The AGI Race: Hassabis and Amodei Disagree on When the World as We Know It Ends

At Davos, Demis Hassabis of Google DeepMind and Dario Amodei of Anthropic sat down at the same table to discuss "The Day After AGI" — and split on the central question: when will the day arrive that makes it meaningful to talk about an "after"? Hassabis is betting on the end of the decade. Amodei allows that unexpected jumps in AI capabilities could start much sooner.

Who argued about what at Davos

In the Swiss mountains, where the fate of the global economy is usually decided, attendees this time tried to work out when that economy will stop being a purely human affair. Davos delivered something the industry had long been waiting for: Demis Hassabis of Google DeepMind and Dario Amodei of Anthropic sat down at the same table. The moderator compared them to the Beatles and the Rolling Stones.

The comparison falls short: unlike rock stars, these two aren't arguing about the charts — they're trying to calculate the date when artificial intelligence catches up with humans or leaps over their heads. The billed topic was ambitious: "The Day After AGI." But before planning breakfast in a world of superintelligence, you need to know when that dawn will actually break.

Just a couple of years ago, AGI (artificial general intelligence) was the province of geeks and science fiction writers. Today it shows up in the quarterly reports of the world's largest corporations.

Hassabis's case: AGI as a tool for science

Hassabis — the man who effectively kicked off the modern AI wave when DeepMind was acquired by Google — has always been a cautious optimist. He builds systems that play Go, predict protein structures, and now try to reason. His vision of AGI is a tool that will solve problems in physics and biology. For him, this is evolution, however rapid: a powerful lever that will help humanity tackle what it cannot handle on its own.

Amodei's case: nonlinear growth and the point of no return

Anthropic's Dario Amodei plays a different game. His company was born out of a split at OpenAI over safety. If Hassabis is a research engineer, Amodei is someone who looks at exponential curves and sees not only progress but also a threat. At Davos, he again pointed to scaling laws: billions of dollars are pouring into compute, and neural networks are dutifully getting smarter, but that growth is nonlinear. The world could wake up to a next-generation model that suddenly shows capabilities no one is ready for, either legally or morally. Amodei stresses that the point of no return is closer than many regulators think.

Why the timeline gap matters

The difference in approaches exposes the industry's core conflict: the belief that the process can be controlled and AI gradually integrated into science and everyday life, versus the fear that the pace of technological development has already outrun the ability of social institutions to adapt. When the moderator asked for specific timelines, the tension was palpable: Hassabis is betting on the end of the decade, while Amodei hints that surprises could start much sooner. Those dates will determine where the next trillions of dollars in investment flow and which laws the EU and the US pass in the coming months.

The era of "toy" chatbots is ending, and the industry is moving toward agents that can act on their own. If the leaders of the two companies that effectively hold the keys to this technology disagree on the fundamental risks, that should give everyone pause: the engine builders are arguing over whether the car has brakes and where the cliff is. Meanwhile, both keep their foot on the gas, because stopping in this race means losing everything. The Davos meeting showed there is no consensus even at the very top, and "the day after AGI" could arrive suddenly for everyone involved.

Hassabis and Amodei agree on one thing: AGI is inevitable and close. But where Google sees it as a super-assistant for scientists, Anthropic warns of a tectonic shift the world isn't ready for.

When do Hassabis and Amodei think AGI will arrive?

Hassabis is betting on the end of the decade. Amodei hints that unexpected jumps in AI capabilities could start much sooner — he stresses that the point of no return is closer than many regulators think.

Why do Hassabis and Amodei disagree on AI risk?

Hassabis sees AGI as a tool and as evolution, however rapid — a lever that will help solve problems in physics and biology. Amodei, whose company Anthropic was born out of a split at OpenAI over safety, points to the nonlinear growth of model capabilities as compute scales and warns that the world may not be ready for it, either legally or morally.

Where do Hassabis and Amodei agree?

Both agree that AGI is inevitable and close. Where they differ is on what it means: Google sees AGI as a super-assistant for scientists, while Anthropic warns of a tectonic shift the world isn't ready for.