The Truth About AI Integrity

About this episode

This episode of My Interface Podcast, features an insightful conversation with Xi Peng, an Associate Professor of Computer Science at the University of Virginia and head of the DeepReal Lab. The discussion tackles the urgent need for "Explainable AI" and how the current education system must adapt to the new AI World.

Just as you would question a calculator that suddenly says 12 times 10 is 121, Xi Peng explains why it is critical to understand how AI models arrive at their conclusions, especially in high-stakes fields like healthcare and autonomous driving. The episode also explores the limits of AI in the physical world and why human connection, emotion, and physical sensation remain irreplaceable.

What You Will Learn

• Explainable AI (XAI): Why treating AI as a "black box" is dangerous for high-stakes industries like radiology, and how researchers are looking inside neural networks to build human trust.

• The 100-Year-Old Education System: Why our current educational models are obsolete in the age of AI, and how schools need to shift from testing answers to teaching students how to ask the right questions.

• Physical AI and Its Limits: Why an AI might know everything in a textbook but can never be your ski coach, highlighting the gap between text-based knowledge and physical-world interaction.

• Prompting as Problem Solving: The essential future skill of breaking down massive, complex problems into sub-tasks that an AI can manage, leaving the high-level design to humans.

Chapters

• 00:07 Introduction to the Interface Podcast and its mission.

• 01:21 Guest Introduction: Professor Xi Peng of the University of Virginia's DeepReal Lab.

• 01:56 The Classroom Dilemma: Students using AI to write bug-free code.

• 03:22 Explainable AI: The "broken calculator" analogy and opening the black box.

• 04:08 Reinventing Education: Adapting a century-old learning system for an AI-native generation.

• 08:04 Physical AI: Why AI struggles to interact with the real, tangible world.

• 11:37 The Ultimate Skill: Breaking down complex problems and asking the right questions.

• 12:15 Advice for Students: Why the internet might be better than textbooks for learning rapidly evolving AI.

• 13:03 Turning the Tables: Xi Peng asks the hosts about their own AI usage and school policies.

• 14:41 Host Reflections: Yvonne and Miya discuss AI detection, critical thinking, and the necessity of human connection.

Quotes from the Episode

"Today's AI is more like a black box... My research interest is to open this black box to figure out how the AI thinks to solve the problem internally."

"We still need to educate the kids... We can't forbid people to use AI, but the question is how to use it while the student can still learn."

"One job that AI probably never can replace human is being a ski coach... because this needs a sensation. You need an interaction with real world."

"Learn from internet, rather than learn from textbooks, if you want to learn AI."

About the Hosts

Yvonne and Miya are high school sophomores on a mission to educate students and professionals on the impact of AI across multiple subject matters by providing insights from top experts.

Where to find Professor Xi Peng and the DeepReal Lab:

• Xi Peng: https://www.linkedin.com/in/xi-peng-74b540b6/

• DeepReal Lab: https://deep-real.github.io/

  
  

New episodes drop every Friday at 3:00 PM.