America’s Autism Crisis and How AI Can Fix Science with NIH Director Jay Bhattacharya

Imagine a system where scientists who fail are celebrated for their productive failures, much like Silicon Valley’s venture capital model that backs bold ideas knowing many will not succeed. This is the transformative spirit Dr. Jay Bhattacharya, director of the NIH, wants to bring to American biomedical science. In a wide-ranging conversation, he details a new $50 million Autism Data Science Initiative and announces two significant public health updates: emerging evidence cautioning against routine use of Tylenol during pregnancy due to a potential link to autism, and policy changes to improve access to leucovorin, an old drug showing remarkable promise in treating a specific folate-deficiency subtype of autism. Underlying these announcements is a broader mission to reset the NIH’s approach, tackling a deep-seated replication crisis and rebuilding public trust through transparency and humility.

A central theme is the urgent need to reform scientific culture, which has become overly conservative and incremental. Bhattacharya argues that the median age for a scientist’s first major NIH grant has shifted from the mid-30s to the mid-40s, stifling the novel ideas often born from younger investigators. He draws a direct analogy to venture capital, advocating for a portfolio approach that embraces high-risk, high-reward projects and stops punishing “productive failure.” To execute this, he is decentralizing some grant-making authority to institute directors, incentivizing them to fund a diversified portfolio aligned with strategic visions rather than just safe, consensus-driven projects.

Rebuilding public trust, severely damaged during the pandemic, is another pillar of his agenda. Bhattacharya emphasizes that trust is earned not by claiming infallibility but by openly communicating uncertainty and relying on rigorous, replicated science. He criticizes past public health mandates that lacked evidence and underscores a new commitment to “gold standard science,” where replication, not just publication in a prestigious journal, is the benchmark for truth. This ethos extends to practical reforms like removing internal permission barriers for NIH scientists to publish and challenging the for-profit journal oligopoly that puts publicly funded research behind paywalls.

The conversation also explores the role of artificial intelligence as a force multiplier in biomedicine, from accelerating drug discovery through protein-folding models to streamlining clinical documentation so doctors can focus on patients. However, Bhattacharya cautions that AI is a tool for augmenting human scientists, not replacing the creative, paradigm-challenging work that drives true breakthroughs. He remains optimistic that by fostering a culture of bold inquiry, rigorous replication, and humble partnership with the public, the NIH can begin to reverse troubling trends like stagnant life expectancy and address the chronic disease burdens affecting millions of Americans.

Surprising Insights

• Celebrating Productive Failure: A core proposal is to actively stop punishing scientists whose experiments fail, instead encouraging them to publish “failure reports” to share lessons learned, mirroring Silicon Valley’s tolerance for failure within a high-risk portfolio.
• The Aging of Scientific Ideas: Data shows the “age” of the ideas in a scientist’s published work increases by about one year for every calendar year they age, meaning the newest ideas disproportionately come from younger investigators—a pipeline currently stifled by funding mechanisms.
• Public Trust and Scientific Certainty: Bhattacharya posits that the public is very capable of understanding scientific uncertainty; the loss of trust stems not from public confusion but from scientists and officials conveying unfounded certainty during crises like the pandemic.
• Old Drugs, New Pathways: The discussion highlighted leucovorin, a cheap, old drug used for decades for other purposes, as having dramatic effects (restoring speech in 20% of cases) for a specific subset of autistic children with cerebral folate deficiency, showcasing the potential of drug repurposing.
• Portfolio Management for Science: The NIH director explicitly applies a venture capital framework to NIH funding, arguing institute directors should be judged on their overall portfolio’s impact and strategic alignment, not on the success of every individual grant.

Practical Takeaways

• Advocate for Replication: When evaluating scientific claims, prioritize findings that have been independently replicated over single, high-profile studies, even those published in prestigious journals.
• Support Early-Career Scientists: Champion funding programs and institutional policies designed to get resources and independence to younger researchers, as they are the most likely source of disruptive ideas.
• Embrace Intellectual Humility: In both consuming and communicating science, lead with the available evidence and openly acknowledge the limits of current knowledge, especially in areas of genuine uncertainty.
• Scrutinize Routine Medical Practices: The Tylenol discussion underscores that even long-standing, common practices during vulnerable periods like pregnancy should be periodically re-evaluated as new evidence emerges.
• Push for Open Science: Support policies and platforms that make taxpayer-funded research freely available to the public, breaking down journal paywalls and accelerating the translation of science into public benefit.