AI in wrong hands could be ‘used for harm’, warns Sir Demis after Nobel win

The British computer scientist founded the AI start-up in 2010, which played a key role in helping solve the mystery of how protein structures form.

British computer scientist Sir Demis Hassabis has warned artificial intelligence in the wrong hands could be “used for harm” after he won the Nobel Prize in Chemistry.

Sir Demis, 48, who founded the London-based artificial intelligence start-up DeepMind, received the honour alongside American John Jumper, a senior research scientist at the company, for their “breakthrough” work on proteins.

In a separate news conference after the announcement, he described AI as a “very powerful” and “dual-purpose” technology which has “extraordinary potential to be one of the most beneficial technologies for humanity”, but at the same time, could also “be used for harm” by “bad actors”.

Sir Demis said: “We have to really think very hard – as these systems and techniques get more powerful – about how to enable and empower all of the amazing benefits and good use cases whilst mitigating against the bad use cases and the risks.”

He said while today’s AI systems are not dangerous, that might change in the future as these technologies get more advanced and acquire “human level” intelligence.

The Nobel Committee for Chemistry sitting beneath pictures of the winners — Johan Aqvist, Hans Ellegren and Heiner Linke, of the Nobel Committee for Chemistry, award this year’s prize to David Baker, Sir Demis Hassabis and John Jumper (Christine Olsson/AP)

Sir Demis said: “We are going to have to be more and more cognisant of these risks and we need to start the research on that in a really big way so that we are prepared for that.”

Along with Dr Jumper, Sir Demis contributed to the development of an AI model which helped solve one of biology’s biggest mysteries that has puzzled scientists for more than five decades: how do protein structures form?

They share the Nobel prize with David Baker, of the University of Washington, who pioneered the method for designing proteins.

In 2020, Sir Demis and Dr Jumper presented AlphaFold2, the AI model the company had developed to help predict the complex structures of proteins.

Since the 1970s, scientists around the world have been trying to work out how a protein folds into a unique three-dimensional shape.

With AlphaFold’s help, researchers have been able to predict the structure of virtually all the 200 million proteins that have been identified.

Since their breakthrough, AlphaFold2 has been used by more than two million people from 190 countries.

The hope is that knowing how proteins – the building blocks of life – work will help pave the way for development of novel drugs to treat diseases such as cancer, dementia and even Covid-19.

Sir Demis said it was “an unbelievable honour of a lifetime to receive the Nobel Prize”.

He said: “I’ve dedicated my whole life to advancing AI, because I really believe in the potential it has to improve the lives of billions of people.

“When we look back on on AlphaFold, it will be the first proof point of AI’s incredible potential to accelerate scientific discovery.

“And it is that potential that really got me into AI in the first place.

Johan Aqvist, right, of the Nobel Committee for Chemistry, uses a chart to explain the work of this year's prize winners, as fellow committee members look on — Johan Aqvist, right, of the Nobel Committee for Chemistry, explains the work of this year’s prize winners, David Baker, Demis Hassabi and John Jumper (Christine Olsson/AP)

“I see AI as, potentially, the ultimate tool for accelerating science and scientific knowledge.”

Sir Demis was born in London in 1976.

A child chess prodigy, he designed and programmed a multimillion-selling game called Theme Park in his teens before going to Cambridge University.

Sir Demis said playing chess from a young age led him to become “very interested” in artificial intelligence.

He said: “I would actually encourage kids to play games, but not just to play them – the most important thing is to try and make them.

“That is one of the best ways into programming and and engineering.”

He received his PhD from University College London, with the journal Science listing his research on imagination and memory as one of 2007’s top 10 breakthroughs.

Sir Demis co-founded DeepMind in London in 2010, which he sold to Google in 2014.

In 2017, he featured in the Time 100 list of most influential people, and earlier this year, he was knighted for his services to AI.

Sir Demis said he was having a “normal morning” when he received the call about the prize.

He said that in a “funny chain of events” his wife first received the call from the Royal Swedish Academy of Sciences, before they were able to reach him “only a few minutes before they announced it”.

The winners share a prize fund worth 11 million Swedish kronor (£810,000).