After three decades having led some of Silicon Valley’s most powerful companies – co-founding LinkedIn and serving on the boards of PayPal and OpenAI – Reid Hoffman has recently turned his attention to healthcare.
Hoffman’s startup, Manas AI, is building an artificial intelligence engine that aims to accelerate the traditionally leisurely process of drug discovery for various cancers. Inspired by a dinner with renowned oncologist Siddhartha Mukherjee, co-founder and CEO of the company, mission statement is to “shift the drug discovery process from a decade-long process to a multi-year process.”
But Hoffman’s enthusiasm for generative AI in particular goes far beyond drug discovery targets and petite molecules. He believes that frontier models – the most advanced large-scale AI models available today from companies like OpenAI and Anthropic – should be the cornerstone of healthcare itself.
“If you as a physician don’t use one or more of the pioneering models as a second opinion, I think you’re bordering on committing malpractice,” Hoffman said during a speech at WIRED Health in London on April 16. “These AI systems, even though many of them are not specifically trained for medicine, have taken in over a trillion words of information. They provide a second opinion with superpowers that no human has.”
Such comments will undoubtedly upset many doctors. Earlier this year important examination concluded that vast language models pose a risk to the general public seeking medical advice due to their tendency to provide wrong and variable information.
Hoffman’s argument is that instead of outsourcing critical thinking skills to artificial intelligence models, humans should operate them as an additional source of information, which he believes can prevent misdiagnosis. He says he personally uses boundary models as a second opinion on matters related to his own health, and insists that his personal concierge doctors do the same.
“You might as well say, ‘No, I think you’re wrong, I think this is it,’” he told WIRED’s health audience. “But if you don’t use that as a second opinion, you’re making a mistake, both as a doctor and as a patient.”
As the UK’s National Health Service struggles under the weight of long waiting lists and workforce challenges, including chronic lack of family doctorsHoffman believes there is an increasingly urgent need to develop a model with a vast tongue that could act as a free medical assistant on any smartphone. He suggests it could also serve as a form of early screening before visits to human doctors.
“We just don’t have enough doctors, most people don’t have access to them, and when you think, ‘How should the NHS be redesigned?’ everyone needs to be connected to that medical assistant,” he said.
While Hoffman has a conflict of interest as an entrepreneur working in drug discovery, he also wants to see artificial intelligence play a broader role in helping the FDA and other regulators evaluate modern drugs, as well as making particularly promising drugs more available to patients.
“As a Silicon Valley resident, I would love to get to the point where the FDA would also do biological model testing and say, ‘Oh, we should speed up this test because there’s less chance of negative consequences,’” he said. “Do I think this will happen anytime soon? Unfortunately not.”
When it comes to Manas AI, human judgment continues to play a key role in a company’s decisions about what goals to pursue. Mukherjee carefully reviews his AI engine’s proposals, Hoffman says, and separates the truly engaging candidates from the “crazy fools.”
While the company’s initial focus was cancer, Hoffman believes the potential of AI discovery engines is much broader, enabling the identification of drug candidates for chronic diseases, but also extremely scarce diseases whose research has traditionally not been as economical for pharmaceutical companies.
“I think in 10 years, every major disease will have target molecules that can at least make a significant difference,” Hoffman said.