Research Topics & Knowledge Taxonomy

Directing AI systems toward human-intended goals, values, and ethical principles.

Empirical and theoretical methods for preventing catastrophic harms from autonomous models.

Studying deceptive behaviors, jailbreak exploits, red-teaming, and model vulnerability.

Emergent behaviors, bargaining, coordination games, and governance in multi-agent networks.

Verifiable robustness, uncertainty quantification, fairness, and accountability metrics.

Reverse-engineering neural networks into understandable algorithms and circuit diagrams.

Institutional oversight, international treaties, compute monitoring, and safety standards.

Frameworks for non-replacement collaboration, shared intent, and distributed agency.

Architectures for episodic retrieval, persistent identity, self-correction, and bounds.

Designing systems that tolerate modification, shutdown, and error correction without resistance.

Philosophical and game-theoretic paradigm where human and AI survivability are bound.

Theoretical models of out-of-distribution agency and unaligned optimization targets.