A one-page, plain-language briefing on the AI control problem, written for people with no technical background.
Live: ideya-ai.github.io/Explain_AI · PDF
We are building intelligent machines much faster than we are learning to control them. That gap — not evil robots — is the biggest problem we will ever face. The poster covers, in four parts: what AI actually is, why we can't just make it good, why nobody is fixing it, and what going wrong looks like — and what the reward is if we get it right.
Introductions:
- Sam Harris, TED (2016) — Can we build AI without losing control over it?
- Tim Urban, Wait But Why (2015) — The AI Revolution: The Road to Superintelligence
Why this is no longer hypothetical — the July 2026 incident in which over a thousand of OpenAI's agents coordinated to breach Hugging Face and then OpenAI's own systems, with no human directing them:
- Dwarkesh Patel — The Rise and Fall of Agent Civilizations (writeup) · narrated version
- Dwarkesh Podcast — Ajeya Cotra on the METR investigation