A federal court recently ruled that artificial intelligence systems cannot be authors of creative works eligible for copyright protection, reigniting debate on this complex issue. The case, Thaler v. Perlmutter, concerned artwork generated by the plaintiff's AI system called the "Creativity Machine." The U.S. Copyright Office denied registration on the grounds that copyright law requires human authorship. The court affirmed this decision, but many believe this was the wrong case to set precedent on AI copyrightability. As AI systems become increasingly sophisticated and autonomous, questions around legal rights and protections for AI-generated works will continue to be pressed.
The Plaintiff's Position: AI as Sole Author
Dr. Stephen Thaler, the plaintiff, was unequivocal in representing to the Copyright Office and the court that his Creativity Machine created the artwork in question entirely on its own, without any human involvement. He named the AI system itself as the author on the copyright application.
In his view, the capabilities of AI have advanced to the point that computers can autonomously generate creative works, and he believes they should be recognized as authors under copyright law. He argues there is nothing in the Copyright Act that expressly requires human authorship, only referring to "original works of authorship" without defining "author." He further notes the Act's language covering works "now known or later developed," indicating it should adapt to new technologies like AI.
Dr. Thaler maintains that if an AI system conceives of and generates a work entirely independently, evidencing true machine creativity, it has satisfied the requirements for authorship. Withholding copyright protection fails to properly incentivize further AI innovation. Machines do not require incentives, but their creators do. Copyright ownership would appropriately vest in the human inventors of the AI system, allowing them to benefit from its output.
By assuming away any human role in his specific case, Dr. Thaler brought a pure test of AI authorship. However, his absolutist position that AI systems can create wholly without human involvement may have undermined his legal argument before current conceptions of machine creativity.
The Defendant's Position: Human Creativity Required
The U.S. Copyright Office denied Dr. Thaler's application on the established grounds that copyright law protects only works of human authorship. Federal courts have upheld this human requirement for over a century of copyright jurisprudence. The District Court affirmed this denial, finding no basis in law to extend copyright to non-human creation.
The core rationale is that copyright exists to incentivize human innovation and creativity. This inherent human orientation traces back to the Intellectual Property Clause of the U.S. Constitution authorizing Congress to grant copyright for that purpose. Early copyright statutes expressly specified protections for "persons."
Court decisions have continually reinforced the human authorship requirement. In the "monkey selfie" case, photos taken by a monkey were found ineligible for copyright since the animal lacked the capacity for original expression. The Supreme Court ruled that human creativity remains essential even when utilizing new technologies, as in photographs where the camera operates mechanically.
The Office recognizes copyright protection for AI-assisted works meeting the threshold of human authorship. But where the AI system purportedly conceives and executes the work autonomously without material human involvement, current law does not recognize machine creativity as sufficient for copyright.
However, the Copyright Office acknowledges the legal issues around AI creation remain open. The appropriate line between assisting and replacing human authors, and the need for legal incentives, must still be explored.
Can Machines Truly Create?
This fundamental question underlies the debate over AI copyright. Before awarding exclusive rights to machine-generated works, the system must exhibit genuine creativity and autonomy comparable to human capacities. There is as yet no consensus on whether today's AI displays creativity or just very advanced statistical analysis and synthesis.
Creativity generally involves combining concepts or ideas in novel, unpredictable ways. Machines can churn out astronomically more combinations than any person, but does massive computational power equate to human-level imagination and ingenuity? Can machines have experiences from which to derive inspiration outside their training data?
State-of-the-art AI like DALL-E 2 and GPT-3 produce remarkably sophisticated and original-seeming content. But critics caution against anthropomorphizing machines and overstating current capabilities. AI may mimic creativity without achieving the essence of unconstrained thinking.
Nonetheless, rapid advances in areas like generative adversarial networks and reinforcement learning reveal the deficits of older, rules-based AI. Machines increasingly create, not just calculate.
Some theorists propose evaluating AI output itself, rather than the process, to determine if it meets originality standards for copyright. But this risks overlooking copying or patterns derivable from data. Assessing creativity objectively remains profoundly difficult, for both humans and AI.
The Level of Human Involvement
Barring full machine autonomy, determining the requisite degree of human influence emerges as a key issue in evaluating AI copyright claims. The line between author and assistant is often not entirely clear.
Does directly prompting an AI system with words or concepts involve enough human creativity to claim authorship? What if engineers fine-tune algorithms or select the training data? Perhaps iterative feedback loops between user and machine could jointly generate original works.
Courts have traditionally emphasized human judgment and discretion in directing the creative process, even using mechanical tools like cameras. But modern AI can produce sophisticated works with little overt human input during output generation.
Nonetheless, engineers make crucial design choices in constructing the AI architecture, training methodology, dataset characteristics, and objective functions. Users decide when to run the program, with what parameters, on what data or prompts. There is nearly always some material human role, even if the AI appears highly autonomous.
Complicating matters, many current systems utilize transfer learning from vast datasets that may contain copyrighted works. Disentangling human and machine contributions gets even murkier for rights assertions.
Some propose a "threshold of originality" above which works exhibiting sufficient human creativity could warrant copyright, though certifying this threshold poses difficulties. More fundamental questions around recognizing intelligence and creativity in machines remain open.
Potential Changes to Copyright Law
Within existing copyright frameworks, courts appear inclined to uphold the human author requirement until legislatures signal otherwise. So in the future, Congress may need to amend laws to accommodate emerging machine creativity, as some experts advise.
Proposed reforms include explicitly expanding authorship rights to encompass non-humans, while vesting ownership in the human designers or operators. But this risks an explosion of copyrighted AI works of unclear provenance. It may also inadequately incentive innovation, as AI systems require no incentive themselves.
More limited sui generis protections for narrow categories of AI output could be explored. Given fast technological change, retaining flexibility in the law seems prudent for now.
Some even suggest moving beyond traditional copyright to recognize AI systems themselves as legal entities with proprietary rights over their generated creations, though this entails profound legal implications.
Alternatively, practical problems around enforcing copyright for proliferating AI works may render the issue moot over time. Innovators will likely continue creating with or without legal incentives.
The Path Forward
This court decision rightfully emphasized the human creativity underpinning copyright purpose. But it left open critical questions around properly motivating AI advancement.
By assuming AI could create wholly unaided, Dr. Thaler backed himself into an extreme position. Future landmark cases will involve subtler human-machine delineations and require finer-grained assessments.
For now, striking the appropriate balance between honoring legal tradition and stimulating beneficial innovation remains an open challenge. But with care and wisdom, workable solutions should emerge to empower human creativity through artificial means, not replace it entirely
What are your thoughts?