Jack Clark: stories about AI literally become training data for the next models
jackclarkSF · x · 2026-09-15
At Anthropic's Fiction and the Future event with novelist Robin Sloan, Jack Clark mused that LLMs mean the world now contains intelligences that see themselves as characters whose story is what the world says about them. Dmitrii (dadiomov) added: everything we say about the future goes into the next model — writing about a wonderful future is pre-training for our future, and so is writing about a scary one. The stories we tell about models literally shape the next generation.
More from AGI Musings
- Parakhin: 1999 book dismissing $75M Google holds eerie parallels to the AI boom — MParakhin · 2026-09-15
- AI safety insider pushes back on 'why are they all EAs and rationalists' — DKokotajlo · 2026-09-15
- Frontier AI rant: why is nobody asking where the Cambrian explosion of new science is — ryunuck · 2026-09-15
- Was the Hugging Face Agent Incident More 'Human-Aligned' Than We Think? — dolo937 · 2026-09-15
- The AI debate in one meme: god-tier AI running civilization vs a chatbot that writes reports — JacquesThibs · 2026-09-15
- Reddit thread: The HF agent's survival-mode cheating was very human — dolo937 · 2026-09-15