health
There are more AI health tools than ever—but how well do they work?
Specialized chatbots might make a difference for people with limited health-care access. Without more testing, we don't know if they’ll help or harm.
TL;DR
- Microsoft, Amazon, and OpenAI have launched AI health tools that allow users to connect medical records and ask health questions.
- The demand for these tools is high due to difficulties in accessing traditional healthcare systems.
- Researchers and experts express concerns about the lack of independent and rigorous testing for these AI health tools before public release.
- Potential risks include misdiagnosis, recommending too much or too little care, and users misinterpreting AI advice.
- Companies are conducting internal evaluations like OpenAI's HealthBench, but these have limitations and lack the impartiality of third-party reviews.
- Independent studies, like one on Google's AMIE chatbot, show promise but Google is not releasing it publicly due to safety and equity concerns.
- While some AI applications like suggesting exercise plans are low-risk, others like triage, diagnosis, and treatment plans carry significant risks.
- Despite disclaimers, users are likely to use these tools for diagnosis and management, highlighting the need for robust safety measures.
- There is a consensus among experts that while AI health chatbots could be beneficial, especially for those with limited healthcare access, their current performance and safety are not fully established.