Invited Talk 2 by Lama Ahmad (Technical Program Manager, Trustworthy AI at OpenAI): Human and AI Evaluations for Safety and Robustness Testing
Lama Ahmad
2024 Invited talk
in
Affinity Event: Muslims in ML
in
Affinity Event: Muslims in ML
Abstract
Evaluating advanced AI systems for safety and adversarial robustness is a critical step in ensuring their responsible deployment. This talk explores the intersection of human and AI-driven evaluations in the context of safety and security testing. We will examine current practices, highlighting how human judgment and AI-assisted tools complement each other in identifying vulnerabilities, unintended behaviors, and emergent risks.
Video
Chat is not available.
Successful Page Load