Protecting Human Creativity from AI’s Grip

Every big AI platform has faced lawsuits for using existing creative works to train their models without permission or compensation, with plaintiffs ranging from Hollywood studios to media outlets to individual artists and authors. For Professor Haochen Sun, Founding Director of HKU’s Programme on Artificial Intelligence and the Law and an expert in technology law and policy, this use of creative works is a legal and ethical problem for both creators and society.

“These works are the foundation of our human civilisation. They embody human thought, from philosophy to literature, music and art, and they should be highly valued. When they’re used in the AI training process, it demeans their value by transforming them into tokens. What would be the implications, then, for the future of human creativity?” he asked.

Using content without permission is problematic on several fronts, he said. First, whoever controls the inputs to the AI training process can shape how AI develops and makes decisions, so the creators who provide that input could be excluded.

Second, there is the possibility that copyright protection could extend to AI-generated material, knowingly or not. While most jurisdictions do not allow this (apart from one judge in the Chinese Mainland who decided that an AI prompt was creative input), Professor Sun believes there is little to stop humans from presenting AI content as their own.

“It’s so easy to conceal this content and pretend it was made by a human,” he said. “I can foresee that an enormous amount of AI-generated content, ranging from text answers to images and videos, could end up being protected by copyright law because the threshold as the law stands – called the ‘originality requirement’ – is extremely low.”

In awe of AI

In such a scenario, ‘AI copyright trolls’ may be incited to launch lawsuits to see what they could extract. Most importantly, content flows to the public domain would be restricted, ultimately harming human creativity.

“When we human beings create new works, we learn from public domain materials and even use them in the process of creation. Borrowing and learning from others’ works and ideas are essential to human creativity,” he said.

AI platforms disrupt that process not only by breaching copyright. They also diminish existing works through their tokenisation in the training process, which suggests the companies do not take human creative content seriously. At the same time, AI-generated content is becoming more powerful on the back of these works. “People have become in awe of AI creativity instead of our own creativity,” he said, despite the greater richness of human creativity.

Professor Sun has identified three reasons why human creativity is unique and merits enhanced protection: humans create new works with intention, while AI responds to prompts; human works represent the embodiment of emotions along aesthetic, intellectual and moral dimensions, while AI does not have these multidimensional experiences; and humans can control the creative process, while AI output is controlled by its algorithms.

“The uniqueness of human creativity is why we need to protect and promote it through more robust regulation of AI systems,” he said.

Reining it in

Professor Sun has made a start on that challenge through three projects that apply legal and ethical principles to nurture human creativity and rein in AI.

One project examines the legal status of AI-generated content under copyright law. He proposes that copyright law should not protect works generated by AI systems and instead, such works should go into the public domain. At the same time, he suggests AI-affiliated content that has been substantially altered by humans to create something new could receive copyright protection.

A second project is about the transparency of AI outputs. Professor Sun identified fundamental differences between the creative capabilities of humans and AI systems by applying relevant philosophical theories, which led him to propose regulating the output transparency of AI-generated content by attaching watermarks, to make them more easily differentiated from human-created works.

The third project aims to elevate the value of highly creative human works and distinguish them from AI by promoting the concept of ‘super creativity’. He proposes major financial prizes as well as legal mechanisms requiring authors or their estates to register works 20 years after they are published.

“I see a tendency for more and more people to outsource their thinking capacity and creative activities to AI systems, and for AI-generated content to be competing with humans. If we don’t do anything to stop this, humans will lose out,” he said, especially given AI’s capacity to generate vast amounts of content.

“We must make an effort to subvert that process, to encourage people to use AI only as a tool, and to emphasise human learning. We cannot simply surrender ourselves to AI.”