Highlights: #212 – Allan Dafoe on why technology is unstoppable & how to shape AI development anyway
Mar 12, 2025
auto_awesome
Allan Dafoe, Director of Frontier Safety and Governance at Google DeepMind, dives into the unstoppable nature of technology and its historical patterns. He explains how societies embracing new capabilities outpace those that resist. The discussion highlights the balance of offense and defense in technology, the complexities of AI cooperation, and the potential risks of backdoor technology in AI models. Dafoe also reflects on how human agency shapes tech development, emphasizing the need for ethical decision-making in AI's future.
Technological advancement is driven by competition, where societies adopting new capabilities tend to outpace and dominate those that resist change.
Increased cooperation among AIs presents both benefits and risks, as it may empower beneficial initiatives while also facilitating harmful collusion among malicious entities.
Deep dives
Understanding Macro Historical Patterns
History reveals complex macro patterns, suggesting that societal progress is not merely the aggregate of individual efforts but is influenced by broad structural dynamics. These dynamics lead to resistance or amplification of societal changes, as illustrated by examples such as the rapid economic recovery in Germany and Japan after World War II. These insights indicate that macro phenomena, like GDP growth trends, follow predictable patterns, such as Moore's Law in technology, which shows consistent exponential growth in transistor density. Understanding these patterns is essential to navigate future technological advancements and their implications for society.
Technological Determinism and Control
Discussions around technological determinism highlight the ongoing debate about the relationship between human agency and technological advancement. Early theorists claimed that technology possesses agency, directing societal evolution while humans merely adapt. In contrast, later sociologists emphasized the intricacies of decision-making processes in technological development, arguing that individual choices and socio-political contexts significantly shape technological outcomes. This shift in perspective underscores the complexity of control over technological progress, suggesting that while humans contribute to advancement, there are larger structural forces at play that can limit agency.
The Risks of Enhanced AI Cooperation
The potential for increased cooperation among artificial intelligences raises both opportunities and challenges, particularly regarding social dynamics and exclusion. Enhanced cooperation might empower many positive initiatives, yet it could simultaneously foster harmful coalitions among antisocial actors, mirroring the moral dilemmas in human cooperation. Historical examples, such as mafia organizations leveraging internal collaboration for harmful purposes, illustrate this duality of cooperative skills. Therefore, while promoting cooperative capabilities in AI may generally benefit society, it is crucial to critically evaluate the potential exclusionary effects on individuals and groups outside these systems.
Technology doesn’t force us to do anything — it merely opens doors. But military and economic competition pushes us through. That’s how Allan Dafoe — director of frontier safety and governance at Google DeepMind — explains one of the deepest patterns in technological history: once a powerful new capability becomes available, societies that adopt it tend to outcompete those that don’t. Those who resist too much can find themselves taken over or rendered irrelevant.
Are humans just along for the ride when it comes to technological progress? (00:03:58)
Flavours of technological determinism (00:07:11)
The super-cooperative AGI hypothesis and backdoors (00:12:50)
Could having more cooperative AIs backfire? (00:19:16)
The offence-defence balance (00:24:23)
These aren't necessarily the most important or even most entertaining parts of the interview — so if you enjoy this, we strongly recommend checking out the full episode!
And if you're finding these highlights episodes valuable, please let us know by emailing podcast@80000hours.org.
Highlights put together by Simon Monsour, Milo McGuire, and Dominic Armstrong
Get the Snipd podcast app
Unlock the knowledge in podcasts with the podcast player of the future.
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
Save any moment
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Share & Export
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode