This post is prompted by a new very provocative and ‘spicy’ article written by Benjamin Sobel for the Chicago-Kent Law Review titled “Copyright Accelerationism” (a brief summary). This paper, if I could summarise in a line or two (to the extent of being very reductionist), states that Copyright maximalism is not that bad. That it should be applied across the board to all, machines and humans equally, without fear or favour nor ill will. That if applied to its full extent, without mercy to all, it is one possible way to achieve meaningful reform of Copyright law, as it will leave no other choice.
In this post, I will explain the core argument of the paper and then discuss my own thoughts about being an Indian copyright Accelerationalist.
Copyright Accelerationism, aka taking your precepts very seriously:
Copyright law, at least as it exists in the US, has created a two-track system for machine and human learning. Where machine reading/learning has been treated as fair use, or arguments are advanced to that effect, both in the US and India (see here and here). On the other hand, human learning is subject to the strictures of copyright law ostensibly to preserve a captive market for the copyright owners to extract rents and licensing fees (for example, see here and here). Hence, “romantic” readers who critically read, understand, reflect and write follow-on works are caught in the net of copyright law, captive users who have to pay a price to access copyrighted works, don’t get any extra privileges, while mindless machines swallow entire texts with no worry in the world.
In light of this dichotomy, and the fact that human readers do not have the privileges that are extended to machines in some instances (for example, learning from shadow libraries), he suggests that Copyright law should not distinguish between human and machine learning. It should treat both of them equally and the strictures of Copyright law should apply in equal vein to both forms of learning.
This suggestion sets the stage for the idea of Accelerationism or more specifically, Copyright accelerationism:
Copyright accelerationism demands that we insist on the rigid application of contemporary copyright doctrine and thereby heighten its contradictions. Doing so will achieve two things. First, it will make it undeniable that the copyright system is unworkable in the present day. Second, it will actually make the copyright system unworkable and, in turn, catalyze systemic reforms that are long overdue.
Thus, the point is that whenever contradictions emerge within copyright law (for example, when fundamental precepts of copyright law come to blows with a new technology), do not carve out exceptions to manage the contradiction; instead, fully embrace it, apply it in full with no exceptions. Thus, by taking forward the logic of copyright precepts to the fullest extent, one might create conditions for meaningful reform, not settle for just tinkering in the margins. Copyright accelerationism, as Sobel clarifies, is still legalism and operates within the broader legal system, which is relatively stable. Its key insight is to take the percepts of Copyright law seriously and apply it to machines, i.e GenAI; if possible, go even further. This position must be contrasted with the copyright incremental-minimalists, who are the inverse, who seek to enforce exceptions, expand their scope, and fight against attempts at increasing the scope of copyright. For example, my article in the India Forum on the recent ANI v. YouTubers controversy is a classic incremental-minimalist response to an attempt by some commentators to make fair dealing defence the end-all and be-all of copyright defence. Sobel urges us to think otherwise, at least in the current GenAI age. With GenAI grappling with Copyright law (debate and litigation have also reached the Indian shores), Sobel argues that there is reason to embrace the accelerationist position, by implementing the precepts of copyright fully and completely to GenAI, one can pit the unstoppable force against the unmovable object, thus setting the stage for a un-classic Goliath vs Goliath showdown.
Hence, a copyright accelerationist’s classic response to those seeking reform of copyright law to carve out an exception for Gen AI trained on copyrighted content is the following: “Sounds like you might’ve made a bad bet. What would you be willing to do to salvage it?” thus forcing the GenAI lobby to reform the Copyright system for us all and not just for machines.
Thoughts on an “Indian” Copyright Accelerationalist:
It is very tempting to me at this juncture to embrace Sobel’s thesis and choose the Copyright Accelerationalist pony. With news of Sci-Hub being blocked and the general absence of provisions for further facilitating human learning (here, here, here and here), one is tempted to say, This is not enough. Maybe we should go ahead and ban all other shadow libraries, make it harder for people to use technological measures to circumvent the ban. Argue that the new AI committee should demand that the strictures of a right-holder-centric Indian copyright law should be applied in toto to Gen AI with no exceptions. Why should machines have the benefits of exceptions that humans don’t have? There should be no specific Section 52 exceptions crafted in for GenAI. Let us fully embrace the Copyright law’s fundamental precepts and hope for meaningful change when it becomes too unmanageable.
Sobel also forces us to think about another important aspect: why should we privilege machine readers over human readers? And adorn it with benefits human reading does not enjoy? What is so special about it? Is it because they have a powerful interest group and lobbies walking the corridors of power, arguing for their interests? Is it for facilitating technological progress? But I am unsure as to how blocking Sci-Hub can enable technological progress. Needless to say, Sci-Hub is a low-cost and affordable way for Indian researchers (see here and here) to access scholarly publications. It is in line with the theory that shadow libraries are a symptom of access inequality sanctioned by law. When the traditional copyright system fails to ensure access, shadow libraries pop up, bridging the “access gap”. While options like an opt-out regime are being pursued for machine reading/learning, why have similar discussions that can enable human learning not been discussed? For example, taking public deposit requirements seriously.
In the end, Sobel’s gambit relies on a powerful AI lobby that can change Copyright law for us all. But in the Indian context, one is sceptical of how foreign tech companies will initiate changes that are beneficial for Indian readers. One might still want to be an accelerationist and hope that socio-political and student movements might emerge in response to such contradictions within Copyright law. Maybe the online outcry over the Sci-Hub ban might translate into something meaningful. But again, there is no guarantee it will emerge, nor does it make reason to think they can bring forth meaningful change, even though they might have done so in the past. In the end, one might still be struck with a copyright maximalist regime with no meaningful change. Hence, the light at the end of the tunnel is not sunlight but the headlight of an incoming train. But Sobel’s paper is creative, novel and forces questions that we all ought to consider – if we take copyright as seriously as we pretend to, why aren’t we doing what that requires: perhaps there’s a valid reason not to and perhaps we should focus more on those reasons.
I wish to thank Swaraj for his valuable input !