This conversation features Dario Amodei, CEO of Anthropic, and Demis Hassabis, CEO of Google DeepMind, discussing AGI timelines and societal implications. The discussion was moderated as a follow-up to their previous Paris conversation, focusing on when artificial general intelligence might arrive and its consequences for humanity.

The conversation covers AGI development timelines, with both leaders maintaining their previous predictions while acknowledging rapid progress in AI-assisted coding and research. They discuss the critical importance of AI systems that can build other AI systems, potentially accelerating development beyond current projections.

Key topics include labor market disruption, geopolitical competition with China, the risks of advanced AI systems, and the need for international cooperation on safety standards. Amodei references his essay Machines of Loving Grace and previews an upcoming piece on AI risks framed around humanity's 'technological adolescence.'

AGI Timeline Predictions: 2026-27 Still on Track

Amodei maintains his 2026-27 prediction for Nobel laureate-level AI across many fields, driven by models that can code and conduct AI research to create self-improving loops.

"We might be 6 to 12 months away from when the model is doing most, maybe all of what SWEs do end to end" - Amodei on current coding capabilities.

Hassabis holds to 50% chance of human-level cognitive AI by decade's end, noting verification challenges in natural sciences versus mathematics and coding.

The critical unknown is whether self-improvement loops can close without human intervention, which both leaders see as the key accelerant for AGI development.

Anthropic's Exponential Revenue Growth and Market Position

Anthropic achieved 10x revenue growth over three years: $0 to $100M (2023), $100M to $1B (2024), projected $1B to $10B (2025).

Amodei expresses confidence in independent model makers' viability, citing exponential relationships between cognitive capability and revenue generation.

Google DeepMind has regained competitive position after declaring "code red," with Hassabis crediting their research depth and renewed startup mentality.

Both companies are characterized as "researcher-led" organizations focused on solving important scientific problems rather than pure commercial objectives.

Labor Market Disruption: Half of White-Collar Jobs at Risk

Amodei maintains prediction that half of entry-level white collar jobs could disappear within 1-5 years, starting with software and coding roles.

Hassabis sees near-term job creation offsetting disruption, recommending students become "unbelievably proficient" with AI tools as better than traditional internships.

Both leaders acknowledge minimal current labor market impact but expect accelerating displacement as AI capabilities compound exponentially.

Post-AGI employment represents "uncharted territory" requiring new approaches to meaning, purpose, and wealth distribution beyond economic considerations.

Geopolitical Competition and China Strategy

Both leaders strongly advocate restricting chip sales to China, with Amodei comparing current policy to "selling nuclear weapons to North Korea because Boeing profits."

Amodei argues chip restrictions would transform US-China AI competition into manageable competition between companies rather than nations.

The administration's approach of binding China to US supply chains through chip sales is viewed as fundamentally misunderstanding the technology's significance.

International cooperation on minimum safety standards is deemed "vitally needed" given AI's cross-border impact on all humanity.

AI Safety Risks and the Technological Adolescence Framework

Amodei previews upcoming risks essay framed around Contact's question: "How did you get through this technological adolescence without destroying yourselves?"

His previous Machines of Loving Grace essay outlined AI's positive potential for curing cancer, eradicating tropical diseases, and advancing scientific understanding.

Key risk areas include autonomous systems smarter than humans, individual misuse for bioterrorism, nation-state misuse, economic displacement, and unknown unknowns.

Both leaders reject "doomerism" while acknowledging serious risks, emphasizing that technical safety problems are "tractable" with sufficient time and collaboration.