Mustafa Suleyman: The AGI Race Is Fake, Building Safe Superintelligence & the $1M Agentic Economy | EP #216 - Moonshots with Peter Diamandis Recap

Podcast: Moonshots with Peter Diamandis

Published: 2025-12-16

Duration: 1 hr 26 min

Guests: Mustafa Suleyman

Summary

Mustafa Suleyman argues there is no 'race' to achieve AGI, emphasizing a collaborative approach to building safe superintelligence while discussing the economic potential of AI-driven agents.

What Happened

Mustafa Suleyman, CEO of Microsoft AI and co-founder of DeepMind, challenges the concept of a 'race' to achieve Artificial General Intelligence (AGI), stating that technological advancement is not a zero-sum game. He emphasizes the importance of safety in AI development, explaining that Microsoft is building a world-class superintelligence team and focusing on tools like Copilot to integrate AI into consumer services. The episode delves into the economic benchmarks for AI agents, with Suleyman proposing a modern Turing test where AI agents are judged by their ability to generate economic value, such as turning $100,000 into $1 million. Suleyman reflects on his time at DeepMind, describing the early years as a grind with little commercial success, and contrasts this with the rapid growth of AI's capabilities since 2022. The discussion touches on the unexpected affordability of AI technology, noting the drastic reduction in inference costs and the open-sourcing of models that cost billions to train. Suleyman stresses the need for international cooperation on AI safety, suggesting that the threat of a powerful, uncontrollable AI could unify global powers. He also explores the ethical and philosophical complexities of AI, including the potential for anthropomorphizing AI agents and the implications of AI achieving legal personhood. The conversation concludes with a look at the future of AI in various fields, including science and education, and the potential for AI to accelerate research and innovation.

Key Insights

• Technological advancement towards Artificial General Intelligence (AGI) is not a zero-sum game, as emphasized by the focus on safety and collaboration rather than competition.
• A modern Turing test for AI agents involves evaluating their ability to generate economic value, such as turning $100,000 into $1 million, as a benchmark for their capabilities.
• The cost of AI technology has decreased significantly, with inference costs dropping and models that cost billions to train being open-sourced, making AI more accessible.
• International cooperation on AI safety is proposed as a means to address the potential threat of uncontrollable AI, which could serve as a unifying factor among global powers.