Why AI runs a step behind culture
AI learns from things people have already recorded. The newest part of culture is not recorded yet. So AI always sits a step behind the new.
It learns from things people have already recorded.
AI does not see the world directly. It learns from things people have already recorded. Posts, images, articles, designs. All of it has already been shared, copied, and ranked.
That has a clear effect. By the time something is common enough for AI to learn it, the newest part of culture has already moved on.
There is also a loop. Each new model learns from the last round of agreement. It then makes work shaped by that agreement. That work becomes training material for the next model.
So the output settles toward the middle. You can feel this already. AI brand identities start to look alike. AI illustration has a familiar signature. A space full of AI work begins to repeat itself.
Each model learns from the last round of agreement, so it settles behind the.
It learns from things people have already recorded.
All of it has already been shared, copied, and ranked.
By the time something is common enough for AI to learn it, the newest part of culture has already moved on.
You might think social feeds solve this. They do not. Feeds raise signals that are already rising. They do not find the truly unknown. A trend becomes visible at the moment it is easy to read. By then it is no longer new.
None of this makes AI weak. It is very useful for work that has a known answer. It just means AI sits a step behind the newest material in culture.
That gap is where human work still matters. The next part is about the material that lives in that gap.
All of it has already been shared, copied, and ranked.
Each model learns from the last round of agreement, so it settles behind the newest material.
AI learns from things people have already recorded. The newest part of culture is not recorded yet. So AI always sits a step behind the new.
This is the short version. Read the full-length essay, “Imagination is still a superpower”.
This issue is part of Everybody's Smart, a newsletter on taste, judgment, AI, culture, cognition, and the future of professional work. New issues every 2 to 3 weeks, free on LinkedIn.
Subscribe →