Can You Trust AI? Workshop

About this event

Virginia Bennett Room. Millions of people use AI tools like ChatGPT every day for school, work, research, and everyday questions. But there's always one important question in the back of our minds: Can I actually trust this answer? This workshop explores how modern AI systems work, why they sometimes produce misleading or incorrect information, how often they do so, and what current AI safety research reveals about making these systems more trustworthy. Through live demonstrations and real-world examples, you'll learn practical strategies for evaluating AI-generated information, simple techniques for verifying whether an AI response is factual, and how to recognize when AI is giving answers that sound convincing or simply tell you what you want to hear instead of what is supported by evidence. The goal is not to make you trust AI more, but to help you understand when you should trust it, when you shouldn't, and how to confidently verify its answers so you can use AI as a powerful tool without being misled. No programming or technical background is required. Whether you're a student, educator, professional, parent, or simply curious about artificial intelligence, this session will help you better understand how AI works, when it can be misleading, and how to use it responsibly. About the instructors: Mrinal A. is an AI Safety & Interpretability researcher whose work on AI deception and reasoning was spotlighted at NeurIPS and AACL-IJCNLP, some of the world's leading artificial intelligence conferences. His research has been presented to more than 400+ AI researchers, educators, and industry professionals receiving over 19 citations from UPenn, CMU, Microsoft, and MBZUAI and he has over four years of experience teaching students ages 12–18. His work focuses on AI safety, deception detection, multi-agent reasoning, interpretability, and building more trustworthy AI systems. Abbas Raza is the CTO of RefocusAI and brings an industry perspective on how AI is being…

Description from aclibrary.bibliocommons.com