Insights

AI’s Learning Curve Clashes With Proprietary Data: Processing Originality Raises Legal, Ethical Questions

Data Privacy, Financial

By: Rebecca L. Rakoski, Esquire
       Patrick D. Isbill, Esquire

Originality is often talked about but difficult to attain. An equation that adds effort and time with patience attempting to battle any final calculation. The marketing of new artificial intelligence (AI) learning technologies promises a spectacular increase in productivity, challenging the limiting factors of effort and time, but perhaps also with a cost to human creativity and innovation. News reports have again highlighted how AI is producing art by using algorithms to study, i.e., arguably copy to some legal analysts, the long-practiced techniques of thousands of artists in order to then generate art of its “own.” Fundamentally, AI is racing to process all of this individual creative work, and yet using this type of technology could also put it on a fast-track collision course with copyright infringement and intellectual property laws.

These large reams of data, whether it be the styles of different artists, the unique voices of writers, the complex nuances of computer code, or even the scientific reasoning of researchers, are being processed at a stunning rate to essentially achieve what could conceivably be termed synthesized productivity and ultimately one might argue synthesized creativity. No longer a wholly human construct but rather manufactured and then authenticated by a machine, or at least a fenced in or potentially undisciplined notion of what that means to the artificial intelligence. In short, synthesizing data to deliberately produce a result may someday altogether alter the definition of what is now understood to be authentic and previously uncharted human innovation.

Digital Life Imitates Art

Largely understood as a term introduced in the mid-nineties and credited to an article in the Harvard Business Review, “disruptive innovation” is typically an all-too-familiar term or theory used to generally describe innovation that initially creates a new market to displace established ones. It is likewise more conceptually complex than the nonchalance it can commonly be afforded. One that is, at times, thoughtlessly attached to AI or other emerging technologies but frequently casts an unfavorable light on what might be described later as arguably unforeseen or undesirable results. So whereas disruptive AI technologies are narrow in focus and can be categorized as opportunistic, progressive leaps are instead ones incorporating what already works with what can be made better. When used without forethought, disruptive AI thinking can also suggest an uneven direction and also unpredictably when it comes to strategy.

Whether steady or unbounded, progressive technologies imply a degree of considered thought at least toward ideals like responsibility and examination. In other words, efficiently improving and building on what is already there, acknowledging those steps may indeed be immense, rather than recreating the wheel for its own sake. Even so, AI is undeniably outstanding for what it can process and then analyze, but should there be greater caution when allowing it to produce what it will subsequently replace/displace or more concerning what it will disrupt?

Part of the legal argument for AI violating intellectual property rights or copyright infringement should include its unfair advantage over its human counterpart when it comes to processing times. It is simply not a one-for-one comparison. For AI, its abilities to learn are markedly streamlined, using large reams of proprietary data from digital sources like the internet to formulate its “own” style without accreditation or at a minimum acknowledgment to the originating sources is an obvious distinction. This ability to rapidly process data and then create, or in some instances re-create, material will only accelerate with the inevitable addition of quantum computing. All achieved in an instant of course relative to what would certainly be multiple lifetimes for human beings. It is difficult to comprehend that decades, or even centuries, of learning and the evolution of common thought may well be reduced to less than a microsecond. While admittedly the possibilities are exciting and should at least be explored, it must also be conceded that the results will be anything but predictable.

This immediately raises several important questions. One being will this meteoric ability to process these large reams of proprietary data then lead to innovation, or an entirely original idea from an intelligence that is artificial? The ethical questions begin to mount too when considering that AI now needs data from external sources to be foundationally innovative or express creativity. This “help yourself digital data buffet” has already been heavily scrutinized when it comes to data privacy in law, confidential personal health information in medicine, trademark and copyright infringement in technology, and consumer data collections for business analytics. But what happens when human creativity is elbowed out by an artificial one? Will AI then launch new creative ideas from the data it generated? And how can such a product be original when the data used, or arguably copied, to create it was proprietary? Lots of questions with few answers so far other than originality are arguably being co-opted under the auspices of artificial learning without adequate recognition of origination and/or compensation at the moment.

