New AGI benchmark: Demis Hassabis proposes ‘Einstein test’—Ultimate challenge to prove true intelligence

Demis Hassabis, the CEO of DeepMind Technologies, has proposed an ultimate benchmark for defining Artificial General Intelligence (AGI).

While discussing AGI during the panel discussion, Hassabis said, “My definition of AGI has never changed, which I can tell you is as the system that can exhibit all the cognitive abilities humans can.”

Given the prominence of the human brain in general intelligence, all the systems should be able to beat this architecture of intelligence.

According to the 49-year-old, although today’s AI systems are impressive in working, they lack general consistency across all the boards as sometimes they fall short of solving simple math problems.

Hence, Hassabis has proposed the “Einstein test” , a true mark of AGI. “The kind of test I would be looking for may be training an AI system with a knowledge cut off at 1911 and then see if it can come up with the general relativity problem like Einstein did in 1950. if yes, it’s AGI,” he added.

The CEO of DeepMind is also of view that AGI will potentially arrive within the next five or ten years, possibly after 2030 as he called the year 2026 another pivotal moment for AI.

While speaking at AI Impact summit 2026, Hassabis said AGI has the potential to reshape the world faster than any prior technological shift. 

According to Hassabis, “If I were to try and quantify what’s coming down the line with the advent of AGI. It's going to be something like ten times the impact of the Industrial Revolution, but happening at ten times the speed — probably unfolding in a decade rather than a century."