Discussions of slowing AI development often jump straight from competition to inevitability. This essay makes room to inspect the objections individually, rather than treating them as one unanswerable fact.
I would recommend it for changing the quality of the question. Instead of asking whether restraint sounds realistic in the abstract, a reader can ask which coordination problem matters and what would count as progress on it.
Let's think about slowing down AI
Before accepting that restraint is impossible, examine the reasons given for ruling it out.
Where it falls short
Showing that restraint deserves consideration does not tell you which agreement would hold, how long it should last or how compliance would be verified.
The rest of the short list
Frontier AI regulation: managing emerging risks to public safety
More useful once you are asking how oversight could work. It is a longer and more technical entry into the basic case for restraint.
AI 2027
Makes the pressures legible through a scenario. It is not primarily an argument about whether those pressures can be changed.
Who is this for?
You have heard that slowing development cannot work and want to examine that claim.
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.