OpenAI's Karan Singhal: 300M people weekly ask ChatGPT health questions, medical intelligence costs will plunge
a16z Podcast · rss · 2026-10-20
a16z podcast conversation with OpenAI's Karan Singhal on the state and long-term vision of AI in healthcare.
Key points:
- More than 300 million people use ChatGPT for health and wellness questions each week. Singhal sees that as just the beginning: the long-term goal is AI that helps people navigate health proactively while making trustworthy medical intelligence dramatically more accessible.
- OpenAI trains healthcare expertise directly into its general-purpose frontier models rather than shipping a separate medical model, working with hundreds of medical experts to evaluate safety and disagreement.
- AI is positioned as a complement to doctors, not a replacement; there are integrations with Epic and AI copilots for clinicians.
- Looking ahead, as personal agents start navigating appointments, health records and other fragmented parts of the system, the cost of medical intelligence could keep falling — enabling AI to act as a safety net, surface insights from years of health data, and accelerate discovery and validation of new treatments.
More from AGI Musings
- Security researcher: "Ban AI models, regulate GPUs" is the wrong answer to AI threats — basedjensen · 2026-10-20
- AI debate: is the tool neutral, or does blame lie with the person wielding it? — gerardsans · 2026-10-20
- Biohub's Alex Rives on virtual cells and the $1.8B push to model life — The Cognitive Revolution · 2026-10-20
- Biohub's Alex Rives on virtual cells and the $1.8B push toward a world model of life — The Cognitive Revolution · 2026-10-20
- Reddit debate: should an AI Manhattan Project replace the commercial race to ASI? — Few_days_b4_death · 2026-10-20
- BCG report: South Korea joins top-tier "AI Pioneer" group for the first time — JungWooHa2 · 2026-10-20