In this talk, Stephen Miller explores the evolving role of artificial intelligence in modern vehicles and the critical question of how much AI is truly needed to deliver scalable, and cost-effective driver assistance systems. Drawing on his experience in ADAS product management and high-performance embedded computing, he examines the balance between increasing model complexity and real-world deployment constraints such as compute efficiency, latency, and system reliability. The session highlights how AI must be purpose-driven rather than over-engineered, with a focus on selecting the right architectures and sensor processing approaches to meet automotive safety and performance requirements.
