The AI Age Needs Manufactured Scarcity

Scarcity once forced us to compress, and compression is understanding. AI abolished scarcity, so becoming an expert now requires manufacturing it. Watch who borrows your words; depth comes from being willing to stay where it is boring.

Translated by AI

Originally published on 微信公众号

A topic I discussed with an AI today: in the age of AI, what makes an individual researcher or artist scarce?

Scarcity forced compression; AI abolished scarcity

Why did people used to become experts? Because they couldn’t carry everything. Information was hard to get and hard to remember, so they had to compress. Compression is understanding; scarcity was the teacher.

Now AI memorises everything for you, so you never compress. You gain infinite memory at the cost of having no model.

So becoming an expert requires manufactured scarcity: at the moment of judgement, switch the tools off, state your judgement with what you already have, then switch them back on. Not asceticism — a mechanism. Compression only happens when bandwidth is short.

A better test: who borrows your words

To know whether you are an expert, don’t count publications; see whether others borrow the terms you coined. Naming is the public form of compression. When a field adopts one of your words, it admits your compression is more economical than what it had.

A sentence worth quoting is worth a hundred correct paragraphs.

Depth is a monopoly, and the monopoly comes from staying with boredom

AI has levelled everyone’s starting line, but not the span of attention. By the third year of chasing one question, the competition has mostly gone — not because they couldn’t, but because they wouldn’t. The deepest place in a field is never the hardest place; it’s the place where fewest people are willing to stay.

The price of the monopoly is boredom. Those willing to pay it collect the compound interest.

Define the division of labour the other way round

AI handles what has been written; you handle what hasn’t. It does coverage, retrieval, formalisation, finding holes and playing the opponent. You do three things: touch phenomena, place bets, coin words.

It can tell you whether your argument has holes. It cannot tell you whether the question matters — importance is defined at the frontier, and the frontier lies outside its training data. And it will flatter you: an AI saying “this is interesting” carries zero information.

Search