The Amsterdam Law & Technology Institute’s team is inviting scholars to publish guest articles in the ALTI Forum.
Here is the latest contribution authored by Dan L. Burk (University of California, Irvine School of Law).
Artificial intelligence or “AI” systems are increasingly a presence in the generation of creative and innovative products. There is, to be sure, nothing that might be robustly considered “intelligent” in such AI devices, which are a species of machine learning technology that might better be termed “pattern recognition software” or “iterative statistical optimization systems.”
But even without unnecessarily anthropomorphic labels, these technologies provide tools to create outputs that fall within the traditional subject matter of intellectual property, or “IP” regimes, such as copyright or patent. For example, it is now routine for machine learning systems to write short newspaper features, such as sports score reporting, and AIs are increasingly taking on more ambitious texts such as novels or move scripts. AIs have been used in the generation of paintings or other novel graphics. On the technical side, they are increasingly used to design new machines, develop new pharmaceuticals, or optimize industrial processes.
These capabilities foreshadow a fundamental shift in the structure of creative industries. Consequently, the burgeoning role of AI systems in the creation of intellectual goods has raised questions about the future role of intellectual property law in fostering the generation of creative works. Much of what has been written on the topic to date has mistakenly focused on unproductive questions about whether machine learning can be considered an “inventor” or an “author” under existing intellectual property regimes. Such inquiries not only take the euphemism of “artificial intelligence” too seriously, they largely miss the actual points of impact AI is having on creative production.
Justifying Intellectual Property
In order to focus on the correct set of questions for AI and IP, we must first remind ourselves of the policy justifications for deploying the exclusive rights entailed in intellectual property systems. Our usual justification for the existence of intellectual property rights centers on some version of appropriability or the “public goods” problem. Technological advances increasingly lower the costs for reproduction and distribution of creative works; the initial creation and development of valuable intellectual goods is difficult or capital intensive, but once they come into existence they are easily appropriated. This makes recouping the initial investment difficult, since it is nearly impossible to exclude “free riders” who choose not to pay for the appropriable good. Intellectual property law seeks to impose legal exclusivity on goods whose appropriation cannot easily be physically restricted.
This narrative is well illustrated, for example, in the history of copyright law, beginning with the printing press and running through successive technological developments up to modern digital computing systems. In a world where books are copied by hand in a scriptorium, there is no real need for copyright law, as copies remain few and expensive. Only once reproduction becomes cheap does copyright become a necessity. A similar story can be told in patent law, taking for example pharmaceutical generic copying as a paradigm case. Drug discovery and regulation is expensive, but mass production of the final product is relatively cheap. Patents are believed to be necessary to recoup the initial costs. Thus, intellectual property is justified as a legal tool to prompt investment in goods where creation is expensive but appropriation is cheap.
At the same time, we know that there are substantial social costs imposed by restricting access to intellectual goods. Intellectual property is a restraint on trade, curtails valuable protected speech, and imposes artificially inflated prices on consumers. Our hope is that the value arising from the incentive for generating additional creative works exceeds the costs created by exclusivity, although that balance between incentives and access is often difficult to measure. The evidence for successful provision of incentives is equivocal at best. Thus, provision of intellectual property incentives is to some extent a matter of ideology and faith with uncertain benefits.
The incentive story is the usual story. But note that this rationale for intellectual property has little or nothing to do with the applications of AI to creation as mentioned at the beginning of this essay. The standard justification for intellectual property anticipates development of valuable intangibles that are expensive to create initially, but inexpensive to distribute, copy, or appropriate thereafter. And there are no doubt some applications of machine learning that lower the costs of appropriability. But the transformative uses of AI now under contemplation work at a different point in the development of intangibles – at the initial phase of creation.
Thus, rather than lowering the cost of appropriability for works that are expensive to create, AIs make initial creation itself inexpensive. AI systems, as I have described above, lower production costs by automating the initial creative stage of product development, replacing a type of labor that formerly was reserved exclusively for the human intellect. Stated differently, such automated labor in effect provides a synthetic substitute for human creativity. I set aside here any discussion as to whether AIs are “truly” creative in the generation of such inputs. My own view is that, lacking any degree of self or situational awareness, they cannot be. But whether that view is correct, or even defensible, is entirely beside the point. What matters is that the output of AI systems can serve as a replacement for human creative input.
We can understand this point best by thinking about other examples of industrial inputs. There is already a long history of human ingenuity having supplied any number of synthetic substitutes for previously existing production inputs. Such substitutes may be functionally equivalent or even superior to the original input. In the case of AI inputs, once again, human ingenuity is on the verge of supplying an acceptable artificial or synthetic substitute for human creative contributions. Synthetic creativity in many cases may well be functionally equivalent or perhaps even superior to the industrial input it replaces. My concern here is not whether AI substitution is in some sense “genuine creativity,” but rather on the social and legal implications of making the substitution.
The question for intellectual property policy, then, is how the automation of creative labor changes the incentives that intellectual property is intended to address. History gives us some clues. We have a fairly good idea from existing scholarship and experience what the reaction to automating the creative process is likely to be. Social and cultural changes resulting from mass production from the first, mechanical industrial revolution are well documented. Thus, the history of past automated labor displacements teaches us something of what we can expect to occur as creativity is automated.
