21-02-2024

Participant of the "Kurk Lietuvai" programme: I do not see the threat of AI to the creator

“There is a tension between creators and artificial intelligence companies. OpenAI claims that it is impossible to create artificial intelligence tools such as ChatGPT without copyright-protected material. Thus, why is there such a reluctance to remunerate creators for their creative contribution to successful development of AI technologies?”, Agnė Minkštienė, participant of the Kurk Lietuvai programme, asks rhetorically. At the Ministry of Culture, she implemented  the project “Artificial intelligence and problematics of creators – what is the role of Ministry of Culture?“. This topic will also receive attention at Vilnius Book Fair, where Minkštienė will attend the discussion “Do Androids Dream of Electric Sheep?” at 2 p.m. on Friday.

Why did you choose to analyse the issues of artificial intelligence and the creator in your project? What determined your interest in this topic?

The interest in artificial intelligence (AI) arose very naturally. When ChatGPT became popular in November of 2022, I also decided to try it. I was immediately fascinated by how this software works. I felt very quickly that it is a technology that will change how we work, how we plan our leisure time. I started using it gradually in my work, as well: AI wrote great abstracts of articles that were required and created formulas for the Excel software. It did all of that very quickly and precisely. Since I was working in the English language, the possibilities seemed truly staggering. I heard in seminars how AI makes various tasks at work more efficient, how the routine, lower paid jobs are being reduced and there is a rising demand for specialists who are capable of working with AI tools.

Since I’m a Lithuanian, I became very interested how Lithuania is preparing for this technological transformation. Thus, as I was coming to the selection of the Kurk Lietuvai programme, I was already carrying the topic of AI with me. I wanted to work with it very much. The potential title of the project, “Artificial intelligence and problematics of creators – what is the role of Ministry of Culture?”, seemed very philosophical. I’ve earned a degree in philosophy, and philosophy was also close to me, so I decided to take on this challenge without even slightest idea of where it can lead me.

Which aspects of this topic seemed the most problematic to you?

After conducting an analysis of the current situation and communicating with many wonderful people from various fields (AGATA, AVAKA, LATGA, Lithuanian Audiosensory Library, Ministry of Culture, Lithuanian Game Developers Association, Martynas Mažvydas National Library of Lithuania, the art and culture magazine Nemunas, VMU etc.) and after analysing various sources, three main areas of problematic points emerged: philosophical (can AI be considered a creator, can it be called intelligence, how much freedom does AI have while creating new content etc.), technical (how to select, how to check, how to teach etc.), and legal (the consumer’s right to know, assignment of authorship, indication of the used sources etc.).

I realised that the issue of payment for the use of the creators’ content in AI systems has hardly been solved in the world. It is a huge problematic point. There are various court processes going on in the world because the creators feel underappreciated, exploited, and left on the margins.

The creators have supposedly been allowed the opportunity to “reserve the rights” to their work by indicating it clearly “using appropriate means”, in other words, the so-called “opt-out” mechanism. It should ensure that the creators’ content won’t end up in AI learning collections if the creator does not want it. However, it is not completely clear what those “appropriate means” are. The authors’ work may be protected in the breadths of the internet by other persons as well, thus, even if one copy will indeed be marked “using appropriate means”, another one might be open. It is also unclear whether different markings are equally understood and honoured. There are no mechanisms that regulate the upholding of this provision. Basically, removing a work from the AI system, when it is already in it, is a complex process which requires “re-teaching” of the AI model, which is very expensive and inefficient.

Interestingly, the Government of Poland, which was late to transpose into its law the Directive (EU) 2019/790 of the European Parliament and of the Council of 17 April 2019 on copyright and related rights in the Digital Single Market, now says that it had additional time to appropriately evaluate it taking into consideration the context of generative AI. According to this country’s Government, the exclusion of text and data cannot be applied on the use of copyright-protected content in AI training. The copyright issue has become quite intriguing and full of unexpected turns.

Perhaps the creators themselves have already taken some measures to protect their work?

Creators started installing tools that block the reading of content, “poisoning their data”, which, after reaching AI systems, “makes them sick” and they start generating unpredictable content.

There is a tension between creators and AI companies. OpenAI claims that it is impossible to create artificial intelligence tools such as ChatGPT without copyright-protected material. Thus, why is there such a reluctance to remunerate creators for their creative contribution to successful development of AI technologies?

