- ChatGPT has 230 million people using its tool for health and wellness advice each week, according to OpenAI.
- OpenAI aims to improve the quality of the models' health answers, as stated by Singhal, who felt a 'responsibility' to enhance accuracy when he joined.
- Singhal leads a high-stakes goal to ensure that the health tool changes people's lives for the better and avoids calamity.
- OpenAI's latest model, GPT-5.5 Instant, scored better than both physician-written answers and GPT-4o in tests, indicating improvements in health advice accuracy.
- There has been a 71% drop in responses flagged for inaccuracy over the last two months, showcasing progress in the tool's reliability.
OpenAI's health advice tool, which boasts over 230 million weekly users, is on a mission to enhance its accuracy and reliability. Singhal, who leads this initiative, expressed a strong sense of responsibility to improve the quality of health-related responses.
In a bid to measure and enhance AI health capabilities, Singhal helped launch HealthBench, a series of evaluations developed in collaboration with a physician group. This initiative aims to ensure that the AI provides health advice that can genuinely change lives and avoid potential calamities.
OpenAI's latest model, GPT-5.5 Instant, has reportedly outperformed both physician-written answers and its predecessor, GPT-4o, in recent tests. Furthermore, the company noted a significant 71% drop in flagged inaccuracies over the past two months, indicating a positive trend in the tool's reliability.45
Singhal also highlighted the rapid adoption of AI technologies in healthcare, stating, 'If you think about the adoption of technology in healthcare broadly, it's actually incredibly, insanely fast.' This reflects a growing acceptance of AI tools among healthcare professionals, suggesting a promising future for AI in health advice.
“OpenAI's ChatGPT tool for health advice is utilized by over 230 million people each week. The company is focused on enhancing the accuracy and quality of its health-related responses.”
