OpenAI has promoted its transcription tool, Whisper, as possessing near “human level robustness and accuracy.” However, Whisper has shown a significant drawback: it frequently generates fabricated text—referred to as hallucinations—which can include inappropriate or harmful commentary on race, violence, and even invented medical treatments. These inaccuracies raise serious concerns among experts, as Whisper is increasingly...
Premium Membership Required
You must be a Premium member to access this content.
Already a member? Log in here
