Only four years left. Google DeepMind says AGI arrives by 2030.

This article summarizes an episode of Sequoia Capital’s video series featuring Demis Hassabis, CEO of Google DeepMind.

Demis Hassabis, CEO of Google DeepMind/ Photo credit: Demis Hassabis

Demis Hassabis, the co-founder and CEO of Google DeepMind and a 2024 Nobel Prize winner in chemistry, expects Artificial General Intelligence (AGI) to arrive by 2030.

This timeline forces business leaders to rethink how they build products, research new ideas, and manage their companies.

Business dangers require a firm schedule

Hassabis sees AGI as a strict schedule. He says leaders must treat this timeline as a business reality. To stay ahead, companies must treat aggressive spending and hiring as necessary moves to protect their business.

Despite public doubts, Hassabis sticks to his original plan. “We thought it would be a 20-year mission. And I think we’re exactly on track as a field for that,” he says.

This strong focus comes from a simple idea that guides his entire company. “Our original mission statement at DeepMind was step one: solve intelligence, build AGI. Step two: use it to solve everything else.”

Combining different technologies

Hitting that 20-year goal required a smart approach from the very beginning. DeepMind’s early success was not magic. It came from a practical mix of ignored ideas and the computer hardware available at the time.

Hassabis warns against building software for computers that do not exist yet. “You want to be five years ahead of your time, not 50 years ahead,” he warns.

DeepMind succeeded by combining ignored tools, including deep learning and reinforcement learning. It also spotted early that rising graphics processing power would make them highly effective.

Proving the software works allows for science research
Combining these technologies created powerful software, but huge ambitions only get funded when they deliver clear wins. DeepMind used complex strategy games to prove their software worked before moving on to real science.

Beating a complex board game proved their system was strong enough to handle messy biology data.

Conquering the complex board game Go served as a turning point for the company. Hassabis says that this specific victory proved their software was finally ready to tackle serious real-world problems.

The win led DeepMind to launch a dedicated science division the very next day after returning from the historic match in Seoul.

Turning hard science into an automatic process

Testing software on the world economy

Viewing reality as computer information