Stanford Professor aims to bring aviation-level safety to AI systems

Stanford aeronautics professor Mykel Kochenderfer is pioneering AI safety research for high-stakes autonomous systems, drawing parallels between aviation’s remarkable safety evolution and today’s AI challenges. As director of Stanford’s Intelligent Systems Laboratory and a senior fellow at the Institute for Human-Centered AI, Kochenderfer develops advanced algorithms and validation methods for autonomous vehicles, drones, and air traffic systems—work that has become increasingly urgent as AI rapidly integrates into critical infrastructure and decision-making processes.

The big picture: AI safety requirements vary dramatically across applications, from preventing physical collisions in autonomous vehicles to ensuring language models don’t produce harmful outputs.

• Kochenderfer illustrates this evolution by comparing early aviation technology to modern commercial flight, noting that “right now one of the safest places to be is up around 30,000 feet in a metal tube.”
• This transformation required decades of incremental improvements, risk-taking, and iterative testing—a pattern now repeating with AI systems.

Key research priorities: Kochenderfer’s team focuses on developing quantitative validation tools that can identify potential failures before deployment in real-world environments.

• Their work includes creating efficient simulation methods and high-fidelity models that accurately represent complex scenarios autonomous systems might encounter.
• This validation approach is critical for applications where safety and efficiency are non-negotiable, such as air traffic control and unmanned aircraft.

Major concerns: Premature dependence on AI systems before they’re properly understood presents significant risks to society.

• Kochenderfer emphasizes the importance of comprehensive testing and validation before deploying AI in critical applications.
• His work addresses the challenge of ensuring AI systems don’t hallucinate or provide harmful information—issues that become more concerning as these technologies gain wider adoption.

Current projects: The Stanford team is actively working on validation methodologies for large language models and other AI systems to prevent harmful outputs.

• This includes developing educational resources for students, industry professionals, and policymakers about AI safety principles.
• Their work builds on previous successes like the Airborne Collision Avoidance System X (ACAS X), demonstrating how rigorous safety engineering can be applied to AI.

Why this matters: As AI increasingly enters high-stakes environments, from healthcare to transportation, developing robust safety validation frameworks becomes essential for preventing potentially catastrophic failures while allowing beneficial innovation to continue.