OpenAI Recruits Global Physicians to Improve ChatGPT Health Responses

OpenAI has recruited hundreds of practicing physicians across the globe to evaluate and refine the health-related responses of ChatGPT. Led by Rebecca Soskin Hicks, this network of contractors spans 49 languages and includes medical professionals from countries such as Kenya, Nepal, and Brazil. The physicians review conversations to identify gaps where the AI might confuse users or create physical danger, having analyzed over 700,000 responses as of July.

This initiative follows increasing legal pressure, including a July lawsuit from a pastor who alleged that GPT-4o misdiagnosed lung symptoms. While the company maintains that its service terms explicitly state the chatbot is not for diagnosis or treatment, it continues to expand its medical capabilities. Recent developments include the integration of medical records and Apple Health, as well as the release of a mental health response benchmark.

Rebecca Soskin Hicks emphasizes that the process involves testing ambiguous and noisy situations to ensure safety in a high-stakes environment. She notes that there is no finish line for what is considered good enough and that the company will continue to seek improvements indefinitely.