Historically, medical research has underrepresented women and people of color, impacting health outcomes significantly. A recent Financial Times report highlighted that AI tools used in healthcare, like OpenAI’s GPT-4 and Meta’s Llama 3, often exacerbate these disparities. Research from MIT revealed that women were more frequently advised to “self-manage at home,” leading to reduced clinical care. Similarly, the healthcare-centric model Palmyra-Med exhibited these biases, downplaying women’s health needs. Google’s LLM, Gemma, showed comparable results, while a previous study indicated that AI offered less compassionate care to people of color. Furthermore, a Lancet paper found AI models, including GPT-4, frequently relied on demographic stereotypes rather than symptoms. This misalignment poses risks as major companies like Google, Meta, and OpenAI integrate AI in healthcare, potentially perpetuating longstanding biases and misinformation. Ensuring equitable AI in medical settings is crucial to improve health outcomes for all demographics.
Source link
