Dr. Minhee Jun, Assistant Professor in the Department of Computer and Electrical Engineering, recently gave a selected oral presentation at the NIST-University Research Summit, held at the National Institute of Standards and Technology (NIST) in Gaithersburg, Maryland.
Her presentation, “Toward Trustworthy Physical AI: Human Demonstrations, Task-Centric Digital Twins, and TEVV for Robot Learning,” detailed a vision for overcoming major bottlenecks in AI training and deployment.
Key highlights from her presentation include:
- Scalable Physical AI Training: Highlighting the shift from reinforcement learning to imitation learning and proposing human task demonstrations as a scalable foundation for training next-generation Physical AI systems.
- Privacy-Aware Data Collection: Leveraging smart-glasses recordings and video repositories to capture human task demonstrations while using advanced masking techniques to protect sensitive information.
- Task-Centric Digital Twins: Transforming these human demonstrations into simulation-ready virtual environments, reducing the need for costly and risky real-world experimentation.
- Trustworthy AI Evaluation: Supporting Testing, Evaluation, Validation, and Verification (TEVV) through reproducible datasets and benchmark generation to ensure AI reliability.
- National Collaboration: Exploring future partnerships with NIST in digital twin technologies, evaluation science, and standards development for trustworthy physical AI.
Dr. Jun's presentation highlights The Catholic University of America's ongoing contributions to the national conversation on trustworthy artificial intelligence and robotics. It also showcases the research being conducted at the Center for Artificial Intelligence Research (CAIR) within the College of Engineering, Physics, and Computing.
Learn more about Dr. Minhee Jun and her research: Faculty Profile + CAIR Website