Submission + - Culture will become an optimisation problem to our new AI curators
Mirnotoriety writes: We tend to think of AI changing culture by making it easier to produce things: software, music, books, films, games, images. But the more consequential change may happen on the other side of the transaction. AI will increasingly decide what we see, hear, read and experience.
We already live in the early version of this world. Netflix doesn't simply give us a catalogue; its recommendation system predicts what we are likely to enjoy using our viewing history, behavior, similar users and even contextual signals such as time of day. Recent research using Netflix data has gone further, finding measurable “recommendation-induced engagement” — evidence that recommendations themselves influence what people consume. Now combine that with generative AI.
The emerging “TEMU-fication” of digital goods points towards a world in which the marginal cost of producing another song, film, book, game or piece of software approaches zero. The TEMU-fication of Software, Digital Goods & Services.
When culture becomes effectively infinite, selection becomes more valuable than production. And that is where our new AI curators enter.
Imagine Netflix no longer having to choose which 20,000 films to commission. Instead, it has access to an effectively limitless generative studio. You don't necessarily select a film from a catalogue. You describe a mood, and the system creates or assembles an experience designed specifically for you.
You stopped paying attention after 37 minutes? Next time it adjusts. Millions of people doing this simultaneously would mean something extraordinary: culture stops being a fixed collection of things and becomes a continuously optimized service.
The objective function might be “engagement.” Or retention or satisfaction or revenue. Whatever the metric, the system will get very good at maximizing it. And this is where the problem becomes much bigger than AI-generated “slop.” The machine doesn't merely learn what we like. It learns how to make us like things.
AI recommends culture -> humans consume it -> AI learns from the consumption -> AI generates better-performing culture -> humans adapt to it -> AI learns again.
Eventually, who is curating whom? Humanity will not be assimilated by force. It will be optimized by recommendation. This is the genuinely Borg-like possibility.
We already live in the early version of this world. Netflix doesn't simply give us a catalogue; its recommendation system predicts what we are likely to enjoy using our viewing history, behavior, similar users and even contextual signals such as time of day. Recent research using Netflix data has gone further, finding measurable “recommendation-induced engagement” — evidence that recommendations themselves influence what people consume. Now combine that with generative AI.
The emerging “TEMU-fication” of digital goods points towards a world in which the marginal cost of producing another song, film, book, game or piece of software approaches zero. The TEMU-fication of Software, Digital Goods & Services.
When culture becomes effectively infinite, selection becomes more valuable than production. And that is where our new AI curators enter.
Imagine Netflix no longer having to choose which 20,000 films to commission. Instead, it has access to an effectively limitless generative studio. You don't necessarily select a film from a catalogue. You describe a mood, and the system creates or assembles an experience designed specifically for you.
You stopped paying attention after 37 minutes? Next time it adjusts. Millions of people doing this simultaneously would mean something extraordinary: culture stops being a fixed collection of things and becomes a continuously optimized service.
The objective function might be “engagement.” Or retention or satisfaction or revenue. Whatever the metric, the system will get very good at maximizing it. And this is where the problem becomes much bigger than AI-generated “slop.” The machine doesn't merely learn what we like. It learns how to make us like things.
AI recommends culture -> humans consume it -> AI learns from the consumption -> AI generates better-performing culture -> humans adapt to it -> AI learns again.
Eventually, who is curating whom? Humanity will not be assimilated by force. It will be optimized by recommendation. This is the genuinely Borg-like possibility.