OpenEvidence, the $15 billion start-up whose AI chatbot answers doctors’ clinical questions, plans to use its technology to find new drugs, its chief executive, Daniel Nadler, said, as Rafe Rosner-Uddin reported in the Financial Times on Wednesday.
The tool is free to verified doctors, and the company makes its money from advertising. Mr. Nadler said that model wasn’t the long-term prize.
“The end game is literature-based drug discovery,” he said, referring to the practice of mining published scientific research for new findings. “The economic value of that as a company is going to dwarf by orders of magnitude whatever we could have charged hospitals or advertisers.”
Investors have been paying up for that pitch. OpenEvidence raised $250 million at a $15 billion valuation in a round led by Andreessen Horowitz and Byers Capital, which Mr. Nadler confirmed in a Sept. 28 email to Becker’s Hospital Review. Forbes reported on Sept. 30 that the company had confirmed the raise. In January, a $250 million round led by Thrive Capital and DST valued it at $12 billion, and Mr. Nadler told CNBC at the time that 40 percent of U.S. physicians used the tool and that it had passed $100 million in annual revenue the year before.
The company has also hired Alexander Tarashansky, who created Claude Science, an Anthropic platform for research scientists. He will work on OpenEvidence’s clinical side.
The bigger AI labs want the same doctors. OpenAI has moved into the field with ChatGPT for healthcare. Anthropic has released its own health care tools, but it recently teamed up with OpenEvidence to bring them to 100 low- and middle-income countries.
Nworah Ayogu, a partner at Thrive Capital, said doctors’ choice to use the product was the company’s edge. “The fact that doctors are adopting and using a piece of software of their own volition means there is a degree of differentiation that OpenEvidence can really lean on,” he said.
The research on how well it works is mixed. A study by the Mayo Clinic, which is itself an investor in OpenEvidence, found that the tool “provided accurate, evidence-based recommendations” in all five patient case studies it sampled. Its effect on treatment decisions was more limited, the study found.
A paper published in Nature in June found that OpenEvidence did worse on clinical benchmarks than general-purpose models such as Google’s Gemini and OpenAI’s GPT.
Dr. Eric Oermann, a neurosurgeon at NYU Langone, said little research had been done on how individual AI tools perform, and that hospitals had scant guidance on how to put them to use in clinical settings.
There is also a worry, which the Financial Times described as widely held, that junior doctors could lean on OpenEvidence and similar systems before they’ve learned the basics of particular conditions and how they’re treated.
Mr. Nadler waved off the benchmark results. “It’s a fun, friendly competitive pressure to play with benchmarks... [But] there’s no question of what AI tool physicians trust and use,” he said.