AI Use Might Equate to Stealing Intellectual Property, Implying Multiple Perpetrators
The United Kingdom's government is considering amendments to its copyright laws, a decision met with resistance from the artistic and creative sectors, particularly the Make It Fair campaign. The campaign claims that teaching artificial intelligence (AI) through copyrighted materials amounts to the covert, uncompensated appropriation of creative work. However, such an assertion raises questions about the ethics of learning itself, as AI, much like humans, uses existing knowledge to generate new insights. Instead of viewing AI as a form of theft, the UK should cultivate an accommodating copyright regime that fosters economic development and its aspirations to become an AI powerhouse.
Critics argue that AI training on copyrighted content without compensation amounts to stealing and demands licensing. They also contend that this practice strips creators of control and accentuates the broader mistreatment of content owners by AI companies. However, these concerns rest on the misunderstanding that human learning has historically been built on past creations without solicitation, payment, or attribution.
The current British copyright law does not prevent individuals from learning from established works to produce something new. It also carefully distinguishes between plagiarism and seeking inspiration, a boundary that encourages creativity, learning, and innovation. Just as artists, researchers, and musicians have always evolved upon prior work, creators have long profited from this adaptability. Notably, even Sir Paul McCartney, an active participant in the ongoing AI and copyright debate, borrowed Little Richard's vocal style for the hit song She Loves You, an act formally deemed permissible under UK copyright law.
AI learning operates on the same principle as human learning, albeit at a larger scale. AI systems that analyze large volumes of publicly available content to identify patterns and generate new, groundbreaking outputs represent an extension of this long-protected human tradition. The only replication that transpires is the formation of the training dataset, a mandate that could fall under the same exception that search engines use to provide search results. As it is often impossible for humans to trace how specific sources influence a final work, AI is similarly incapable of pinpointing what inspired a particular creation, especially considering that it synthesizes materials from multiple sources, thus diluting individual contributions.
Advancements in AI present greater benefits to creators than restrictions. For instance, when OpenAI introduced a new image generator, users inspired by Studio Ghibli's animation style started creating art in its likeness, raising concerns over originality. Nevertheless, Ghibli itself derived inspiration from the shin-hanga artistic movement, which was inspired by traditional ukiyo-e art and Western impressionism. Like Ghibli, shin-hanga thrived by reinventing past creativity in novel ways. Rather than devaluing Ghibli's work, AI breathed new life into its creative origins, fueling cultural engagement and even boosting the parent company's share value during the "Ghiblification" trend.
McCartney himself employed AI in the production of the documentary series The Beatles: Get Back and the hit single Now and Then, formerly known as "the last Beatles song," which sampled unfinished recordings from late band member John Lennon. The custom-made AI software isolated Lennon's vocals, allowing McCartney to continue working on Beatles music decades after Lennon's demise. The single reached number one on the UK Official Chart, marking the band's first time returning to the top spot in 54 years.
Even under a lenient copyright regime, creators remain protected in areas where copyright law has always protected them. For example, if someone were to utilize OpenAI's new tool to develop a film that closely mirrors the style and narrative of prior Studio Ghibli films, this would break the current UK copyright rules, as copyright protection continues to safeguard rightsholders against profiting from another creator's success. Training AI does not challenge this principle.
UK policymakers should resist calls to alter copyright law based on short-term, self-serving concerns from select creators. Instead, they should prioritize the broader public interest in fostering the UK's international competitiveness in AI. To achieve AI supremacy, the UK should refrain from erecting superfluous barriers to the utilization of publicly available data for training AI systems. The government should disregard the House of Lords' proposed copyright amendments to the Data Use and Access Bill and affirm that text and data mining is permitted under the UK's copyright law. Simultaneously, the government should continue to enforce existing copyright protections on AI-generated outputs.
The UK government faces a straightforward decision. Employing publicly available data for AI training is not a theft, but a vital component of innovation. If the UK aims to carve out a global leadership role in AI, it should develop a progressive copyright framework that nurtures, rather than inhibits, AI development. Initiating this change involves adopting a modern, tolerant approach to copyright that protects creative rights without hindering the UK's long-term technological prospects.
- Critics argue that using copyrighted content for AI training without compensation may be perceived as stealing and demand licensing, suggesting that this practice strips creators of control.
- Notably, even Sir Paul McCartney, a prominent figure in the AI-copyright debate, has employed AI in his work, such as the creation of the hit single "Now and Then."
- AI learning, while on a larger scale than human learning, operates on the same principle, analyzing large volumes of publicly available content to generate new outputs.
- If the UK wishes to establish itself as a global leader in AI, it should develop a copyright framework that fosters innovation, rather than imposing superfluous barriers on the use of publicly available data for AI training.
