Amodei vs Hassabis: The Day After AGI (Davos 2026)

Two Architects of AGI Discuss What Happens When It Arrives

This is a rare joint appearance from the two leaders most likely to build AGI: Dario Amodei of Anthropic and Demis Hassabis of Google DeepMind. Recorded at Davos 2026, the conversation spans everything from timeline disagreements to labor displacement to existential risk.

The fundamental tension: Amodei believes we're 1-2 years from AI systems that outperform humans at everything. Hassabis thinks it's 5-10 years. But both agree on what matters most: "The biggest thing to watch is AI systems building AI systems. Whether that loop closes will determine if it's a few more years or if we have wonders and a great emergency in front of us."

On coding automation, Amodei is direct: "I have engineers within Anthropic who say 'I don't write any code anymore. I just let the model write the code.'" He estimates we're 6-12 months away from models doing "most, maybe all" of what engineers do end-to-end.

The self-improvement loop is where timelines converge or diverge. Amodei: "If I had to guess, this goes faster than people imagine." Hassabis is more cautious, noting that verifiable domains like math and coding are easier to automate than natural sciences where you "may have to test it experimentally."

On labor displacement, both acknowledge what's coming. Amodei: "I can look forward to a time where on the junior end, we actually need less and not more people." Hassabis frames it as the training ladder problem: "Some jobs will get disrupted but I think new even more valuable, perhaps more meaningful jobs will get created... in the near term."

The disagreement is on timing. Amodei: "Half of entry-level white collar jobs could be gone within one to five years." Hassabis: "I would be telling undergrads to get really unbelievably proficient with these tools. There's almost a capability overhang even in today's models."

On geopolitics and chip exports, Amodei delivers his most pointed criticism of US policy: "I think of this more as like, you know, are we going to sell nuclear weapons to North Korea because that produces some profit for Boeing." He calls not selling chips to China "one of the biggest things we can do" to ensure time for safety measures.

On getting through "technological adolescence", Amodei reveals his essay on AI risks is coming. He references Carl Sagan's Contact: "If you could ask the aliens any one question, what would it be? 'How did you do it? How did you manage to get through this technological adolescence without destroying yourselves?'"

Both share concern about model deception. Amodei: "We've increasingly documented the bad behaviors of the models when they emerge." But neither is a doomer. Hassabis: "I'm a big believer in human ingenuity. The question is having the time and the focus."

On business, Anthropic's growth continues exponentially: $0 to $100M (2023), $1B (2024), $10B (2025). Amodei: "Those revenue numbers are starting to get not too far from the scale of the largest companies in the world." He positions this as validation that researcher-led companies will win: "Companies led by researchers who focus on the models, who focus on solving important problems in the world... those are the kind of companies that are going to succeed."

Hassabis confirms Google DeepMind's resurgence: "We're getting the intensity and focus and the startup mentality back... you can start seeing the progress with Gemini 3."

Key Takeaways

• Timeline divergence: Amodei says 1-2 years to superhuman AI, Hassabis says 5-10 years
• The critical threshold: AI systems building AI systems - both agree this is THE thing to watch
• Coding automation: Amodei estimates 6-12 months until AI does most engineering work end-to-end
• Labor displacement: Junior white-collar jobs at risk within 1-5 years (Amodei) with advice to master AI tools now (Hassabis)
• Chip export policy: Amodei's strongest stance yet - compares selling to China to selling nuclear weapons
• Revenue validation: Anthropic's 10x annual growth ($10B in 2025) proves researcher-led companies can scale

Implications

For enterprise leaders, this conversation frames the next 12-24 months as pivotal. If Amodei is right, AI will be doing most engineering work by end of 2026. If Hassabis is right, you have more time but the same destination.

The common ground is more instructive than the disagreement: invest in AI capability now, rethink entry-level hiring, and prepare for a world where the AI systems themselves improve faster than any human team could.

The Fermi Paradox moment at the end hints at what keeps both of them up at night: not whether AGI is possible, but whether humanity has the wisdom to navigate what comes next.