same techniques that were used in AlphaStar have gone on to help improve the training models of Waymo’s self-driving cars.
It’s also a scientific mission that DeepMind is proud to be pursuing in the UK. This country has a history of innovation in this area - and today brings together incredible, diverse talent from all over the world. We value the UK’s world-leading universities and multicultural environment, which allows companies to build the interdisciplinary teams they need to thrive.
Thank you Computer Weekly for the work you do explaining and analysing technological advances to your readers and thank you again for this award. Congratulations to all the other winners - here’s to more exciting breakthroughs in 2020.
