people
Roman Yampolskiy
AI safety and security researcher advocating precautionary controls and 'boxing' approaches
Roman Yampolskiy is a computer scientist at the University of Louisville known for early and persistent work on AI safety and security. Trained in computer science (BS/MS at Rochester Institute of Technology and a PhD from the University at Buffalo), he founded the Cyber Security Lab at Louisville and has published both technical work on intrusion detection and several influential books and papers on the risks of advanced AI, including Artificial Superintelligence: A Futuristic Approach Roman Yampolskiy — Wikipedia.
Yampolskiy stands out in the safety community for emphasizing a security‑minded approach: he has proposed practical mitigation concepts such as "boxing" potentially dangerous systems, introducing engineered weaknesses or "Achilles' heels," and systematically enumerating worst‑case scenarios to guide defenses Building safer AGI by introducing "artificial stupidity" (arXiv). He also encouraged formalizing the study of different kinds of minds (what he called "intellectology") as a way to reason about capabilities and limits.
On the scale of risk, Yampolskiy is explicit and precautionary. He has publicly warned that highly advanced AI could be uncontrollable and has advocated pauses or careful governance for large experiments; in interviews he has given high probabilities to extreme outcomes and has spoken about scenarios that would cause mass unemployment or existential catastrophe — claims that are reported in news coverage and should be read as his assessments rather than settled predictions Business Insider: Why this AI researcher thinks there's a 99.9% chance AI wipes us out.
Scholars and commentators treat Yampolskiy as a prominent early voice stressing that AI security is distinct from conventional software security: the incentives to conceal capability, the difficulty of verifying massive models, and the potential for rapid capability amplification all call for institutional responses beyond standard engineering practice. His proposals — while debated in the research community — help sharpen policy discussions around inspection, licensing, and pre‑deployment verification University of Louisville Q&A: UofL AI safety expert says artificial superintelligence could harm humanity.
Sources
- Roman Yampolskiy — Wikipedia
- Building safer AGI by introducing "artificial stupidity" (arXiv)
- Business Insider: Why this AI researcher thinks there's a 99.9% chance AI wipes us out
- University of Louisville Q&A: UofL AI safety expert says artificial superintelligence could harm humanity
- Roman Yampolskiy's homepage
- Roman Yampolskiy — Wikipedia
- Building safer AGI by introducing "artificial stupidity" (arXiv)
- Business Insider: Why this AI researcher thinks there's a 99.9% chance AI wipes us out
- University of Louisville Q&A: UofL AI safety expert says artificial superintelligence could harm humanity
- Roman Yampolskiy's homepage
Notable quotes
- "99.9% within the next hundred years" — reported in coverage of his Lex Fridman interview (YouTube) where he discusses extinction probabilities.
Videos