How Microsoft Copilot's 'Critique' feature reduces AI hallucinations
What's the story
Microsoft has introduced new features in its Copilot research assistant, allowing users to leverage multiple AI models within the same workflow. The latest update is part of the tech giant's efforts to improve its AI offerings and drive adoption. The new feature, dubbed "Critique," will enable Copilot's Researcher agent to pull outputs from both OpenAI's GPT and Anthropic's Claude models for every response.
Feature functionality
'Critique' feature aims to reduce AI hallucinations
The "Critique" feature works by having GPT generate a response, which is then reviewed for accuracy and quality by Claude before being presented to the user. This bi-directional workflow could be implemented in the future, allowing GPT to review drafts created by Claude as well. Nicole Herskowitz, Corporate Vice President of Microsoft 365 and Copilot, said this multi-model approach will speed up user workflows while reducing AI hallucinations (where systems produce false information).
Feature introduction
New 'Council' capability for side-by-side model response comparison
Along with the "Critique" feature, Microsoft is also launching a new capability called "Council." This will let users compare responses from different AI models side-by-side. The updates come as Microsoft expands access to its new Copilot Cowork agentic AI tool for members of its 'Frontier' program, which gives customers early access to some of its latest AI features.
Tool details
Microsoft recently unveiled Copilot Cowork based on Claude Cowork
Microsoft had recently unveiled Copilot Cowork, a tool based on Anthropic's popular Claude Cowork product. The move comes as part of the company's strategy to capitalize on the growing demand for autonomous AI agents. Despite strong competition from rivals like Google's Gemini and autonomous agents such as Claude Cowork, Microsoft is racing to improve its Copilot assistant for better adoption.