OpenAI’s Whisper transcription tool has been criticized for producing “hallucinations,” or fabricated text, which can include harmful content and inaccuracies, especially concerning medical transcriptions and captions for the Deaf and hard of hearing. Despite warnings from OpenAI against using Whisper in high-risk areas like healthcare, many medical centers have adopted it to transcribe patient consultations, raising concerns about patient safety and accuracy. Experts are calling for regulatory measures to address these issues, emphasizing the need for improvements in Whisper’s accuracy and the potential dangers of reliance on AI in critical decision-making contexts.
Fabricated AI-Text
By Ian Vielma
Image from OpenAI
Ian Vielma
Current Computer Science Major @ Fordham University. Working in IT Security and Assurance as a Risk Assistant.