Robots in the skies (and they use Transformer models)

In this episode, Ryan Donovan introduces the topic of artificial intelligence (AI) in defense technologies, welcoming Nate Michael, CTO of Shield AI. The discussion sets the stage for exploring the technical aspects of AI-enabled autonomous systems, focusing on how such technologies are being employed in sensitive military operations.

Nate shares his background, detailing his passion for robotics sparked by childhood shows like Transformers. With a PhD from the University of Pennsylvania and experience as a faculty member at Carnegie Mellon, he explains his focus on resilient intelligence and developing systems that can operate independently in uncharted environments.

Nate explains Shield AI's Hivemind technology, which enables drones and other autonomous vehicles to think and act independently in complex environments. He describes it as an 'autonomy factory' that supports the development of AI capabilities and emphasizes the importance of edge-level intelligence.

The conversation shifts to the limitations of on-device AI, particularly regarding constrained compute power, power, and space. Nate discusses how Shield AI designs modular frameworks to meet various operational requirements while balancing the need for intelligent decision-making under adverse conditions.

Discussion on the human role in the operation of autonomous vehicles reveals how operators interact with drones from a safe distance, ensuring that critical decisions remain in human hands. Nate emphasizes the importance of keeping humans in the loop, particularly during high-risk missions.

Nate addresses the probabilistic nature of AI models and the necessity for rigorous quality assurance measures. He explains the techniques employed to ensure deterministic behavior in decision-making processes and the need for extensive simulations before real-world deployment.

Nate assures that Shield AI does not create weaponized systems. He highlights the importance of ethical guidelines that govern the use of autonomy in military contexts, focusing on the need for responsible engineering standards and alignment with U.S. governmental policies.

Concluding the episode, Nate discusses the versatility of Shield AI’s technology, indicating its potential applications beyond defense, including commercial uses. He stresses the commitment to developing high-performance autonomous systems while expanding their applicability to various industries.