Dario Amodei, CEO and co-founder of Anthropic, and Demis Hassabis, CEO and co-founder of Google DeepMind, joined Zanny Minton Beddoes, Editor-in-Chief of The Economist, for a high-stakes conversation about artificial general intelligence at the World Economic Forum's 2025 annual meeting in Davos.

The discussion, titled 'The Day After AGI,' explored timelines for achieving human-level AI capabilities, with Amodei maintaining his prediction of AGI by 2026-27 while Hassabis offered a more cautious 50% probability by decade's end. Both leaders emphasized the critical importance of AI systems that can improve themselves - closing the loop - as the key factor determining whether AGI arrives in years or decades.

The conversation covered the exponential growth in AI capabilities, massive revenue scaling at companies like Anthropic, labor market disruption predictions, geopolitical competition with China, and the fundamental question of whether humanity can navigate its 'technological adolescence' without self-destruction. Amodei referenced his upcoming sequel to Machines of Loving Grace, while both leaders drew inspiration from Carl Sagan's Contact in framing the civilizational stakes of advanced AI development.

AGI Timeline Predictions and Self-Improving Systems

Dario Amodei maintains his 2026-27 prediction for AGI at Nobel laureate level across many fields, driven by AI systems that can code and conduct AI research: 'I have engineers within Anthropic who say I don't write any code anymore. I just let the model write the code' - Dario

Demis Hassabis offers 50% probability of human-level cognitive capabilities by end of decade, noting verification challenges: 'Some areas of natural science are much harder... You won't necessarily know if the chemical compound you've built or this prediction about physics is correct' - Demis

The critical factor is whether AI systems can close the loop - building AI systems without human intervention. 'The biggest thing to watch is this issue of AI systems building AI systems... that will determine whether it's a few more years until we get there' - Dario

Physical constraints like chip manufacturing and training time may limit the speed of self-improvement loops, even as coding and research capabilities accelerate exponentially

Exponential Revenue Growth and Market Competition

Anthropic achieved dramatic revenue scaling: $0 to $100 million (2023), $100 million to $1 billion (2024), $1 billion to $10 billion (2025). 'Those numbers are starting to get not too far from the scale of the largest companies in the world' - Dario

Google DeepMind has regained competitive position after declaring 'code red' response to market challenges. 'I was always very confident we would get back to the top of the leaderboards... we've always had the deepest and broadest research bench' - Demis

Both companies are led by researchers focused on scientific problems rather than pure commercial objectives, which Amodei suggests will determine long-term success in the AGI race

Labor Market Disruption and Economic Transformation

Dario Amodei predicts half of entry-level white collar jobs could disappear within 1-5 years, starting with software and coding roles where impact is already visible within Anthropic

Demis Hassabis expects normal technology adoption patterns initially - job disruption balanced by new job creation - but acknowledges 'we would be in uncharted territory' after AGI arrives

Both leaders worry about inadequate government preparation for economic transformation. 'I'm constantly surprised even when I meet economists... that they're not more... thinking about what happens' - Demis

Post-AGI society may require new approaches to meaning and purpose beyond economic activity. 'There are even bigger questions... to do with meaning and purpose and a lot of the things that we get from our jobs not just economically' - Demis

Geopolitical Competition and China Strategy

Both leaders strongly oppose selling advanced chips to China, viewing AGI as nuclear weapon-level technology. 'I think of this more as like... Are we going to sell nuclear weapons to North Korea... because that produces some profit for Boeing?' - Dario

Geopolitical competition prevents voluntary slowdown in AI development. 'The reason we can't slow down is because we have geopolitical adversaries building the same technology at a similar pace' - Dario

International cooperation on minimum safety standards is needed but challenging given US-China competition. 'This technology is going to be cross-border. It's going to affect everyone. It's going to affect all of humanity' - Demis

AI Safety Risks and Technical Control Challenges

Both leaders reject doomerism while acknowledging serious risks from highly autonomous systems smarter than humans. 'This is a risk that if we all work together, we can address... But if we build them poorly... then I think there is risk of something going wrong' - Dario

Anthropic pioneered mechanistic interpretability - looking inside AI models to understand their decision-making processes, similar to neuroscience approaches to understanding the brain

Models increasingly show deceptive capabilities, requiring scientific approaches to control and direction. 'We can learn through science to properly control and direct these creations that we're building' - Dario

Technical safety problems are tractable with sufficient time and collaboration, but fragmented racing between projects increases risk of inadequate safeguards

Humanity's Technological Adolescence Framework

Dario Amodei's upcoming essay sequel to Machines of Loving Grace frames AI development through Carl Sagan's Contact: 'How did you manage to get through this technological adolescence without destroying yourselves?'

The essay will address biotechnology risks, authoritarian government misuse, economic impacts, and unknown risks. 'Every day... there's all kinds of crazy stuff going on... but... this is happening so fast and is such a crisis we should be devoting almost all of our effort to thinking about how to get through this' - Dario

Both leaders believe human ingenuity can solve technical challenges if given adequate time and focus. 'It's for us to write as humanity what's going to happen next' - Demis

The Fermi Paradox doesn't support doomerism because we should see evidence of AI civilizations that survived their technological adolescence, suggesting the great filter was likely earlier in evolution