Unpredictable Patterns #125: AI in the noise society

Culture, authors, schools of thought and why we need to think more about moral rights

Dear reader,

I am just back from Almedalen - the political festival Sweden celebrates at the beginning of Summer on the island of Gotland every year. Lots of AI there - and interesting discussions about issues like national security and sovereignty and AI - a subject we will come back to, I am sure. This week’s note, however, discusses noise, AI-slop and moral rights as a surprising framing for thinking through reputation and cultural growth.

Noise wars

One of the earliest papers I wrote was titled “Noise Wars: Is the Answer to the Machine in the Noise”. It was published in 2003 as a contribution to the a conference that year. It was later reworked and published as “Noise Tactics in the Copyright Wars” (2006) in International Review of Law Computers & Technology, 20(3). The subject of the paper was the way that copyright holders had started to inject fake files into peer-to-peer networks to break them and make them useless for would-be teenage pirates. I was interested in two things: first the tactic itself, since it seemed to betray an acute understanding of how information systems could be attacked through noise of different kinds, and then second what that meant for the way in which the copyright litigants could use as evidence in court: if they submitted a list of file names, and it could be demonstrated that at least some of those files were noise files injected by the copyright holders, well, then that list in itself could hardly be good evidence, could it?

I would later expand this question about the role of noise in understanding public policy in the information society to what became my dissertation, where I looked at the uses of noise in preserving privacy and undermining free speech.

Now, I have not thought much about noise as a general policy problem since, but it seems clear that it could be interesting to revisit this issue in the context of artificial intelligence overall. What triggered me to make some notes about this was a recent essay by the brilliant Harry Law. In “The Slop Must Flow” he makes, if I simplify, the argument that what we call AI-slop is simply an exernality that is generated by the fact that it is easier and cheaper to make art. He ends the essay with:

The uncomfortable truth is that we are the heirs to a trend hundreds of years in the making. Art is getting cheaper to produce, and there’s more of it to enjoy, critique, and discuss.

If slop is the cost, then so be it.

He also notes that the notion that AI cannot make art seems more like a nervous statement from a creative elite than an actual philosophical insight.

The reason is simple: art is a way of separating us from them. When anyone can produce and distribute creative work, it threatens the integrity of a social apparatus whose legitimacy depends on deciding what counts as good taste. Art, as John Berger saw, is a servant of ideology.

Brave defenders of artistic life are less concerned with protecting art than their dubious positions as arbiters of cultural value. When a TikTok dance provokes more soul searching than a performance at the Royal Ballet, you have to wonder about the vitality of the cultural hierarchy.

Both of these observations seem quite right to me, and they strike, in some ways, at one of the core arguments for copyright - an argument that now seems to be wholly void: that we need copyright to incentivize creativity.

In the strongest version, this claim simply states that unless there is money in it for the creator they will stay their creative urges and remain silent much to the detriment of the rest of mankind - and this claim is then used to legitimize the baffling terms of protection, for example, that extend way beyond the life span of the creator (where no amount of incentives seem to be very effective to ensure continued creativity).

In reality, however, there seems to be little overlap between originality, quality and compensation in the overall content landscape - suggesting that this idea is one that we should abandon as fast as we can.

But there is another aspect of the challenge of AI-slop that interests me, and that is that it seems like a new kind of noise - cultural noise. And I do wonder if there may be a surprising role for copyright to play when it comes to dealing with this new challenge.

Cultural noise

I want to propose that we examine the following argument:

Some books are better than others, some music pieces are superior to others. Quality is not wholly subjective, but culture as a whole is a signal. This signal can be diluted by noise - and when that happens, it becomes more difficult to discover the valuable cultural artifacts and consume them. There are many different ways of assigning value to cultural artifacts, but such value is not arbitrary. Culture has at least one function that we should seek to protect - it enables social, psychological, political and philosophical progress. 

Now, I can see the objections. We may reject the idea that there are any objective values at all in culture, we could reject the idea that culture has any function - or if it does we can reject the idea that we should preserve that function. We can even reject the idea that there is culture at all. All of those positions can be taken, argued for and examined - but here I am going to assume them away, and study the impact of cultural noise.

At a very minimum I will suggest that cultures that preserve these functions are stronger than cultures that are largely negated in noise.

Cultural noise can be defined as artifacts or content that makes it harder to convey any coherent message culturally. It slowly erodes the possibility of a culture overall, by undermining the shared understanding that culture requires, but also by just turning people off from consuming culture overall. When culture becomes to noisy, it can be consumed as a way of spending a few idle hours, but it does not engage. It does not suggest that we are in any way responsible for taking a cultural stance.

The onslaught of AI-slop could, here, be a kind of cultural noise. It dilutes the function culture has, and makes it harder for societies to learn and discover new ideas and ideologies. It is not just music, movies and art in general, it is also political philosophy and literature, slowly drowning in noise.

