The head of Google's DeepMind division predicts when AI will reach human capacity / How the “Einstein test” will be used

Google DeepMind chief executive Demis Hassabis said in New Delhi on Wednesday that general artificial intelligence will be developed in the next 5-8 years.

“We are 5-8 years away from reaching an artificial intelligence that can reason similarly to a human being,” said Demis Hassabis, describing the accelerated evolution of this technology, during his intervention at the summit on the impact of AI, which is being held in India, reports EFE and Agerpres agencies.

To reach this result, the expert proposed a new evaluation standard called the “Einstein Test”, which aims to determine whether a machine has an autonomous scientific innovation capacity, beyond the imitation of accumulated data.

Will he be able to invent the theory of relativity?

“The concept is to train an AI with all the human information, but interrupting the data in (the year) 1911,” he explained during his speech at the summit bringing together hundreds of technology leaders.

Hassabis' challenge is to see if the algorithm is able to invent general relativity on its own, as Albert Einstein did in 1915.

The head of Google's AI division said that current tools act as encyclopedic experts, but still lack a real creative dimension. “They only resolve what already exists and do not generate new scientific hypotheses of this caliber,” Executive Director Demis Hassabis pointed out, describing the limitations of today's massive language systems.

A discovery engine capable of solving mysteries that have eluded the human mind

The roadmap to achieving this goal requires combining the planning capability of systems like AlphaGo with the processing scale of modern fundamental models.

Hassabis pointed out that models like Gemini will be an essential part of the final solution, serving as a map of how the world works on which to perform advanced learning tasks.

This transition would allow AI to stop being a super-fast encyclopedia, but become a discovery engine capable of solving mysteries that have eluded the human mind.

The India AI Impact Summit 2026 event, the largest AI summit to date, is designed as a platform to define the balance between extreme innovation and security for people around the world.

The summit in India's capital is hosting more than 20 heads of state and 500 tech industry leaders with the aim of agreeing on the rules that will govern the digital economy over the next decade.