Insights, breakthroughs, and updates from TerraNova MU's 8 specialized research labs
Exploring the escalating competition between adversarial attack development and defense mechanisms. Our Adversarial AI Lab analyzes recent breakthroughs in jailbreak attempts against GPT-4 variants and discusses countermeasures now being deployed.
How distributed consensus protocols are being adapted for collaborative AI training. The Byzantine Fault Tolerance Lab reveals new approaches to achieving agreement in decentralized LLM training across multiple nodes with Byzantine failures.
Establishing ethical frameworks for autonomous weapons systems. The Autonomous Systems Ethics Lab addresses the critical question of meaningful human control and proposes governance structures that balance operational effectiveness with ethical oversight.
Assessing the impact of quantum computing on military AI systems. Our Quantum-AI Security Lab details timeline projections for quantum computer deployment and the cryptographic vulnerabilities affecting defense infrastructure.
Advanced techniques for ensuring language model safety beyond standard RLHF approaches. The Natural Language Governance Lab explores constitutional AI, mechanistic interpretability, and multi-layered safety architectures for next-generation models.
Developing international standards for facial recognition deployment. The Computer Vision Ethics Lab proposes governance frameworks addressing bias, accuracy thresholds, and national security considerations across allied nations.
Privacy-preserving distributed learning for classified environments. The Federated Learning Privacy Lab demonstrates how federated learning enables collaborative AI training across military departments without exposing sensitive data.
Making AI recommendations transparent for defense command. The Explainable AI Lab presents techniques for generating human-understandable explanations in real-time decision support systems for military operations.
Announcing a collaboration to advance AI safety research. TerraNova MU partners with Anthropic to jointly develop governance frameworks and safety techniques for frontier AI systems used in defense and government sectors.
Highlights from our first quarter 2024 publications. We published 34 peer-reviewed papers across top venues including USENIX Security, IEEE S&P, and ACL, covering AI governance, cryptography, and autonomous systems ethics.
Shaping the next generation of AI governance professionals. TerraNova MU launches an expanded certification program in partnership with leading universities, offering online and in-person training in AI governance, policy, and security.
Deep dive with our Byzantine Fault Tolerance Lab director. In this exclusive interview, Prof. Dr. Elena Martinez discusses her research on consensus mechanisms resilient to Byzantine failures and their applications in distributed AI systems.