AI Safety & Alignment Center

Researches AI alignment, mechanistic interpretability, robustness, and evaluation methods - including RLHF, red-teaming, and scalable oversight - to ensure advanced AI systems remain safe, controllable, and beneficial when deployed in high-stakes scientific and societal settings.

⚙️ GeoVerse System Administration Division 🖥️ AI & Computing Sciences Division 🔬 Basic Sciences Division ⚡ Intelligent Geophysical Exploration Division 🛢️ Resource & Energy Engineering Division 🌍 Applied Geoscience Solutions Division 💼 Economics, Policy & Strategy Division 🎓 Education & Training Development Division 🚀 Innovation & International Collaboration Division

Machine Learning Center Natural Language Processing Center Computer Vision Center High-Performance Computing Center Quantum Computing Center Generative AI & Foundation Models Center Reinforcement Learning & Agents Center AI Safety & Alignment Center AI for Science Center Robotics & Embodied AI Center

🔑

⭐ Center Chief

Norbert Wiener

Center Head

💡 Cybernetics founder, early warnings on AI risk

🔑

1835–1903

Auguste Kerckhoffs

Researcher

💡 Formulated Kerckhoffs's Principle — a system must remain secure even if everything about it except the key is public knowledge — the foundational security-engineering assumption underlying adversarial robustness and red-teaming evaluation of AI systems

🖥️ AI & Computing Sciences DivisionAI Safety & Alignment Center

🔑

1903–1972

W. Ross Ashby

Researcher

💡 Founded cybernetic control theory (Law of Requisite Variety, the Homeostat), establishing how a regulator must model a system's variety to control it — the systems-theoretic root of AI alignment

🖥️ AI & Computing Sciences DivisionAI Safety & Alignment Center

🧑‍🔬

🔑

1906–1985

Bruno de Finetti

Researcher

💡 Founded the subjectivist (Bayesian) theory of probability and the exchangeability theorem, and defined calibration as the criterion for a good subjective probability forecaster — foundational to evaluating whether an AI system's confidence estimates are trustworthy

🖥️ AI & Computing Sciences DivisionAI Safety & Alignment Center

🔑

1908–1988

Lev Pontryagin

Researcher

💡 Founded optimal control theory (Pontryagin's Maximum Principle), the mathematical framework for steering a dynamical system toward a goal under constraints — directly relevant to the theory of maintaining control/corrigibility over autonomous AI agents

🖥️ AI & Computing Sciences DivisionAI Safety & Alignment Center

🔑

1928–2015

John Forbes Nash Jr.

Researcher

💡 Founded non-cooperative game theory and the Nash equilibrium, the formal framework for analyzing strategic interaction between agents — the game-theoretic foundation of AI safety debate and multi-agent oversight protocols

🖥️ AI & Computing Sciences DivisionAI Safety & Alignment Center

🔑

1724–1804

Immanuel Kant

Researcher

💡 Founded deontological ethics (the Categorical Imperative — act only on principles universalizable and treating persons as ends, not merely means), one of the two dominant ethical frameworks (alongside utilitarianism) debated for encoding constraints in AI systems

🖥️ AI & Computing Sciences DivisionAI Safety & Alignment Center

🔑

1921–2002

John Rawls

Researcher

💡 Founded modern theory of distributive justice (justice as fairness, the veil of ignorance, the difference principle), the philosophical foundation most cited in algorithmic fairness research for defining what a 'fair' AI decision procedure should achieve

🖥️ AI & Computing Sciences DivisionAI Safety & Alignment Center

🔑

1903–1993

Hans Jonas

Researcher

💡 Formulated the 'imperative of responsibility' — that the unprecedented power of modern technology demands a new ethics of long-term, precautionary responsibility to future generations, foundational to AI governance frameworks

🖥️ AI & Computing Sciences DivisionAI Safety & Alignment Center

🧑‍🔬

🔑

1904–1985

Donald O. Hebb

Researcher

💡 Proposed the theory that neurons wire together via correlated activity ('cells that fire together wire together'), founding the study of how distributed representations encode concepts — the theoretical root of interpreting learned features in neural networks

🖥️ AI & Computing Sciences DivisionAI Safety & Alignment Center

🔑

1917–2008

Leonid Hurwicz

Researcher

💡 Founded mechanism design theory, showing how to design incentive structures so that self-interested agents' rational behavior reveals their true preferences and produces desired outcomes — directly foundational to inverse reinforcement learning and incentive alignment

🖥️ AI & Computing Sciences DivisionAI Safety & Alignment Center

🔑

1903–1930

Frank P. Ramsey

Researcher

💡 Founded subjective expected-utility theory, providing the axiomatic basis for representing an agent's preferences as a utility function derivable from choices — the mathematical foundation of reward modeling from human feedback

🖥️ AI & Computing Sciences DivisionAI Safety & Alignment Center

🧑‍🔬

🔑

1925–2019

Charles Perrow

Researcher

💡 Founded Normal Accident Theory — that catastrophic failures are inevitable in tightly coupled, complex systems regardless of safeguards — a foundational framework for analyzing systemic risk in complex, opaque AI systems

🖥️ AI & Computing Sciences DivisionAI Safety & Alignment Center

🔑

1915–1990

J. C. R. Licklider

Researcher

💡 Founded the vision of 'man-computer symbiosis' — humans and machines cooperating with complementary strengths under meaningful human oversight — the foundational framework for trustworthy, human-compatible AI system design

🖥️ AI & Computing Sciences DivisionAI Safety & Alignment Center

🧑‍🔬

🔑

1936–2001

Robert W. Floyd

Researcher

💡 Founded formal program verification (Floyd-Hoare logic, assigning preconditions/postconditions to prove program correctness), the mathematical foundation for formally verifying that safety-critical AI systems satisfy specified properties

🖥️ AI & Computing Sciences DivisionAI Safety & Alignment Center

🔑

1922–1983

Herman Kahn

Researcher

💡 Founded systematic future/scenario analysis for low-probability, catastrophic risks ('thinking about the unthinkable'), pioneering the methodology of existential-risk analysis later applied to advanced AI

🖥️ AI & Computing Sciences DivisionAI Safety & Alignment Center