When AI Can’t Afford to Be Wrong: Inside Bot Auto’s Autonomous Trucking Mission

Everyone’s talking about AI that writes code and answers questions. But what happens when AI has to operate in the physical world, where mistakes don’t just generate weird sentences, they have real consequences?

On this episode of Innovation Heroes, host Ed McNamara sits down with Tete Xiao, VP of Engineering and AI at Bot Auto, to explore what it actually takes to deploy AI that touches the real world. Tete helped create Meta’s Segment Anything model—a breakthrough that changed how machines understand images—and now he’s applying a decade of cutting-edge research to one of AI’s hardest challenges: making trucks drive themselves safely on U.S. highways.

Key takeaways:

• Why physical AI is “broadly underhyped” despite billion-dollar robotics headlines
• How simulation lets autonomous systems explore scenarios too dangerous to test in real life
• What “truly driver-out” operations mean—and why it’s so much harder than having someone in the backseat
• Lessons IT leaders can apply when deploying any AI system with real-world consequences

For anyone thinking about AI in manufacturing, healthcare, logistics, or security, this conversation is essential listening.