David Silver, the computer scientist who co-founded DeepMind and is renowned for teaching machines to master complex games like Go, has issued a stark warning about the trajectory of artificial intelligence. Speaking at the Web Summit in Lisbon on Thursday, Silver described AI as potentially the most significant technological advancement in the history of the human species.
Despite his pivotal role in developing AlphaGo, which famously defeated world champion Lee Sedol in 2016, Silver emphasized that his work focuses solely on solving specific problems rather than creating sentient machines. He clarified that his goal is not to build a ‘Terminator’-style robot, but rather to address narrow, defined challenges through machine learning.
However, Silver expressed concern that the rapid pace of progress is outstripping humanity’s ability to understand and manage it. He posed a critical question to the audience: ‘If you don’t know what you’re doing, someone else will.’ He argued that the future of AI is not predetermined and that society must actively shape its development.
Reflecting on the early days of his project, Silver recalled the initial resistance he faced. When he proposed teaching a computer to play Go, colleagues advised him to choose simpler targets such as checkers, chess, or even Pong. He noted that at the time, achieving superhuman performance in Go was considered unlikely within the next five to ten years, yet the technology advanced far faster than anticipated.
Silver suggested that while some outcomes are beyond our control, humans still retain the agency to influence the direction of AI. He urged that society should aim to ‘make sure that we go there,’ rather than passively allowing the technology to evolve without guidance. His comments highlight the growing tension between the relentless speed of innovation and the urgent need for strategic oversight in the AI era.
I just wish we had more concrete policy ideas instead of general warnings about the trajectory.
Finally, a realistic take from inside DeepMind. No sci-fi doom, just sobering responsibility for the pace.
If we don’t guide it, someone else will. Who exactly is ‘someone else’ in this scenario?
Remember when they said Go was too complex? We’re probably underestimating the next leap just the same.
His point about speed outpacing understanding is exactly why I’m nervous. We need governance now, not later.