Litigant AI

The other side of the legal argument is whether AI should be afforded copyright protection for its generation of what some argue is its original content. Needless to say, this argument triggers a corresponding question of whether AI can “own” or hold a copyright? It is not an easy answer for all the reasons previously outlined, but thus far courts have been reluctant to budge on the finding that AI is not entitled to such protection. Furthermore, the findings for now state that works must have human creators to qualify as copyrightable while leaving just enough room though for additional argument in the future on the issue of whether the content was created with sufficient human creative input. This might be interpreted as a possible signal for a corresponding recognition of the human origin of the data behind AI processing, as well as the need to assign discernible proprietary rights.

Based on the flood of recent reports involving AI technologies like ChatGPT, it has reignited the imagination of innovators and thought leaders across almost every industry around the globe. Much of these deliberations however have frankly been turning gears for some time now in areas like the law, e.g., when it comes to brief writing and case research to conceivably streamline the grind of litigation. In addition, AI should be considered for the reverberations it will create in cybersecurity, and the effect it could have on areas like professional and products liability. For example, AI’s proposed use in medicine is well documented. Assisting in diagnosis or aiding in surgery have been talked about concepts for a while. But what about implications for informed patient consent or the redefinition of tort law liability related to professional services?

Another pressing question along those same lines involves products liability and the potential corruption of AI through deficient cybersecurity controls leading to unintended production which is unknown, or could have been known, to the distributor. There is a temptation to suggest that the AI itself could be a party litigant to a cause of action, but an equally strong resistance to consider what such a representation would look like for its legal counsel. For healthcare providers, there must be a distinction for liability purposes, i.e., AI would be relied upon to essentially synthesize a diagnosis whether partial or in full, as to what degree a licensed doctor is directing a diagnostic course and to what corresponding degree AI is weighing in and then followed.

Conclusion

Progress would seem to be the tortoise to disruption’s hare by today’s logical metric when it comes to the integration of artificial intelligence. AI should work with what has been done in order to make possible what can be done and ultimately aid in expanding, rather than replacing, the limits of human innovation to realize what will be done. AI after all is only as creative as the data it is initially given, often with flagrant proprietary legal rights attached to that data. The results are therefore unequivocally an artificial construct, notwithstanding the acceleration of process or the worthwhile result.

Reasoning and inspiration are a human construct that inexplicably must have the freedom to shift arbitrarily to find what to us now may seem an illogical result but later is identified as a fitted solution. Artificial algorithms have its advantages to be sure, but will unlikely be able to calculate what is on the surface unexplainably impractical but equally right minded. Legal guardrails and ethical discussions need to be ongoing and flexible to meet this moment in part for the sake of equity to those understandably wary innovators involuntarily providing the data AI needs today to perform with the synthesized, albeit calculated, creativity it promises for tomorrow. Promising to produce innovative and original material from a vast digital world of what should already be recognized as original content with intellectual property rights without rendering at the same time its human contributor/author obsolete for failing to match either its speed or measurable intelligence should be given serious mindful thought.

All in all, it is said that good things come to those who wait. Progress implies that the considered path will in the end deliver the better result over what may instead be hastily produced or erroneously designated as disruption and thereafter the production of an undesired result, or one not easily remedied. Most innovators in business and technology will agree that speed and efficiency are not synonymous terms and in fact are two dissimilar concepts. Discussions surrounding the application of AI across multiple industries are definitely not a new construct. Legal analysts have been talking about the potential and ethical pitfalls of AI for years, but like any discovery, the potential inches closer to reality over time, and still it would seem another sizable step has been taken. We are more than capable of meeting the moment, much like we have in the past, and to progress to an approach that favors constructive use of AI to responsibly evolve established methodologies rather than impetuously trying to force disruption for its “own” sake.

Reprinted with permission from the March 14, 2023, issue of the New York Law Journal. Further duplication without permission is prohibited. All rights reserved. © 2023 ALM Media Properties, LLC.

 This article does not constitute legal advice or create an attorney-client relationship. The information provided herein may not be applicable in all situations and should not be acted upon without specific legal advice based on particular situations.

Data Privacy, Financial

Categories

Follow XPAN Law Partners