Five things we learned about DeepMind's Demis Hassabis on Radio 4s Desert Island Discs

by

DeepMind co-founder Demis Hassabis found his time at Cambridge University fun (Source: DeepMind)

DeepMind, the British tech company bought by Google for millions, has more often than not let its work speak for itself.

But, co-founder Demis Hassabis has now shared insight on his life as an early chess champion, popular games designer, scientist and successful startup founder speaking on Radio 4s classic show Desert Island Discs.

Here's what we learned...

1. He has unusual sleeping habits

Asked if the he had any unusual habits, Hassabis detailed the way he sleeps - which means not doing so until 4am.

"I generally sleep at around four in the morning, I'll get into work around 10am. I'll do a full days work in the office, come back for dinner, I spend a bit of time with the family and then start a second days work at 10pm or 11pm and then go on to the small hours of the morning."

He added that this time was normally reserved for research, reading academic papers and creative thinking having learned "to maximise the way I think".

2. He's a fan of Blade Runner

Fulfilling every stereotype one might have of a man who works in the world of AI perhaps, the classic Ridley Scott robot movie has influenced the scientist and entreprenuer.

"I just love everything about this film and it was very informative for me when I saw it as a teenager because it brought artificial intelligence to life in this really beautiful visual world - sumptuous visual world - I would say," said Hassabis, adding that he loved the famous speech at the end of the movie.

3. Going to Cambridge was "like a holiday camp"

Heading to Cambridge after designing a hugely successful computer game as well as being one of the world's top chess champions as a kid, Hassabis' time at the prestigious university was almost a break.

"When I got to Cambridge, I had an amazing three years. It was the best three years of my life, because I think for me it was like a holiday camp. I'd basically been working my whole life," he said.

As a professional chess player, he never had a "normal summer holiday" and worked rather than backpacking between taking his A-Levels and heading to uni.

"I was determined at Cambridge to have a great social time as well as learning a lot of things," he said, choosing Prodigy to represent the times he came home after partying all night.

He still managed to graduate with a double first.

4. On robots taking over

"I think there's a lot of research left to go but we have to think about what goals we give these systems, what values we give these systems and how we make sure that they stick to the goals that we give them. And the way I think about AI is as this amazing tool that we can use to enhance our own goals as humans," he said. 

Hassabis believes AI will be built with benevolent intent, understanding in the years to come "the types of control systems required, how to check and interpret what these systems are doing, and I think we will solve those problems on the way to building AI".

He also noted the small community of researchers around the world currently able to build such systems meant there would less likely be any overtly negative programming.

"Once we understand these systems better and understand what we would be legislating for, there would be some stronger form of governance that would be agreed by world governments," he said.

5. On Google's acquisition

Hassabis did "a year of due diligence" on how the relationship would work after being acquired 

"For me it was nothing to do with the money, in fact our investors mostly didn't want to sell even though it was a lot of money, it's because I was convinced that by joining forces with Google we could accelerate the progress of the mission," he said.

"By using Google's power and resources we could explore this space of algorithms much faster, I could hire more research scientists and the whole kind of mission and the whole research programme would accelerate. And that's what's happened over the last three years." 

Listen to the show in full here.

Tags: DeepMind