Demis Hassabis predicts AGI arrival within two to three years

Demis Hassabis, co-founder and CEO of Google DeepMind, has updated his timeline for the emergence of Artificial General Intelligence (AGI), suggesting it could arrive within the next two to three years. This is a significant acceleration from his previous estimates of three to five years, or even five to ten years earlier in the year.

Hassabis defines AGI as a model possessing the full cognitive range of a human brain, capable of original thought and stable intelligence. To measure this milestone, he proposes the ‘Einstein test’: determining if an AI, trained only on human knowledge available prior to 1901, can independently rediscover the groundbreaking physics breakthroughs Albert Einstein achieved in 1905, such as special relativity and the photoelectric effect.

Beyond mere imitation, Hassabis aims to use AGI to understand the fundamental nature of reality and potentially derive new layers of physics. He noted that while Google previously hesitated to release early large language models to protect its reputation for accuracy, the pursuit of AGI remains central to DeepMind’s mission of solving complex scientific challenges.