Doctors horrified after Google's healthcare AI makes up body part
15 days ago
- #Google Med-Gemini
- #Medical Errors
- #AI in Healthcare
- Health practitioners are concerned about the widespread use of error-prone generative AI tools in medicine.
- AI 'hallucinations'—made-up facts and lies—are a significant issue, with one error taking over a year to be caught.
- Google's Med-Gemini AI incorrectly identified a non-existent brain part, 'basilar ganglia,' in a research paper.
- The error was flagged by a neurologist, but Google only fixed its blog post, not the research paper.
- AI's falsehoods could have devastating consequences in healthcare settings, despite Google's claims of 'substantial potential in medicine.'
- Experts warn that AI's tendency to make up information without admitting uncertainty is dangerous in high-stakes domains like medicine.
- Google continues to push AI in healthcare, including error-prone features like AI Overviews for health advice.
- Human oversight is crucial to monitor AI outputs, but inefficiencies and risks remain if outputs go unverified.
- Some experts argue AI should have a much higher accuracy bar than humans in healthcare.