Physical and Spatial AI

Through four focused sessions, participants will explore both the technical underpinnings and the strategic implications of these developments. The program provides senior leaders with a clear, cross-domain understanding of how generative AI, probabilistic modeling, and advanced system design are translating into operational advantage, while highlighting pathways to resilient, secure, and scalable AI-enabled systems. 

California DREAMS: Microelectronics for Defense, Space, and AI  

California DREAMS (CA DREAMS) is a USC-led microelectronics hub advancing domestic prototyping and fabrication capabilities for analog and digital microelectronics. The discussion highlights the hub’s technical capabilities, collaborative operating model, and strategic relevance to defense, space systems, and AI-driven technologies. 

Spatial AI for Mobility: From Next-Visit Prediction to Foundation Models 

Spatial AI is increasingly challenged to move beyond task-specific mobility analytics toward general-purpose foundation models. Using TrajGPT, a transformer-based model trained via self-supervised learning on large-scale trajectory data, the discussion illustrates how generative models can support multiple downstream mobility intelligence tasks and concludes by examining geospatial objects as a pathway to richer, transferable representations for next-generation Spatial AI systems. 

AI-Enabled Inference of Material Interfaces in Hypersonic Environments 

A multi-university research effort is advancing generative, probabilistic AI methods to infer material interface behavior in hypersonic environments. As new data from flight tests, laboratory experiments, and high-fidelity material simulations become available, the project addresses the challenge of synthesizing these heterogeneous and uncertain data sources within a unified inference framework, while assessing the relative contribution of each source. It establishes an AI-enabled inference approach for hypersonics that can be extended to other complex engineering systems. 

Safe and Effective Autonomy in the Age of AI 

Embodied AI–enabled systems, including autonomous vehicles and robotic platforms, are reshaping defense operations across air, ground, and maritime domains. With reduced human oversight, these systems must make reliable decisions under uncertainty, coordinate across distributed teams, and remain resilient in adversarial environments. 

The discussion presents a neuro-symbolic approach integrating logic-based reasoning with modern AI to enable trustworthy autonomy. Formal specifications make safety and mission objectives explicit and verifiable, while scalable verification and monitoring provide strong assurances and expose rare failure modes, advancing systems that are safe, reliable, and resilient by design.