Standing on the shoulders of giants: the conflict between creatives and AI

written by Lee Curtis, Partner and Chartered Trade Mark Attorney – HGF

It is reported that in a letter to Robert Hooke in 1675, Sir Isaac Newton stated, “If I have seen further it is by standing on the shoulders of giants”. Now it is likely that Sir Issac Newton was not the first to have uttered a form of that saying. Forms of the saying date back to the 12^th century, but what he was illustrating that advancements in human knowledge and crucially the art of learning are based on the previous discoveries of others and that one learns to degree from the previous efforts of others.

To a degree all human students learn by copying or at least reading and absorbing the creative works of others. However, with this comes a conflict between the learning process and creative works of others. At its basic core intellectual property rights are there to prevent copying. However, intellectual property law has accepted the conflict behind ‘learning’ and IP rights and provided carve outs, especially with regards copyright laws for private study and indeed non-commercial research. For example, under UK Copyright Law a student can copy books for private study and indeed there are text and data mining carve outs for “computational analysis” which are for the “sole purpose of research for a non-commercial purpose”. However, what happens when it is an artificial intelligence application which has to learn?

AI Applications learn via the input of millions, if not billions, of pieces of data and images. For example the diffusion model of artificial intelligence learns via diffusing images, or in other words breaking them up, into pieces and then reconstituting them and via that process learns what certain images respects. However, what if those images are subject of copyright rights?

Matthew Sag who is a Professor of Law in Artificial Intelligence Machine Learning and Data Science at Emory University Law School in Atlanta has framed this issue as the so-called “Snoopy Problem”. Ask an AI application to create a cartoon dog and it has a tendency to create an image which is very similar to Snoopy, which is subject of copyright, as it tends to be the image most widely available on the internet of a cartoon dog. Many AI applications having ‘learnt’ from data supplied to them via scraping the internet. Others have referred to this problem as ‘The Italian Plumber’ problem, and I will leave you to guess which copyrighted character that refers to, but it does relate to a computer game!

AI applications pose further complications for IP law in that in the main they are used for commercial purposes, not simply for non-commercial research. Now creatives rightly contend why should AI application providers benefit from their creative efforts to enable their AI applications to learn and provide commercial applications which could compete or replace them or more simply benefit from their creative efforts without recompose.

However, in its written submissions to the House of Lords Communications and Digital Select Committee inquiry on Large language models, OpenAI, the creators of ChatGPT stated on the 5^th December 2023:

“Because copyright today covers virtually every sort of human expression – including blog posts, photographs, forum posts, scraps of software code, and government documents–it would be impossible to train today’s leading AI models without using copyrighted materials. Limiting training data to public domain books and drawings created more than a century ago might yield an interesting experiment, but would not provide AI systems that meet the needs of today’s citizens.”

In short, if an AI application were limited to simply learning from materials which were out of copyright, noting that copyright can last up to seventy years from the authors death, then it might produce very outdated responses. What if the impressionist artists of the nineteenth century had been restricted from learning from medieval art or what if the Beetles had not heard the music of Elvis Presley? Sgt Pepper’s Lonely Hearts Club Band or Strawberry Fields Forever might not have been penned.

So how can we provide a resolution to this conflict between how AI applications learn and the legitimate rights of creatives? There are presently multiple disputes between the courts around the world between AI application providers and creatives concerning the issue of AI learning, notably the case presently before the English High Court between Getty Images and Stability AI, which on present timings is unlikely to be decided before 2025. So creatives have already resorted to the courts. However, are there solutions for this conflict from government or indeed via commercial tools?

It is likely to be near impossible to resolve the conflicts to satisfaction of both sides. The previous government went through various proposals for resolution, all of which failed.  At first there were proposals for a general carve out from copyright law to enable AI Application to copy for learning purposes. This proposal obviously encountered significant resistance from the creative sector and was shelved. The Government then attempted to devise an ‘AI Code of Conduct’ via the United Kingdom Intellectual Property Office, but discussions between representatives of the creative and AI sectors failed to find a resolution.

In May 2024, Viscount Camrose, the then Minister of IP, indicated that some modifications to existing copyright legislation was likely, that international standards would have to be agreed and that technology might form part of the solution. On the later score, the researchers at the University of Chicago did at the end of 2023 release an application called Nightshade which potentially ‘poisons’ AI applications which scrapping websites for information. Also a number of IP rights owners, notably Sony Music, have attempted to ‘opt out’ of their materials of being used by AI application. Finally, some IP rights owners such as the Financial Times and Axel Springer have licenced their data to be used via licensing deals.

So what is likely to happen? Well it would seem that in the words of Sir Issac Newton it is likely that creatives may have to allow some standing on their shoulders, but they should be compensated for the privilege.