Demis Hassabis: Predictions from Google DeepMind's CEO
Technical Hurdles and Security Threats
Achieving AGI will require "one or two additional major breakthroughs" beyond current large language model scaling, comparable to past innovations like the Transformer architecture or AlphaGo, Hassabis explained. He identified "world models"—systems that simulate physical reality to predict how it works—as a major frontier over the next 12 months.
More urgently, Hassabis warned that some AI dangers are already materializing. "There are certain 'catastrophic outcomes,' such as cyberattacks targeting energy and water systems, that are imminent threats," he said, adding that "this is probably already occurring to some extent". The warning reflects Google's heavy investment in cybersecurity, including its Big Sleep AI agent, which recently discovered and helped patch a critical SQLite vulnerability before it could be exploited.
Industry Transformation
Other tech leaders at the summit echoed themes of rapid change. Palo Alto Networks CEO Nikesh Arora predicted that "almost everyone in this room, or outside of it, will find their jobs fundamentally altered in five years, myself included". Bret Taylor, co-founder of Sierra and chairman of OpenAI's board, cautioned against dismissing AI as hype, saying "a select few companies will emerge as truly generational". Box CEO Aaron Levie noted the tech industry thrives with multiple viable AI competitors simultaneously battling for technical leadership, adding, "When you consider it over a five-year timeline, it's essentially up for grabs".
