In short

The founding essay of Generative Futures rejects techno-determinism, arguing AI's societal impact stems from political choices, not inevitability. The rise of transformer LLMs was contingent on cheap capital, GPUs, fracked energy and waning social-media trust. Drawing on Acemoglu and Autor, it argues AI can augment rather than replace workers if societies choose, and urges us to develop a positive vision for AI now.

This is the founding statement of Generative Futures, the essay that names the problem the whole project is trying to solve.

We are stuck in a debate between two bad options: inevitability, the belief that the effects of AI are unavoidable and we can only adapt, and nostalgia, the urge to reject new technology outright. Both surrender our agency. The premise of Generative Futures is that there is a third path: technologies are malleable, and their impact is the product of a negotiation we can still influence.

The essay traces how contingent the current path of AI really is, how cheap money, GPUs, and cheap energy shaped which ideas won, and argues that whether AI replaces or empowers people is a political choice, not a technical destiny. If that’s right, the most useful thing we can do is start articulating a positive vision now, before the next crisis decides it for us.

Key takeaways

  • Technologies have no fixed trajectory; their impact reflects collective political choices, not inevitability.
  • Transformer LLMs 'won the hardware lottery' amid specific economic conditions rather than being optimal or inevitable.
  • Whether AI replaces or augments workers (per Acemoglu and Autor) is a policy decision that can reduce inequality.
  • Without positive visions for AI articulated now, societies default to whatever markets and geopolitics produce.
  • The project aims to make AI legible to non-specialists so broader constituencies can shape its future.

Read the full piece

This is a summary. Read the complete essay, with all the sources and argument, on Substack.

Read on Substack →

Frequently asked questions

Is AI's impact on society inevitable?
No. Technologies are malleable and their effects result from political choices. Radio, for example, served both democracy and Nazi propaganda, the same technology, different negotiated outcomes.
Why did transformer-based AI become dominant?
It emerged from contingent conditions, GPU availability, cheap capital from quantitative easing, low-cost fracked energy, and declining trust in social media, rather than being inevitable or technically optimal.
Can AI reduce inequality instead of causing job loss?
Yes, if steered by policy. Economists like David Autor argue AI could augment workers and rebuild the middle class rather than simply replacing labour.

People & ideas in this piece

Joseph WeizenbaumDaron AcemogluDavid AutorBernard StieglerWill DaviesDario AmodeiWill MacAskillMilton FriedmanTechno-determinismPolycrisisHardware lotteryQuantitative easing

Topics: The Political Economy of AI