Microsoft Unveils 'Correction' to Tackle AI Hallucinations

The new feature aims to detect and correct inaccuracies in AI-generated content by aligning outputs with verified documents.

The 'Correction' feature is integrated into Microsoft's Azure AI Content Safety system.
It works by flagging and revising incorrect AI outputs based on grounding documents provided by users.
The feature is designed to improve the reliability of AI responses, particularly in critical fields like medicine.
Experts caution that while 'Correction' may reduce errors, it does not eliminate the root causes of AI hallucinations.
The feature is available in preview, with usage beyond 5,000 text records per month incurring additional costs.