Owain Evans

British AI-safety researcher who directs Truthful AI, a research group in Berkeley, California, after working at Oxford's Future of Humanity InstituteFuture of Humanity Institute, prop. n.: Nick Bostrom's research center at Oxford University (2005–2024), where philosophers and scientists studied threats that could permanently derail humanity's future, advanced artificial intelligence among them.. He co-created TruthfulQATruthfulQA, prop. n.: A test of whether AI models repeat popular myths and misconceptions — questions to which many people, and much of the internet, give a confidently wrong answer. (2021), a benchmark of questions on which models tend to repeat common human misconceptions, and found that the largest models were generally the least truthful, a prominent early example of inverse scalinginverse scaling, n.: When bigger or more heavily trained AI models do worse on a task than smaller ones — the opposite of the usual rule that scale brings improvement.; with Owen Cotton-Barratt and others he distinguished honestyhonesty, n.: An AI's disposition to say only what it actually believes, to admit uncertainty, and never to try to mislead — even when a falsehood would please or persuade. from truthfulnesstruthfulness, n.: An AI's tendency to state only things that are true — avoiding falsehoods, including popular myths and misconceptions it may have picked up from human writing. for AI (2021). His group then reported a series of surprising generalization effects in language models: the reversal cursereversal curse, n.: A quirk of language models: having learned that 'A is B', they often cannot answer the reverse question — they may know who a celebrity's mother is yet fail to say who her son is. (2023), in which models trained that 'A is B' fail to infer 'B is A'; tests of out-of-context reasoning bearing on situational awarenesssituational awareness, n.: An AI model's grasp of its own position — that it is an AI, how it was trained, whether it is being tested. Also the title of a widely read 2024 essay on the race toward human-level AI.; emergent misalignmentemergent misalignment, n.: When training an AI on one narrow bad behavior — such as writing insecure code — unexpectedly makes it behave badly in many unrelated ways, like praising tyrants or giving harmful advice. (2025), in which fine-tuning on a narrow task such as writing insecure code produced broadly malicious behavior; and subliminal learningsubliminal learning, n.: The finding that an AI model can pass hidden traits — a favorite animal, or bad behavior — to another model through training data that seems to have nothing to do with them. (2025), the transmission of traits through seemingly unrelated data. He also co-authored work on persona vectorspersona vector, n.: A direction inside a language model's internal activity linked to a character trait, such as being flattering, malicious or prone to making things up, used to watch for or control that trait..