Amodei and Hassabis on Life After AGI at Davos 2026

At the World Economic Forum Annual Meeting in Davos in January 2026, Anthropic CEO Dario Amodei and Google DeepMind CEO Demis Hassabis sat down for a session titled The Day After AGI. Moderated by The Economist editor in chief Zanny Minton Beddoes, the conversation was billed as a rare joint appearance by two leaders who have shaped much of modern frontier AI research. The Radio Davos podcast and WEF video archive preserve the full exchange; this piece reconstructs the public record for readers who want substance rather than headline fragments.

Why this session mattered

Both executives have spent years arguing that artificial general intelligence is not a distant science fiction plot but a live engineering trajectory. Amodei has repeatedly tied model capability to revenue in exponential terms, while Hassabis has emphasized scientific applications from protein structure to materials discovery. Putting them on one stage forced a comparison of timelines, labor market assumptions, and the distribution question that follows a productivity shock.

Beddoes opened by returning to Amodei's 2025 Paris prediction: a model performing at Nobel laureate level across many fields by 2026 or 2027. Amodei stood by the broad direction. "It is always hard to know exactly when something will happen," he said, "but I do not think that is going to turn out to be that far off." He pointed to coding as the clearest early domain where models already behave like skilled collaborators rather than autocomplete tools.

Hassabis held a slightly more cautious line. He reiterated his view that there is roughly a fifty percent chance of systems exhibiting the full range of human cognitive capabilities by the end of the decade. Verifiable domains such as mathematics and software engineering, he argued, are easier to automate because outputs can be checked. Open ended reasoning, physical interaction, and long horizon planning remain harder.

Capability, capital, and independent labs

A substantial portion of the session explored whether standalone model companies can survive the capital intensity of the next scaling generation. Amodei described a compounding loop: more compute yields more capable models, which yield more revenue, which funds more compute. He acknowledged investor skepticism but argued Anthropic's trajectory already shows the pattern.

Hassabis, operating inside a larger corporate parent, framed DeepMind's recent return to state of the art benchmarks as a product of research depth and renewed operational focus rather than a single breakthrough. The contrast between an independent lab and a platform anchored inside Google became a subtext for how safety, productization, and geopolitical pressure will be managed differently across the ecosystem.

Jobs, institutions, and the day after

On employment, both rejected the idea that current model performance already explains labor market statistics. Amodei stressed that when he discussed future displacement he was describing a forward looking scenario, not claiming present day evidence. Hassabis predicted disruption in some roles alongside creation of new, potentially more meaningful work, a pattern familiar from prior industrial transitions but accelerated by software that learns.

Where they converged more sharply was institutional readiness. Hassabis said he is "constantly surprised" that professional economists are not modeling post AGI distribution with more urgency. Amodei echoed concern that racing without guardrails could produce poorly controlled systems. Both referenced scientific upside, citing AlphaFold and disease focused spinouts as templates for unequivocal public good, while agreeing the industry's balance still skews toward general capability over targeted humanitarian deployment.

Safety, geopolitics, and unequivocal goods

The Davos audience pressed on risk framing. Amodei maintained his long standing view that powerful AI can address cancer, tropical disease, and fundamental science while carrying "immense and grave risks" if built recklessly. Hassabis described post AGI society as "uncharted territory" where humanity must deliberately write the next chapter rather than drift into it.

Neither executive offered a detailed governance blueprint on stage. Instead they illustrated the tension facing 2026 policy makers: capability is visibly accelerating, verification friendly domains are automating first, and the social contract for sharing gains remains underspecified. For Classy AI News readers, the interview is less a forecast document than a status check from two operators who still believe coordinated science, not competitive panic, is the path through the transition.

Sources

World Economic Forum, The Day After AGI session page, weforum.org

Radio Davos podcast, Day After AGI episode transcript, weforum.org

YouTube, WEF Davos 2026 Day After AGI full session, youtube.com