OpenAI's Whisper AI Faces Criticism for Inaccurate Medical Transcriptions

The widespread use of Whisper in healthcare settings raises concerns over fabricated and potentially harmful transcription errors.

Whisper, an AI transcription tool by OpenAI, is criticized for generating false information, known as 'hallucinations,' in medical settings.
Researchers have found that Whisper often adds non-existent content, including violent and racial remarks, to transcriptions.
Despite OpenAI's warnings against using Whisper in high-risk domains, the tool is employed by over 30,000 clinicians and 40 health systems.
Nabla, a company using Whisper, deletes original audio recordings, preventing verification of AI-generated transcripts.
Experts emphasize the need for caution and higher standards in using AI for critical tasks like medical transcriptions due to potential misdiagnoses.