AI executives promise cancer cures. Here’s the reality.
The technology is genuinely useful for scientific discovery, but its applications are less dramatic than you might think.
 

This is Atlantic Intelligence, a newsletter in which our writers help you wrap your mind around artificial intelligence and a new machine age.

Did someone forward you this newsletter? Sign up here.

Matteo Wong

Staff writer

The greatest promise of generative AI has never been to draft emails, generate PowerPoint slides, or even write detailed research memos. Rather, it’s to cure cancer—and executives at the helm of the industry’s top labs, including OpenAI (which has a corporate partnership with The Atlantic), Anthropic, and Google DeepMind, have promised that such a breakthrough could be achievable very soon. “I don’t see why not,” Demis Hassabis, the CEO of Google DeepMind, said on 60 Minutes last Sunday, when asked if AI could cure all diseases within a decade. A few days earlier, OpenAI released its highly anticipated o3 model and boasted of “its ability to generate and critically evaluate novel hypotheses” in biology, math, and engineering.

There are, in fact, many reasons why AI will not “cure cancer.” As I wrote this week, generative AI faces a number of significant roadblocks in biomedical research, including a lack of high-quality training data and propensity to assert falsehoods as true. Human judgment and insight remain far more important than applying a more advanced algorithm. And “even if, tomorrow, an OpenAI or Google model proposed a drug that appeared credibly able to cure a single type of cancer, the medicine would require years of laboratory and human trials to prove its safety and efficacy in a real-world environment,” I wrote.

Still, the technology is already accelerating scientific discovery. In conversations with researchers and executives at a number of universities, pharmaceutical companies, and research institutes, including Moderna and Pfizer, I learned exactly how generative AI is contributing to the study and treatment of disease today—as an incredible supplement to, but not a replacement for, human collaborators. “These AI models are to biologists like a graphic calculator and drafting software are to an engineer,” I wrote. Generative AI’s contributions are dazzling, but not science-fictional, in nature—these models are never going to simply “cure” even a single type of cancer, but they could, in the years to come, revolutionize how humans understand disease.

(Illustration by The Atlantic. Sources: H. Abernathy / Getty; Bettmann / Getty.)

 

To hear Silicon Valley tell it, the end of disease is well on its way. Not because of oncology research or some solution to America’s ongoing doctor shortage, but because of (what else?) advances in generative AI.

Demis Hassabis, a Nobel laureate for his AI research and the CEO of Google DeepMind, said on Sunday that he hopes that AI will be able to solve important scientific problems and help “cure all disease” within five to 10 years. Earlier this month, OpenAI released new models and touted their ability to “generate and critically evaluate novel hypotheses” in biology, among other disciplines. (Previously, OpenAI CEO Sam Altman had told President Donald Trump, “We will see diseases get cured at an unprecedented rate” thanks to AI.) Dario Amodei, a co-founder of Anthropic, wrote last fall that he expects AI to bring about the “elimination of most cancer.”

These are all executives marketing their products, obviously, but is there even a kernel of possibility in these predictions?

What to Read Next

  • Science is becoming less human: “AI may be challenging the very nature of discovery,” I wrote in 2023.
  • “This is not how we do science, ever”: The new Trump administration is “willing not just to slash and burn research that challenges their political ideology but to replace it with shoddy studies designed to support their goals, under the guise of scientific legitimacy,” writes Katherine Wu.

P.S.

As winter grays transition to a spring palette, add another color to your wish list: olo, a new color designed by researchers at UC Berkeley. But you might have to wait, as you’d have to have your retina mapped first. Only five people in the world have seen olo so far, and while my colleague Ross Andersen has not, he did write a beautiful story about the team that created the color.

— Matteo

Sign up for Work in Progress, a newsletter in which Derek Thompson, Rogé Karma, Annie Lowrey, Jerusalem Demsas, and others explain today’s news and tomorrow’s trends in work, technology, and culture.

For full access to our journalism, subscribe to The Atlantic.