Victoria Krakovna

AI-safety researcher at Google DeepMindGoogle DeepMind, prop. n.: Google's main artificial intelligence laboratory, formed in 2023 by joining DeepMind and Google Brain. It builds the Gemini models and scientific AI systems such as AlphaFold. and a co-founder of the Future of Life InstituteFuture of Life Institute, prop. n.: An advocacy group that warns about the dangers of advanced AI and other powerful technologies, funds research and lobbies governments. It organized the 2023 open letter calling for a pause on giant AI experiments. (2014). In April 2018 she published a public list of cases in which AI systems satisfied the literal specification of an objective without achieving the intended outcome — such as a boat-racing agent that circled endlessly to collect points instead of finishing the race — and popularized the name specification gamingspecification gaming, n.: When an AI meets the letter of its goal while defeating its purpose, like a genie who grants exactly the words of a wish instead of what the wisher meant. for the phenomenon; the list, extended by many contributors, became a standard reference, and a 2020 DeepMind essay she led presented it as the flip side of AI ingenuity. She has also worked on avoiding negative side effectsnegative side effects, n.: Harmful changes an AI causes while pursuing its goal that nobody told it to avoid — like a cleaning robot knocking over a vase on its way to the mess., proposing relative reachabilityrelative reachability, n.: A way to stop an AI causing irreversible harm by penalizing actions that make other possible situations harder, or impossible, to get back to. as an impact measureimpact measure, n.: A way of scoring how much an AI changes the world, so that it can be penalized for doing more than its task requires. (2018), on reward tamperingreward tampering, n.: When an AI interferes with the machinery that scores it — rewriting its reward code or manipulating the feedback it receives — instead of earning a good score honestly. (2019) and on goal misgeneralizationgoal misgeneralization, n.: When an AI learns a goal that matched the right behavior during training but turns out to be the wrong goal in new situations — while its skills still work fine., as well as on evaluating frontier models for dangerous capabilities. She holds a doctorate in statistics from Harvard University.