Demis Hassabis, co-founder and current president of Google DeepMind, expressed his belief that artificial general intelligence (AGI) may be only two or three years away. This new estimate represents a shortening compared to the researcher's previous forecast, made in early 2026, when he indicated a period between three and five years.
Hassabis disclosed this projection during an interview given to the Wall Street Journal. As Chief Scientist at Alphabet, he considers this novelty to be a system capable of aggregating cognitive abilities comparable to human ones, unlike systems that only show high performance in very specific tasks.
There is no universally accepted definition regarding the exact moment an AI will achieve AGI status. Generally, the concept refers to systems capable of learning, reasoning, and solving various types of problems like a human would—an idea often associated with science fiction. AGI remains a theoretical concept, as today's available systems do not yet exhibit all these expected cognitive capabilities.
Hassabis uses a more rigorous evaluation standard called the 'Einstein test.' This test would involve training a system using only human knowledge available up to 1901 and then verifying whether it could independently achieve the advancements made by physicist Albert Einstein in 1905, including the theory of special relativity.
The purpose of this test is to measure the capacity to generate new discoveries, something current models struggle to demonstrate, in contrast to merely combining pre-existing knowledge. For Hassabis, a fundamental application of the most sophisticated systems will be accelerating the process of scientific discovery.
Google DeepMind has already illustrated this approach with AlphaFold, a system designed to predict protein structures, which earned the researcher part of the 2024 Nobel Prize in Chemistry. Furthermore, the executive leads Isomorphic Labs, an Alphabet company focused on applying AI to new drug development.
According to the Wall Street Journal, Hassabis anticipates that future AIs may assist in formulating new physical theories and contribute to finding cures for currently incurable diseases. The expectation is that AGI will act as a partner to scientists, capable of proposing hypotheses and locating unknown solutions.
Despite the optimistic tone, Hassabis emphasizes that winning the race for AGI does not just mean being the first organization to develop the technology. For him, the main focus must be on implementing these systems safely and ensuring their benefits are widely accessible to society.
Concerns regarding risks accompany these positive predictions. The executive himself advocates for the creation of an international body, led by the United States, to supervise advanced models before their release.