Authors and AI principles
ALCS has developed a set of principles which set out our position on technologies such as AI in terms that are broadly applicable and understandable. We have and will continue to engage on this issue in detail, specifically on various technologies and uses, as we have in other relevant position papers and consultations. These principles are intended to be general and applicable to various technological developments, as well as comprehensible to an audience unfamiliar with AI. We state some of the most prominent concerns identifiable now, but the principles are not limited to current challenges.
We want to ensure that discussion and policy-making around AI is not swept up in the often fantastical thinking that accompanies such ‘technological advance’ but considers instead its limitations and the potential impact on the irreplaceable value which creators, such as writers, contribute overall. We do not want to obstruct or decry technological innovation. Rather, we want to ensure that the future is one where human creators work with new technologies and where their unique contribution is properly recognised.
To achieve this, ALCS believes that the following principles must be integrated into Intellectual Property (IP) policy development:
1. Human authors should be compensated for their work and have transparent information of uses of their work – new technologies and types of uses should respect the established precedent that there must be fair payment for use and transparency regarding how works have been used.
There should be no use without payment. While many systems of remuneration reflect the ways in which we consume media at the time, the principles of copyright remain: that authors should share in the enjoyment of their work, especially financially. Licensing continues to be an effective way to adapt to, support and remunerate a wide range of uses of original works. As authors works are often used in ways to develop AI technology, authors must also have transparent information regarding when and how their work has been used.
2. Technological developments in machine learning should not be seized upon as a circumvention of creators’ rights and AI generated work should be labelled as such
The role of technology as a tool of creators must be acknowledged, as it is both reliant on and complementary to human work and human authorship. As with other media and technological developments, consideration must be given to the rights of human creators. The technological tools with which an author interacts should not change their fundamental, time-honoured rights. Additionally, where a work has been generated by AI rather than written by a human author, this must be made clear in any publishing and selling of such a work.
3. Licensing terms should be carefully considered, not just for the ‘first use’, but also for times when an AI program would effectively reuse a work repeatedly
Authors’ work can be used to programme technology, which would then create derivative works, such as an AI program reading a selection of novels before repeatedly creating similar books. In this sense, an author’s work is used multiple times or in different ways; in analysis and subsequently in the creation of other derivative works. This means that consideration of authors’ rights should not end when an AI program initially analyses their work, as it effectively uses it on an ongoing basis to create additional derivative works. This must be a consideration in the terms of licensing the ‘first use’ to programme an AI system.
4. Stakeholders must fully acknowledge the limitations of AI in a wider social context.
While the capabilities of artificial intelligence are impressive, they may also reproduce, reinforce, or exacerbate societal biases. Being reliant on existing media, it has been widely found that AI-created content without human perspective may reflect harmful stereotypes and repeat false information. The extent and nature of these biases have yet to be fully documented, and this must be considered both in their application and in policymaking regarding their use. It has been acknowledged that AI requires ongoing human oversight, scrutiny, and expertise for ethical development. AI cannot self-regulate, and it often cannot apply context and nuance to information in the same ways in which humans can.
5. Policy-makers and industry should responsibly pursue uses of AI that support creators’ contributions
The approach to developing the opportunities AI presents should not diminish or side-line, the social, cultural, and economic value that creators contribute. This would have a devastating impact and would cause significant harm to the creative community upon which our creative industries will continue to rely.
Artificial Intelligence (AI) is increasingly discussed as a way in which authors’ works could be further exploited. To achieve a positive future where human authors engage with AI to create new works, some key principles must be considered.
AI is reliant on a huge amount of work created by human authors. It should not be treated as something independent of, or a substitute for, an author’s original work, insight, and creativity. Whether use of that work comes from human programmers of the software, or where author-created content is used to “teach,” an AI program knows nothing for itself and only reuses authors’ work. The original work is essential, and authors must be acknowledged for their creations, as should all writers, artists, composers, and producers. AI is wholly reliant on their output, understandings, judgments, observations, and other data on which it draws.
We want to see a future where creators can make the most of AI, while avoiding significant harm from a misguided approach to its exploitation or missing the opportunities that it can provide. Human creativity should be remunerated as part for its use by AI. We want to be a world leader in the human and machine learning systems that such collaborations can create, but we also need to reward, and continue to invest in, the uniqueness of British authors and their creativity.
Downloads
ALCS AI Principles | Download |