Anthropic is making biotech boasts, but scientists are wary of hype around ‘curing cancer’

Anthropic has been pushing is biotechnology projects in a bid to drive up its stock price, but scientists warn people need to be realistic.

The AI giant has opened a biology lab in the Bay Area, bought two biotech startups — one for about $400 million in stock — and partnered with drugmakers including Bristol Myers, Squibb and Novo.

CEO Dario Amodei predicted powerful AI could cram 50 to 100 years of biological progress into the next five to 10. This summer, he wrote promising AI will cure cancer had become a “cliché” and that “the thing that will work is actually curing cancer.”

While AI from Anthropic and other companies has greatly benefitted the medical industry, people working in the medical field all their careers caution against hype.

“A good chunk of the timeframe has passed since [Amodei’s prediction], but we are far away from that goal,” Andreas Bender, a professor of machine learning in medicine at Khalifa University in Abu Dhabi, told the Post.

“The human proof point is what counts, such as a clinical trial,” added Bender, who has co-founded several AI drug-discovery startups and has no ties to Anthropic.

No drug developed by AI has yet won full approval from the Food and Drug Administration and the leap from the lab to the human body is where most medicines die.

“You can do an experiment, things work very well, you have all kinds of supporting data from cells or mice or from biopsies, and then it gets to the clinic and fails completely,” a Silicon Valley cancer scientist with over 10 years of experience in the field, told the Post.

“You can pour millions of dollars and thousands of man hours on some method that looks wonderful in the lab, but when it gets to clinical trials, nothing happens. And it’s not because someone was stupid,” the scientist, who asked not to be named, said.

“Every human is drastically different from every other in multiple hundreds, thousands of parameters,” he said. “That’s why so many drugs might work in a very neat, organized system, but once you get to the real world, they won’t.”

Personalized medicines are one of the ways AI — from Anthropic, OpenAI, Google and other labs — is helping to improve the medical field and treatments for people who are ill, particularly those with cancer.

But still only about one in 10 drugs which enter human trials ever reach the market — and for cancer drugs, it’s closer to 1 in 20, industry data shows.

The drug industry’s bosses are blunt in their assessment of cancer being cured anytime soon.

“That’s complete BS,” Jeff Leiden, executive chairman of Vertex Pharmaceuticals, told the Wall Street Journal this week, recalling a tech leader who claimed AI would cure 90% of human disease in seven years. Vertex has run its 10,000-compound library past roughly 50 AI companies, and none produced a working drug, he said.

However, not everyone is as pessimistic. The Silicon Valley cancer scientist doesn’t think Amodei’s timeline is too ambitious.

“It is like criticizing the first planes,” he said. “Of course they are not going to be perfect, but these are the first steps.”

But even he says people, not chatbots, will be making the big calls for many years to come.

“Are the lab results good enough to justify pouring $25 million into clinical trials that could have gone into something else? AI won’t decide that in the next 10 years,” he said.

“So much is at stake that it has to be a human that decides those things, a committee of humans.”

Bender’s message to Amodei: biology is not physics.

Anthropic did not respond to The Post’s requests for comment. In its own announcement, the company acknowledged work is “early” and “mostly preclinical,” meaning the drugs have not yet been tested in people.

Meanwhile, the scientist laid out exactly what would make him a believer.

“Show me the AI that really develops a revolutionary new drug for some unmet medical need,” he said. “Show me something that scientists have been struggling with for 10, 15 years, and all of a sudden the AI can really reliably find a solution for it that no one thought about.”