The answer lives in this podcast
Beyond saving time, where is the biggest opportunity for AI in healthcare and scientific discovery?
The biggest opportunity isn't saving hours — it's solving cases that humans couldn't solve at all. OpenAI partnered with a children's hospital to run frontier reasoning models against pediatric cases that had stumped the medical system for years, with no diagnosis. Those models suggested new pathways, new tests, and entirely new diagnoses, giving families answers they had never received.
Unsolvable for years — diagnosed in a single run
Katrina Mulligan frames this not as an incremental improvement but as a category shift. These were patients who had been inside the healthcare system for years, cycling through specialists and tests, with nothing conclusive. The models weren't replacing doctors — they were doing something doctors hadn't been able to do: seeing patterns across the totality of the case and proposing paths that hadn't been tried.
The result was entirely new diagnoses for families who had given up expecting answers. As Mulligan explains in 3 Takeaways™, this is what frontier reasoning models make possible when organizations are willing to use them creatively rather than defensively.
A "defensive crouch" is the real obstacle — not the technology
Mulligan's sharpest point isn't about the model's capabilities — it's about organizational posture. She explicitly contrasts organizations that "creatively explore" with those that remain in what she calls a "defensive crouch." The technology already exists. The constraint is the willingness to apply it to hard problems.
This is a direct challenge to institutions — including government bodies like the Defense Health Agency — that are cautious about deploying AI in high-stakes settings. Mulligan's argument, detailed further in this episode, is that caution has a real cost: every year without a diagnosis is a year a family lived without answers that were, in principle, reachable.
The children's hospital partnership is presented as a model — a proof of concept for what becomes possible when a frontier organization and a high-stakes institution agree to move at the same speed. It's a template Mulligan believes should be replicated, not treated as an exception.
"I honestly think that I am at least 30% more effective, maybe more than I was a year ago, because of how I've matured my use of these tools. I'm a better leader. I'm a better manager."
Katrina Mulligan — Head of National Security Partnerships, OpenAI.
Mulligan previously held senior roles at the Department of Defense, the National Security Council, and the Department of Justice — including serving as number two at the Pentagon overseeing special operations. She describes her move to OpenAI as the hardest professional transition of her career: leaving the most hierarchical organization on the planet for its exact opposite. That dual vantage point — inside both Washington's national security infrastructure and frontier AI development — makes her one of the few people positioned to speak to what AI actually does at scale, in real institutions, under real pressure. Her perspective on 3 Takeaways™ is grounded in lived operational experience, not theory.
The 30% personal effectiveness figure Mulligan cites is not a benchmark from a study — it's her own direct experience using AI tools daily at OpenAI. It matters here because it grounds the healthcare argument: if a senior executive with decades of institutional experience sees that magnitude of improvement in her own work, the potential in domains like medicine — where the raw complexity is far greater — is a logical extension, not a leap. The full context of that argument is laid out in this episode of 3 Takeaways™.
See also
What was the state of ChatGPT when it first launched and what changed to make modern AI so much more capable?
Mulligan notes that when ChatGPT first launched, there were only 200 people at OpenAI. The foundational shift came from scaling laws: multiply compute by 10x and data by 10x, and you get predictably better model performance — the discovery that has driven every capability leap since.
What is recursive self-improvement in AI and how close is it to becoming reality?
Mulligan explains that recursive self-improvement is the idea that AI models would eventually become good enough to conduct the research that improves the next generation of AI — creating a self-reinforcing cycle of capability growth that would fundamentally change the pace of development.
What makes the AI revolution fundamentally different from every previous technology revolution in American history?
According to Mulligan, this is the first time in American history that a technology of this consequence is being developed exclusively by the private sector — without the government-led coordination that defined transformations like GPS or the Human Genome Project.
Key takeaways
- OpenAI's partnership with a children's hospital produced entirely new diagnoses for pediatric patients who had been undiagnosed for years — the models suggested pathways and tests no clinician had tried.
- The obstacle isn't the technology's capability — it's organizational posture. Institutions in a "defensive crouch" leave solvable problems unsolved.
- Mulligan holds up this hospital collaboration as an explicit model for what becomes possible when an institution agrees to explore rather than wait.
- The opportunity in healthcare and scientific discovery is not efficiency — it's accessing answers that were previously out of reach entirely.
This answer comes from episode #315 of 3 Takeaways™, hosted by Lynn Thoman with Katrina Mulligan.
Listen to the episode on Listenly