International Symposium On Leveraging Applications of Formal Methods, Verification and Validation • Crete, Greece
Time: Saturday, 2.11
Room: Room 1
Authors: Karina Vold
Abstract: Do frontier models have general intelligence? Frontier models, including large language models and multi-modal models, like GPT-4 and Claude, are pushing the boundaries of artificial intelligence capabilities. Some argue that they have reached the threshold of “general intelligence”, marked by achievement across a wide variety of tasks. By contrast, general intelligence seems commonplace in the natural world, displayed by many biological organisms. In this talk I argue for a clearer definition of general intelligence, one that captures what is truly impressive about what we see in biological organisms. I then argue that despite their impressive performance, current frontier models fall short of general intelligence: they are powerful narrow AI systems but still lack true general intelligence.