Culture is Infrastructure
Why the Intelligence Age Has No Institutional Answer to Creative Economies
A few years ago, I become obsessed with (and did a mini podcast series on) art and culture forgery and thievery. Fake wine, forged art, replica watches, and more. Now, being a writer with my own IP in the cultural space, I have once again become obsessed with the impact of LLMs and AI on the value of IP, and subsequently contract law, which eventually feeds into industrial strategy!
This paper contributes to research I’m doing with Alex Fleetwood on cultural institutions and new models of industrial strategy in the Intelligence Age.
In short, the institutions that governed creative economies for three centuries, such as copyright law, collecting societies, arts councils, and public broadcasters, were designed for a world of (1) discrete works, (2) identifiable copies, and (3) traceable distribution.
But… AI training dissolves all three! And still, nobody has built what comes next…!
Consider that two decades of creative work across every language, every tradition, every medium that has touched the open web, have been absorbed into AI models and been reproduced via “model training weights”.
Is this theft? No, because theft implies goods that could be returned.
I argue that this is something slightly different; enclosure: a structural transformation after which the prior arrangement is no longer available. A book hasn’t been copied by AI in the traditional sense, but it has been dissolved.
Additionally, think about this: copyright protects copies; but with AI, there is no copy. Moral rights protect attribution; but with AI there is no attributable act. The Suno/RIAA cases, the Authors Guild suits, the EU’s collapsed text-and-data-mining consultations… These are all surface-layer expressions of the same structural problem: the institutional architecture was designed for a mechanics of culture that no longer exists.
Simply put, the conditions that cultural institutions and laws were designed for just don’t exist any more.
The framework in this paper explains why institutions are coordination architectures. What matters is not whether they’re “good” institutions, but how fast they can change, also knows as their adaptive bandwidth:
Copyright law revises on legislative timescales: decades.
Collecting societies and museums update on administrative timescales: years.
AI capability is changing on a timescale of months.
As we see with institutions outside of culture as well, you cannot have both high stability and high bandwidth; and every major cultural institution was optimized for stability. Thus, the architecture gap is widening, and culture is in the premature regime: it has been technically transformed by AI, but institutionally abandoned as museums, writers and the law cannot keep up.
The paper here applies the architecture gap framework to the domain where the coordination failure is most acute and least addressed. It identifies seven domains any institutional response must span (memory, perception, production, economics, distribution, judgment, and rights) and explains why solving any one in isolation changes the problem in the others.
Why does this matter? In short, it’s because industrial policy for the intelligence age is currently building on a foundation that is actively being dissolved in real time.
Read the full paper here paper →
This article applies the architecture-lag and coordination-architecture frameworks developed formally in two companion papers by me, Sinéad O’Sullivan: “Institutions as Coordination Architectures” and “Market Formation as a Systems Engineering Problem.” Available on request: s@sinead.co
So interesting Sinead, the Irish perspective on copyright is interesting, after all it is the location of one of the first copyright disputes, the copying of a manuscript by Colmcille caused a war, and resulted in his voluntary exile to Iona. 'To each cow, its calf' was the original verdict, which condems AI scraping training data two millenia later.
Sinead, goods which can be returned? So there is a specific offence under the road traffic act in Ireland for taking a car and say joy riding around. The intention according to law is not to permanently deprive the owner of his / her property. The offence is not considered to be one of theft but unauthorised taking. Now if the Gardai establish that the thief has taken the car and say filed off the engine number or changed the number plates these actions elevate ihe offence into the theft category. I'm wondering if its possibkecto apply a similar legal approach to the question of data? Not theft but an offence of unauthorised taking of data. Perhaps civil law is far more complicated?