Demis Hassabis says AGI could arrive in two or three years
🎧 Listen to this story
Demis Hassabis, co-founder and president of Google DeepMind, told The Wall Street Journal that artificial general intelligence (AGI) could be two or three years away. The estimate shortens his previous forecast of three to five years. Hassabis says AGI would have human-like cognitive abilities and could produce discoveries rather than merely combine existing knowledge; his benchmark includes an “Einstein test” based on knowledge available through 1901.
According to The Wall Street Journal, Hassabis expects future systems to help formulate physics theories and find treatments for currently incurable diseases. The executive cites AlphaFold, which was linked to the 2024 Nobel Prize in Chemistry, and argues that the AGI race should prioritize safety and broad distribution of its benefits. The Tecnoblog reports that he also proposes an international entity led by the United States to oversee advanced models before release.
Why it matters · editorial interpretation
Hassabis’s shorter forecast could intensify debate over how to evaluate AGI capabilities, particularly through his “Einstein test.” His proposal for an international entity to oversee advanced models also ties technical progress to decisions about safety and the distribution of benefits.