Skip to content

A Scenario of AI’s Future

Cofounder of Lightcone Infrastructure; builder of LessWrong & the AI Alignment Forum

A general introduction has to do two jobs: explain why people are worried, and make the moving parts memorable. AI 2027 puts those parts into a sequence. Compute, research automation and competition stop being separate talking points and start affecting one another.

I would hand this over first because a reader can point to a particular turn in the story and ask: why would that happen? That is a useful place to begin a conversation. The scenario gives the disagreement an address.

Top read
AI 2027 browser preview with a red title and branching Race and Slowdown chart

AI 2027

Kokotajlo, Alexander, Barnes, Larsen & Dean · 2025

Scenario2 hr

A concrete story gives you something to disagree with. This one leaves the machinery visible.

Where it falls short

A vivid scenario can feel more inevitable than its assumptions warrant. Read the dates as part of a model to interrogate. A coherent story is not, by itself, evidence that the story will happen.

The rest of the short list

COLD TAKES
The Most Important Century
Holden Karnofsky · 2021
SeriesSource facsimile
A gentler way into the subject

The most important century

Holden Karnofsky · 2 hr

Starts further back, with economic growth and the unusualness of this moment. I would choose it for someone put off by a detailed future narrative; it takes longer to reach the mechanisms.

Research report
Is Power-Seeking AI an Existential Risk?
Joe Carlsmith · 2022
  1. Introduction
  2. Six premises
  3. Timelines
  4. Incentives
  5. Probability estimate
  6. Objections
ReportSource facsimile
More explicit about the premises

Is power-seeking AI an existential risk?

Joe Carlsmith · 3 hr

Better for checking the argument one claim at a time. The report asks for more attention before giving a new reader a picture of the whole.

Who is this for?

You want a map of the argument before deciding which part deserves a closer look.

Why you should trust me?

For the last nine years I have built LessWrong and the AI Alignment Forum, where professionals and independent researchers discuss this subject. Along with the pieces themselves, I have read rebuttals, failed ideas and more comment threads than I would recommend.

I built and ran the LessWrong Annual Review, which evaluates the writing each year. I have also worked alongside people in the field as an editor, web developer, event host and occasionally an argumentative commenter at 2am.

I have watched arguments and research areas take shape, and I have watched some of them fail.

Two disclosures
  1. My team built the website for AI 2027, my top pick.
  2. I know many of these authors personally and have helped edit some of their writing. Ruben is a colleague.