At 13 he was described as a child-prodigy, being the top chess player in the world at his age group. Since then Demis Hassabis, the Londoner of Famagusta origin, embarked on a course in which he has been carefully selecting ever more demanding mental challenges, methodically fulfilling the mission he has taken on for himself.

The chess gave way to sim games developing and gaming world championships. Then it was computer science at Cambridge and that was followed by a PhD in neuroscience by UCL, a purposeful plunge into the deepest secrets of the human brain.

The next step was DeepMind, a rather small Artificial Intelligence (AI) company, which gradually started growing in London, away from the spotlights. Until January this year, when DeepMind grabbed the headlines – a rumoured £400m offer by Google made Demis’s and his partners’ company the largest and most ambitious business acquisition of the American giant in Europe.
The 37-year old Demis Hassabis speaks to CNA about his work, his dreams and his Greek Cypriot roots in one his rare interviews from his company’s headquarters tacked away in an unassuming central London building.

As he says with disarming self-confidence, all he has done since a young age were conscious choices, part of a plan that has led him to where he is today. He leads a unique AI company which tries to understand and recreate functions of the human mind and from there to build thinking machines. “If you look at what I’ve done, it looks quite disparate, many things. But everything I did, my games, my neuroscience PhD, all of that, was to gain all the right experiences and know how to build what ultimately became DeepMind. Of course I didn’t know it was going to be called DeepMind at this particular instance of it, but everything I chose to do, all the games I wrote, the early ones with Peter Molyneux and the big famous games like ThemePark, all of them had lots of AI in the games,” he says.

“I was gathering lots of information and experience in how to run companies, how to run teams of engineers, very smart engineers, how to work on really ambitious projects, how to manage all that. All of those things I have taken to DeepMind as well as ideas for what technologies we should use, what the road map should, so all these issues, how to do research I king of learnt through all the other experiences,” adds Demis Hassabis.

His work is complex, but he has learnt how to explain what he wants to achieve. “If we can actually understand and recreate intelligence, then we could use that technology to help us scientists and so on to solve all the other problems that we want to solve, whether that is understanding economics, understanding disease or understanding climate. Any of these really complex systems that as a society we want to be able to understand and master,” he tells CNA. It is, he explains, what he calls a “meta-solution”.

So applications of his AI work could be valuable in financial trading, in climate prediction, “anything that involves high dimensional complex data, but it could even be in the future about things like drug discovery, medicine discovery, where you have many genomics data and you’re trying to figure out what’s the next experiment you could do, this kind of things. And then obviously it could be used in robotics as well, having robots that actually think.”

In pursuing this goal, “a personal dream of 20 years”, he is working on building machines-computers that are able to do what is called “unsupervised learning”. As he explains, “we’re interested in algorithms where instead of programming them how to do things, which is like normal software, we actually program them with the capability to learn for themselves how to do things.” His team uses two special AI techniques, ‘deep learning’, from which the name of his company derives, and ‘reinforcement learning’. “We were the first people to really combine those together in a very fundamental way,” he says of his team, a group about 80 minds strong. “More than 50 of them have PhDs from all the top universities, MIT, Harvard, Stanford, Cambridge, Oxford, UCL, Imperial, all the top list. So I’ve cherry picked the best for these places, with this in mind, to crack these hard problems of AI. It’s kind of a unique organisation.”

So, is his research leading to something that would be on higher level than the human mind? “Potentially,” he says. “The human mind is constrained by biology. That’s why we have a fallible memory, for instance. There are many things that could potentially be improved about how the mind works. At some point there is no reason not to think why these technologies, certainly in some aspects, would be better than the way humans can do things. We’ve already seen that in some very narrow domains, like playing chess. But what we are trying to build is a more general programme that can do this across a wider range of tasks. This is going to be a different kind of intelligence and in theory you could imagine an artificial programme that would be more powerful is some ways.”

His research is also a personal bet, another personal journey through knowledge. “The other reason I’m so fascinated from a more personal point of view is because from a very young age I wanted to understand how our minds work. I mean who hasn’t thought about how your mind does all the incredible things that it does? One my favourite scientists, Richard Feynman, a brilliant physicist said an interesting quote. He said ‘what I cannot build, I cannot understand’. It’s a very true statement. Building something is the ultimate expression of understanding that system. In that sense trying to build an artificial mind I think is the best way of investigating some of the mysteries of our own minds, like what consciousness is, what dreams are for, what creativity is.”

For Demis the acquisition of DeepMind by Google is primarily a vehicle that will allow him to concentrate on what he is really passionate about. “The reason we teamed up, that we joined Google was because we felt we could turbo-charge the mission. With Google’s scale behind us, their computers, all the data, all their resources, all their engineers, it allows us to really focus on the research programme which is the thing I’m passionate about, rather than running companies, which is fun as well, and I like doing that, but it’s time consuming, trying to raise money and all that. Now, I don’t have to worry about that any more, I just concentrate on the research.” He smiles like a boy who opens a door to a room full of presents. “I’m so excited I can barely sleep these days!”

His smile becomes even wider when talk comes to his Cypriot roots. Although he doesn’t speak Greek (“I understand some Greek, because my grandfather lived with us for a long time and he spoke only Greek with my father”), he says he is proud and passionate about his Greek origin. “It makes me feel well that Cypriots are proud of me. I have a Greek way of thinking and what Greece has done for western world from classical times is a huge part of my character, all these inspirational things. My son is called Alexander Ulysses, because I had to give him the Latin version, my wife is Italian. Odysseus is my favourite classical character; he was the smart one, right? And Achilles. It’s just so inspirational.”
His grandfather’s family are from Famagusta. He and his grandmother from his father’s side were both Greek Cypriots. “They had a very nice house in Famagusta. They used to work in London to try and pay for building this dream house, on the beach. That was always their dream. They were in London and they thought that they were going to be here for like ten years or something and then go back eventually. Of course the invasion happened, so then they couldn’t go back. My father grew up as an only child in London. And obviously I was born in London. We used to go back a lot to Greece and all the Greek islands, and I love those places – I recently had holidays in Crete actually. We have a lot of contact with the Greek side of things.”

Demis Hassabis works almost all day and he says he concentrates better in the small hours. But all this hard work has paid dividends. “This is just the beginning in some ways, it’s kind of the end of the beginning, let’s put it that way. It’s the platform I’ve always dreamt about from here. With this platform we have now it’s an incredible chance to do something really amazing, you know. Usually when you plan things out they never work as well as this has worked. For some reason with DeepMind everything turned out maybe better than how we were planning it, which never happens with me. I don’t know if that means we were exactly the right time with the right idea with the right team and all these 20 years of planning maybe made a difference, maybe it was useful for something, all this thinking.”

CNA