Food and According to sources familiar with the meeting, the Drug Administration has been meeting with OpenAI to discuss the use of AI by the agency. These meetings appear to be part of a broader FDA effort to use the technology to speed up the drug approval process.
“Why does it take 10 years to go to market?” FDA Commissioner Marty Makary wrote on Wednesday. “Why don’t we modernize with AI and other things? We just finished our first AI-AI-Asiscrastion scientific review of our product, and that’s just the beginning.”
The rhetoric comes after the annual meeting of the American Hospital Association earlier this week, Makary talked about the potential of AI to help approve new therapies for diabetes and certain types of cancer.
Makary did not specify Openai as part of the program. But sources close to the project said a small team from Openai has been in contact with the so-called government efficiency ministry several times in recent weeks with two colleagues from the FDA and Elon Musk. The panel discussed a program called CDRGPT, which may represent the Center for Drug Assessment, which regulates over-the-counter and prescription drugs in the United States, as well as research GPT. Jeremy Walsh, who was recently appointed as the FDA’s first AI official, presided over the discussion. So far, no contract has been signed.
Openai declined to comment.
Walsh also met with Peter Bowman-Davis, a Yale University vacancies, an undergraduate student who is currently acting AI officer in the Department of Health and Human Services, discussing the FDA’s AI ambitions. Politico first reported on the appointment of Bowman-Davis, who is part of Andreessen Horowitz’s American Dynamics team.
When contacted by email on Wednesday, Robert Califf served as FDA specialist from 2016 to 2017 and again from 2022 to January, saying the agency’s review team had been using AI for years. “It will be interesting to hear what parts of the comments are details of ‘AI assist’,” he said. “There has been a search for shorter review time and broad consensus that AI can help.”
Before Kalif left the agency, he said the FDA was considering various ways in which AI could be used in internal operations. “The final comment approved is only part of a bigger opportunity,” he said.
To be clear, using AI to assist in the final drug review will be an opportunity to compress only a small part of the infamous long-term drug development timeline. The vast majority of drugs fail before FDA review.
Rafael Rosengengarten, CEO of Precision oncology company Genialis, also co-founder and board member of the Healthcare AI Alliance Alliance, said he favors automation and certain tasks in the drug evaluation process, but said there should be strategy guidelines around which data to train AI models and which model performance. “These machines are very good at learning information, but they have to be trained in some way so they can learn what we want them to learn,” he said.
