Episode 0040: Dr. James Hunt

Are Tech Teams Building Safe Systems?

(Or Just Believe They Are)

Real-Time Java, AI & Safety

Are modern software systems actually getting safer? Or are we just getting better at believing they are?

In this episode of DRIVEN, Paul sits down with James Hunt, CEO/CTO of AICAS, to explore one of the hardest problems in engineering: How do you build software that can’t fail?

James has spent decades working on real-time and safety-critical systems. From early computing at Data General to advanced embedded platforms and AI-driven systems today.

This conversation goes deep into the reality behind modern software:

- Why 70% of software errors come from lack of memory safety

- The critical difference between real-time and “fast”

- Why safety requires both determinism AND memory safety

- The challenge of combining AI (statistical systems) with deterministic control systems

- Continuous AI retraining loops inside embedded and automotive systems

- Whether we can ever truly trust AI in safety-critical environments

One of the most important insights:

- AI systems don’t “know” anything...they are statistical models.

- So how do you make something inherently uncertain… safe?

If you’re building robotics, autonomous systems, or large-scale software platforms, this episode will fundamentally change how you think about reliability, safety, and system design.

Don’t forget to like and subscribe for more conversations on the future of robotics, AI, and beyond!

From: Paul Perrone - DRIVEN Host (https://www.driven.show) & Perrone Robotics CEO/Founder (https://www.perronerobotics.com)