Google scientist: AI may not be ‘cure-all’ for discovering new drugs

in San Francisco PROMISES that artificial intelligence will create life-saving drugs faster than humans and healthcare may be overblown, one of Google’s top computer scientists has suggested.

Mark DePristo, who runs the genetics team at Google Brain, which is working on a computer that is able to think like a human brain, said the technology’s role in drug discovery was “in its winter”.

Billions are being bet on the next breakthrough drug using super-smart computers, but Mr DePristo suggested that artificial intelligence (AI) is not the cure-all many are hoping for. Excite- ment around the technology has seen a boom in Silicon Valley health start-ups that claim they will be able to reduce the average 12 years and $2billion pharmaceutical companies spend bringing a drug to market.

It is hoped that computers will be able to trawl genetic patterns and suggest ground-breaking alternatives to drugs that already exist, but are not effective on everyone. However, algorithms alone cannot overcome the obstacle of collecting data necessary to detect these patterns.

Mr DePristo said: “The problem is that you want hundreds of thousands of people and information about them … how do you collect that data? The analytics will involve deep learning [AI] techniques but at the end of the day, the fundamental problem is scaling these measurements. We struggle to do trials with thousands [of patients] and it is clear we need millions.”

The computer expert, who founded a company working on a blood test to detect autism before he was poached by Google, was speaking on a panel during the JP Morgan Healthcare Conference in San Francisco, where biochemist start-ups were out to convince venture capitalists to part with cash.

In 2017 returns on drug discovery investment dropped to its lowest point in eight years. No drug “discovered” by AI has yet to reach clinical trials.