Meet the top OpenAI researcher trying to make ChatGPT into a Google-sized healthcare juggernaut

At OpenAI, the focus on healthcare has become paramount as the company seeks to enhance the capabilities of ChatGPT for medical inquiries. With over 230 million users turning to the platform each week for health and wellness advice, the demand for accurate and reliable medical information is soaring. Leading this initiative is Karan Singhal, a prominent researcher who has taken on the challenge of making ChatGPT a transformative tool in healthcare. Singhal aims to develop a version of ChatGPT that not only provides insightful health advice but also instills trust among users, acting as a reliable guide in their healthcare journeys. He emphasizes that the upcoming GPT-5 model family has been specifically designed to excel in health-related queries from its initial stages of development. Before his tenure at OpenAI, Singhal was instrumental at Google, where he contributed to the creation of the Med-PaLM AI models tailored for medical questions. However, as AI development trends shifted towards general-purpose models, Google reduced its investment in this area. In mid-2024, Singhal joined OpenAI, where he quickly recognized the need for improved health-related responses, prompting him to form a dedicated team of health experts and collaborate with over 200 physicians. One significant achievement under his leadership was the launch of HealthBench, a series of assessments designed to evaluate AI systems' capabilities in health scenarios. Singhal believes that establishing clear evaluation metrics simplifies the process of enhancing AI performance. Recently, OpenAI announced that its latest model, GPT-5.5 Instant, outperformed both physician-generated responses and its predecessor, GPT-4o, in health-related tests, showcasing a 71% reduction in flagged inaccuracies over two months. As healthcare professionals increasingly integrate ChatGPT into their practices, Singhal notes that the adoption of AI technology is progressing rapidly. He contrasts this with the historical reliance on Google Search for health information, which often directs users to static websites. Instead, he views the interactive nature of chatbots as a significant advancement, providing tailored advice through conversational exchanges. One of the main hurdles in offering personalized health information through AI is the lack of context about the patient. Unlike a physician who has access to medical records and a history of patient interaction, ChatGPT must find ways to simulate this experience. OpenAI's recent efforts include a health-oriented feature within ChatGPT that allows users to connect health applications and upload medical records, enabling personalized analysis. Singhal's team recognizes the importance of making AI accessible to all users, not just those who are tech-savvy. He stresses that the pace of adoption will depend on healthcare professionals' readiness to embrace these advancements. As the technology evolves, guiding users toward effective utilization remains a priority for Singhal and his team at OpenAI.