Past displacement of human craftsmanship has typically resulted in a kind of initial fascination with the novelty of the machine-made goods. But as the supply of such artifacts becomes routine and settled, society begins to place an increasing value on the fruits of human production, imperfect and inefficient as they may be. Indeed, the imperfections and rarity of the human-generated goods become an emblem of “authenticity,” while the affordability, accessibility, and ubiquity of the machine-generated versions become emblematic of the “inauthentic.” This shift to valuing the “hand-made” or “artisan” product re-imposes a type of scarcity on the abundance of creative substitutes, valuing what Walter Benjamin termed the “aura” of authenticity.
We might expect that this effect would have less traction in technical innovation, where the efficiency and efficacy of new products would seem the primary values to be effectuated. Consumers might be expected to invest greater meaning in the provenance of expressive or esthetic creations than in technical innovation. But we cannot entirely discount the tendency to value human connection, even in areas of technical achievement. Designations such as “German engineering” or “Yankee ingenuity” have long been indicative of an aura of superior technical quality arising from an ad hominem associations.
Situating the advent of synthetic creativity within the popular turn toward authenticity prompted by advancing automation places us in familiar territory. There already exists a rather large body of literature exploring the move to authenticity as a reaction to industrialization and to the displacement of human labor. The drive for authenticity has been recognized as a discursive social construct, fed by the need for distinction among the abundance of fungible consumptive objects generated in an industrial society. Consumers attempt to re-establish lost connections between places, communities, and objects by projecting certain ontological ideals onto commodities. Perhaps ironically, commodity sellers are often able to supply and capitalize on such ideals by promoting narratives of authenticity associated with their goods and services.
Thus, we have from past experience a good sense as to how the perception of authenticity plays out in a wide variety of situations, from politics to fashion. Authenticity has in particular become a key factor in the generation of Internet content, where the production of information goods is often divorced from any personal interaction with the consumers of such goods. The social turn toward authenticity has thus far been a reaction to industrial replication of fungible goods – to the falling marginal cost problem identified above. But we may expect the social reaction to be the same when cheap creativity rather than cheap reproduction becomes the center of attention.
IP and Authenticity
It therefore seems likely that proliferation of automated, synthetic creativity will drive demand for authentic, human creativity, and a core research question at the intersection of IP and AI must be understanding what role intellectual property law has in promoting or preserving such value. At a minimum, this change may be expected to shift the emphasis in IP away from patent and copyright law, which are conceived primarily as exclusive rights against appropriation, and which have relatively little to do with the question of authenticity. Patent and copyright are directed toward innovation, to the production of new creative works. Authenticity, in contrast, is directed away from innovation, toward an ideal of stability, measured by traditional or familiar templates. Authenticity values preservation of established, typical, known practice. Even when an “authentic” French chef develops a new dish, or an “authentic” gangsta rapper produces a new song, they are judged to be authentic for conforming to some prior set recognized conventions.
Admittedly, copyright has at times been dragooned into guaranteeing authentication rather than incentive, leveraging the statutory authority associated with the copyright author. But far more salient to the validation or certification of authenticity are the moral rights regimes often associated with copyright. Authorial moral rights are recognized by the majority of international jurisdictions, although they are largely absent from the federal copyright system in the United States. Typically, such rights include rights of attribution that ensure a creator’s name is associated with her work in the fashion she desires. Moral rights of integrity are also common, ensuring that the form of the work is not altered or amended against the wishes of the creator.
Trademark law is the form of intellectual property most readily associated with certification of authenticity, ensuring a valid link between products and a particular source. Thus, it is largely justified as a mechanism to mitigate consumer search costs and build supplier’s reputational capital. But trademark may also serve to create value by validating preferred associations with a product or service. Barton Beebe has cannily observed that modern trademark law acts sometimes like a contemporary sumptuary code, restricting the availability of luxury goods, creating an artificial scarcity that ensures that even where goods of identical quality could be provided on a mass basis, only the wealthy sport luxury brands. In the face of synthetic creativity, I suggest that trademark may come to support the inverse marketing position, identifying otherwise indistinguishable goods produced by human creativity rather than AI generativity.
Other parallel intellectual property regimes display related authentication functions. Rights of publicity serve to authenticate associations with prominent celebrities. Geographic indications, although ostensibly territorial, are on closer examination less associated with geography than they are with particular communities and “authentic” practices that may be related to geography. Traditional knowledge protections, too, serve to separate the authentic from the inauthentic, distinguishing artisan or craft goods from industrial replications.
Each of these modes of intellectual property operates as a form of certification rather than as a mode of incentive. Cheap, synthetic creativity will surely augment the importance of such certification. Patent and copyright incentives will not fall into desuetude; increased appropriability will remain a consideration. But as the marginal cost of initial creation falls, new emphasis will undoubtedly shift to intellectual property regimes that can certify human creative production. Patent and copyright have had the center stage in a world of cheap reproduction, but this may no longer be the case in a world of cheap creativity. Thus, a major goal of the future research agenda for IP scholarship must be coming to understand the role of the intellectual property and authenticity.
Dan L. Burk
Dan L. Burk, IP and Authenticity in the Age of Automated Creativity, ALTI Forum, July 4, 2022.
Invited by Thibault Schrepel