The quest for a cancer cure dates back thousands of years. Some of the earliest known research dates to ancient Egypt, where Imhotep, the physician and architect to King Djoser, described a human tumor on papyrus around 2600 BC.
Now, a growing chorus of tech leaders is singing the praises of AI as the key to solving the medical mystery that has puzzled physicians for millennia. It’s what Google president Ruth Porat predicted last October. And it’s why Anthropic CEO Dario Amodei coined the term “the compressed 21st century,” reflecting his view that AI will accelerate medical progress. But some in the medical field think that forecast is at least a bit overshot.
In a recent interview on the Plain English podcast with Derek Thompson, Eli Lilly CEO David Ricks said AI is far from curing the disease.
“If you just ask them to solve biology or chemistry questions, they’re not particularly good at it,” he said. “They’re trained on the human language, not on the language of chemistry, physics, and biology.”
One reason AI investment has reached record levels, rivaling the GDP of some developed countries, is the belief that the technology could enable revolutionary scientific breakthroughs. During President Donald Trump’s press briefing announcing the Stargate Project last year—a $500 billion investment in AI infrastructure through 2029—Oracle executive chairman Larry Ellison said the project could lead to a cancer vaccine, one that could be devised within just 48 hours.
The current reality of AI cancer research
While Ricks has some doubts about AI’s scientific research capabilities, several AI models have made significant advancements in cancer research. Harvard’s Sybil AI model in 2023, for example, accurately predicted lung cancer risk within six years.
And Google DeepMind’s AlphaProteo model has proved instrumental in designing protein binders that target certain molecules, including those associated with cancer. In fact, Eli Lilly uses AlphaFold, another AI system developed by Google DeepMind, and maintains a partnership with it.
But Ricks said current AI capabilities are only a drop in the bucket compared with the need for additional scientific research. “We can get a machine to predict things pretty well, like predicting the structure of a protein,” he said. “But that is one maybe 1,000th of the kind of problems we face in drug discovery.”
The Eli Lilly CEO is placing his bets on tailored AI models to seal the deal on scientific advancements. During the interview, he noted that most LLMs fail to master the nuance required to deal with biology, something he thinks models trained on advanced and specific data could one day achieve.
“The future here is actually to build more and more models of those narrow prediction problems because biology unlike, human language, doesn’t follow all the same rules in the same way,” he said, similar to Google DeepMind’s AlphaFold and AlphaProteo.
Still, Ricks thinks humans, with or without AI, are quite far off in biological R&D, despite the advancements in medicine already achieved. “We’re sort of like a toddler in the language of biology,” he said.