It is a special case of Herbert Simon’s observation that with a wealth of information comes a poverty of attention, and a need to allocate attention efficiently. The wealth of information is the first part of the challenge, what noise does is that it destroys attention allocation. Cultural ideas are objects of collective intentionality - build on collective attention - and when that attention is disrupted by cultural noise no cultural ideas, nor cultural progress, is possible. Even culture itself then becomes impossible to sustain - and is replaced by content alone.

There is a temporal dimension to this as well: we should ask what happens to cultural memory when noise infects the systems. Any culture naturally seems to evolve a canon, but no such body of works can accrue in a society that is constantly under the pressure of even more noise. Culture thus is reduced not just to content, but emphemeral content.

Copyright to the rescue?

One surprising idea to counter this would be to look to copyright. Not the economic rights, but the moral rights.

Moral rights, as codified in international copyright law, affirm the personal and reputational bond between an author and their work. Under Article 6bis of the Berne Convention for the Protection of Literary and Artistic Works, authors are granted the right to claim authorship and to object to any distortion, mutilation, or other modification of their work that would be prejudicial to their honor or reputation.

These rights exist independently of economic rights and are retained even after the transfer of those economic rights. The WIPO Performances and Phonograms Treaty (WPPT) extends analogous protections to performers, ensuring their right to be identified and to object to distortions of their performances (Article 5). Unlike economic rights, moral rights are typically inalienable and, in many jurisdictions—especially civil law countries like France and Sweden—cannot be waived or sold, reinforcing the idea that authorship carries an enduring personal dimension beyond commercial exploitation. But this personal dimension could also be a recognition of the cultural dimension of these works.

The copyright industry has traditionally been obsessed with building systems that track use of works for economic concerns, and digital rights management technologies have been the goto-solution for this. With their roots in the bizarre early Electronic Copyright Management Systems - like the European project IMPRIMATUR (incredibly named!) from the late 1990s - these systems seek to track use of works and ensure that all use is accounted for economically. Such systems are increasingly now built into traditional content delivery systems in different ways, or made obsolete by streaming as a key delivery mechanism.

What would be more interesting, however, would be to build systems that allow us to trace, assert and track the authorship of cultural artifacts - and to ensure that those rights are respected. The act of claiming a work should come with accountability for the views in it as well, suggesting that you have chosen to enter into the dialogues that ultimately underlie the production of culture overall.

You put your name, and your intellectual reputation in play. Moral rights should be two-way: you claim your work, and you are also held responsible for it in some way.

The ability to search and sift content on the basis of reputation and accountability could be one way to revitalize the idea of culture as putting something at stake. A cultural work should be a form of gamble, associated with both possible upside and downside - it should matter. Culture - like other human endeavors, improve with skin in the game.

Allowing for the strong assertion of moral rights for works through technological means would be one way of supporting such a reversal of the race to a noisy end - and it would also allow for constructing new cultural infrastructures.

Authors, schools of thought and AI

We now should be rightfully worried about the return of elitism and a limited set of intellectuals producing culture: what Harry writes is right - we should celebrate the extension of the circle of possible contributors to culture and the fact that it is cheaper to produce art than ever before.

We don’t want to gate authorship on being accepted by a legacy publisher - for example. But what we want to do is to create new networked reputation clusters: where different authors can attest to the commitment of others. This could mirror the way a school of thought works - a common theme or set of ideas that allow for disagreement and debate, but within some kind of common framework. The rise of such virtual schools of thought would be a good alternative to traditional, narrow ways of producing culture.

And it goes without saying that we should expect to see hybrid thinkers and writers emerge - centaur philosophers (like centaur chess players) or musicians that work with various orchestrated artificial intelligences. We may even see single AIs emerge as members of such networked schools of thought, and should, were that to happen, celebrate it. That something is produced by AI is not the problem - the problem is the flat landscape where slop and cultural acts are indistinguishable from each-other.

In this world, moral rights would cover such hybrids as well - and authorship would morph into a network concept, but with retained, shared reputations.

Noise - revisited

Noise is a constant companion to information - and exploring the role of noise in economy, society and culture is important: it allows for a deeper understanding of where we need to build mechanisms for allocating attention, and filters to reform or reinvent existing institutions. Different technologies produce different kinds of noise - and AI is interesting in that it can produce many forms of noise. From hallucinations to low-quality content or slop.

At the end, however, the noise / information ratio for AI as a technology works out massively in the technology’s favor - especially if we combine the cheap production of content and information with attention allocation infrastructures of different kinds.

It will take time for those infrastructures to develop — it took time for the early newspapers in the US to become the revered publicist powerhouses that they really ended up being for a period of time - but we can accelerate the process, if we see the problem not as stopping the slop or reducing the noise, but augmenting the signal and elevating authorship.

Thanks for reading,

Nicklas