Can AI really solve all diseases by 2040?

Nobel Prize winner Demis Hassabis has an extraordinary ambition for AI

Image generated by Google’s Gemini AI program

It was 3am one morning this week, and I was wide awake, insomnia being one of the more aggravating symptoms of my Parkinson’s. Convinced I was not getting back to sleep, I decided to finish listening to a fascinating podcast I had begun earlier.

It was an interview with one of the most prominent figures in artificial intelligence Sir Demis Hassabis, co-founder of the British startup DeepMind, which was sold to Google in 2014 and is now right at the heart of its AI operations. Under him, DeepMind first defeated the world champions of the impossibly complex Chinese game Go and then, as if to prove those cynics wrong who said “yes but show us something more serious than just playing a game” unlocked the mysteries of protein folding.

Before AlphaFold, as DeepMind called its project, researchers working on new drugs or other biology challenges might spend four years working out the precise structure of a protein. Now it was a matter of moments and for this achievement Sir Demis and his DeepMind colleague John Jumper were awarded the Nobel Prize for Chemistry.

He is also unusual for a scientist in being a great communicator and in an interview which roamed far and wide, he spoke of how his childhood prowess at chess led him to want to understand how computers played the game. He explained what a big breakthrough deep learning and reinforcement learning had been - IBM’s Deep Blue had used a brute force approach to beating the chess world champion Garry Kasparov, analysing every possible move. That would have been impossible with the massively more complex game of Go, so AlphaGo learned the game by playing it millions of times, often against itself.

But it was his answer to a question right at the end of the interview about what is coming next for AI which really took my breath away. He said medicine was where he expected to see rapid progress: “I think maybe in the next 10 to 15 years, we can actually have a real crack at solving all disease..” His hope was that in 10 to 15 years “we will look back on the medicine we have today a bit like how we look back on medieval times, and how we used to do medicine then.”

Wait a minute - was he saying that within 15 years Parkinson’s and every other disease would be sorted and that people like me could stop worrying about them? That was a very bold claim, especially when you look back on the last 15 years and find that really very little progress has been made in “solving” Parkinson’s.

But then again Demis Hassabis has a good track record with his predictions. There was huge scepticism when DeepMind set off on its mission to defeat Go, something most experts thought would not happen before 2030. And when in an interview in January 2020 for my book Always On I asked him the “what next for AI in the coming decade” question, he gave this reply:

“What I'm hoping you're going to see in the next phase of AI - especially for us - is some huge breakthroughs in Nobel Prize level winning problems.”

By 2024 he had already ticked off that lifetime goal.

So maybe, given the speed at which AI seems to be advancing right now, his prediction is not that fanciful. But it depends what you mean by '“solving” diseases. Advances like Alphafold should accelerate the work of researchers trying to understand what is happening inside the brain of someone with Parkinson’s and should make the process of identifying molecules which are likely candidates for disease modifying drugs a lot faster.

But as we Parkies have seen with the long and so far fruitless pursuit of a wonder drug for our condition, however quickly researchers identify a potential treatment, the laborious process of going through clinical trials does not seem to be getting any faster.

And while existing AI healthcare innovations, for instance algorithms to triage scans and identify malignant tumours, have been making headlines for at least a decade, few have been deployed at scale in hospitals or GP surgeries.

I am excited by the vision painted by Sir Demis Hassabis, intrigued to see what Isomorphic Labs, the drug discovery firm spun off from Google DeepMind, achieves over the next few years and grateful that the UK has such important work taking place here. But let’s hope that our healthcare system also goes through a frenzy of innovation over the next decade so that, if and when AI does deliver solutions to every disease, it is ready to embrace them.