Get Full Government Meeting Transcripts, Videos, & Alerts Forever!
Get email alerts on the Ai Mitigation topic
No spam. Unsubscribe anytime.
Speakers urge audits, inclusive models and careful prompting to reduce AI bias
Summary
The webinar recommended mitigation practices—source auditing, inclusive models such as Latimer, local LLMs/NotebookLM, careful prompting and feedback signals—to reduce biased outputs and preserve privacy.
Gadeau presented a set of practical mitigation techniques for users and organizations. He recommended inclusive LLMs that incorporate culturally fluent sources (he cited Latimer), building or running models locally for privacy and domain control (NotebookLM was mentioned), and prompt engineering such as specifying diverse groups in image prompts. He emphasized feedback loops: using thumbs‑up or thumbs‑down signals and explaining why to help models learn.
He also presented four FLCC guidelines for everyday users: co‑create (treat AI as a partner), don't let AI be the sole arbiter of truth, avoid entering personal identifying information into AI prompts, and remain attentive to hidden biases. "Don't put personal information in an AI," he said, summarizing the privacy advice.
