Why we don't understand AI models yet
Breaking apart LLMs with Lee Sharkey
The political economy of AI
Building progressive technological futures.
AI is not inevitable. Generative Futures explores how democratic societies can steer technology, not be steered by it. Essays and podcasts with the researchers shaping AI and its alternatives, written to be understood without a technical background.
The idea
Escaping techno-determinism
We are trapped between two bad options: the belief that AI’s effects are inevitable, and the urge to simply reject it.
Almost every technology in history involved a careful, uncertain negotiation that could have turned out differently, and still could. Radio spread both democracy and Nazism. The dominant path of AI was contingent on cheap capital, GPUs, and cheap energy. Generative Futures digs into that contingency to show how malleable the future really is, and tries to articulate a positive vision for what we want from it.
Latest
New essays & podcasts
The radicalism of Positive Alignment
Could it go further?
Why LLMs can't count
Are they System 2 processors yet?
Who is building the agent economy?
Infrastructure born contested
How is AI changing our experience of Time?
It might be the antidote to social media
Why is AI shale-shaped?
The material contingency of the AI boom
Can Europe Catch Up on AI? with Carl Frey
Progress is not inevitable
Straight from the source
Latest from Substack
Pull the newest posts live, including any published since this site last updated that don’t yet have a page here.
Explore by theme
Topics
The Political Economy of AI
The founding idea of Generative Futures: technologies are malleable, not inevitable. Their impact is the product of political choices, institutions, and material conditions, which means a different future is possible.
How AI Actually Works
You cannot steer a technology you do not understand. These pieces open up the technical workings of modern AI: its architectures, its interpretability, and where today’s models break.
AI, Work & the Economy
Whether AI replaces or empowers workers is a political choice, not a technical destiny. These pieces examine automation, productivity, and the economics of an AI labour market.
AI Safety & Alignment
Making AI safe is as much a political and institutional question as a technical one. These pieces explore interpretability, alignment, and human flourishing as a design target.
AI & Geopolitics
AI is being built into the infrastructure of power. These pieces map the US–China race, the contested agent economy, and the politics of compute, sovereignty, and progress.
AI, Time & Society
Beyond economics and geopolitics, AI is reshaping how we experience time, memory, creativity, and learning. These pieces ask what kind of human life we want technology to support.
About Generative Futures
What is Generative Futures?
Generative Futures is a publication by Phil Bell on the political economy of AI. It explores how democratic societies can steer AI and emerging technologies toward a progressive future, rather than treating their effects as inevitable. It publishes essays and podcast interviews with researchers at the frontier of AI and AI safety.
What does "generative futures" mean?
The name rejects two traps: technological inevitability (believing AI’s effects are unavoidable) and nostalgia (simply rejecting new technology). Technologies are malleable, their impact is the product of negotiation between people, institutions, and the technology itself, so many different futures can be generated.
Who writes Generative Futures?
Phil Bell writes and hosts Generative Futures. He interviews AI researchers, economists, and political theorists, and writes plain-language explainers that make the technical AI debate legible to non-specialists.
Where can I read, watch, or listen to Generative Futures?
Essays are published on Substack (genfutures.substack.com). Video interviews are on YouTube (@generativefutures). The podcast is on Spotify and Apple Podcasts. You can also subscribe via RSS.
Get involved
Subscribe for new essays and interviews, or join the reading group, contribute, or suggest a podcast guest.