After discovering that there already are several companies that decided to take initiative and voluntarily are looking for ways to pay remuneration to creators for the use of their content in AI systems (Adobe, Canva, Shutterstock, OpenAI contract with Associated Press, YouTube contract with Universal Music Group), I realised that that was the answer. All AI companies should pay remuneration to creators for the use of their content.

What did you aim to achieve with your project, what goals did you set?

Knowing these remuneration initiatives, I realised that a sustainable solution exists. However, in my opinion, it is necessary to not leave this to market self-regulation, due to a simple reason: if there are no clear legal references to payment of remuneration, most AI companies will choose not to do it. Therefore, I propose to introduce some sort of legal instrument which obligates all AI companies to pay remuneration to creators for the use of their content.

My goal was to describe potential models of remuneration to creators and submit a recommendation to the Ministry of Culture regarding further steps.

When there is no clear single solution or a functioning good practice, the primary impulse is very important which you can use as the jumping off point and plan other work. Thus, I described several remuneration models in detail, evaluating their advantages and drawbacks.

So, what are those models?

In my analysis, Potential models of remuneration to creators for the use of their content in artificial intelligence systems, I singled out two main groups of remuneration models: individual (income sharing, royalty payments, a fund, and their various possible combinations) and centralised (data set platform and adoption of compensatory remuneration).

Individual remuneration models create a personalised relationship between AI companies and creators. They provide the creators with the opportunity to negotiate regarding the amount of remuneration, the usage of the content and the payment conditions. This increases the creators’ motivation to share their content with the AI companies that ensure just and acceptable distribution of remuneration.

However, small, new AI companies and those that develop a specific product would find it difficult to attract creators that are making the appropriate content because they would be unable to offer just as good conditions and remuneration as the big companies. In the case of individualised remuneration model, the most favourable conditions would not be ensured to small or niche creators, because in the negotiations with AI companies those conditions will be dictated by the popular, large creators and the collective organisations that represent the creators’ rights.

Centralised remuneration models are a more universal system whose potential can encompass all creators that wish to participate in the scheme regardless of the level of maturity, quantity, or character of their work. Moreover, all AI companies would have equal conditions for access to high-quality data (in the case of data set platform) and transparent, clear format of remuneration and its predicted amount.

In the case of centralised models, for AI companies and creators or the collective organisations that protect their rights, one contract would be sufficient with a remuneration model administrator who would ensure fair cooperation, transparency, sustainability, and continuation of relations.

Since centralised models encompass very many elements, in my analysis, I evaluated similar solutions that already exist in Lithuania, considering the potential of their application in collection of remuneration from AI companies.

What threats to authors are posed by the increasingly prevalent artificial intelligence?

No less than three creative occupations are included in the top 15 occupations that will be most affected by augmentisation and automatisation. They are editors, graphic designers, and translators. Already, the artistic content generated by AI is winning competitions and is widely used in various projects.

However, I do not see a big threat to the creator. On the contrary: new sources of inspiration and different creative solutions will emerge. Yes, there will always be those who will categorically reject work with AI tools. And that is not a bad thing: after all, when photography emerged, painting did not disappear. When cars appeared, horses became a symbol of luxury. There will always be people who strive for authenticity and who will pay more for the opportunity to make use of human creativity.

It is important to understand that AI tools will become more and more popular, they will improve, and the product created by them will be increasing in quality. Thus, a creator who is making a commercial product should listen and learn to work with AI tools. Businesses won’t wait multiple weeks or months in order for their advertising idea to be implemented. The divide between “here and now” and “reflected, felt out” creation will become as clear as ever.

How, in your opinion, should synergy be achieved between creators and artificial intelligence companies?

AI companies need high-quality and often copyright-protected content: data. Businesspeople have already started calling data the gold of this century; thus, it is important to deal with data appropriately. We have to know where and how they are used. It is important to ask for the authors’ permission to be allowed to work with their content, appropriately remunerating them for this.

You have conducted analyses of the current situation and the good practices of foreign countries and organisations. What are their essential aspects?

The subject of my project is still barely discussed at national levels, even though there have been some reactions stating that it is done quietly.

Among EU countries, so far only France has taken on this issue loudly: a proposal known as No. 1630 aims to change France’s Code of Intellectual Property, taking into consideration the challenges in copyright that are raised by AI. Specifically, it stipulates creation of a tax regime for work that was made using AI and utilising copyright-protected material. The goal of this proposal is to effectively manage copyright issues and ensure appropriate remuneration to authors of original work.

At the moment, the U.S. Copyright Office is examining how the consent of copyright-owners and remuneration for the use of their work in the AI training process should be regulated. Various licensing models and their impact are considered, possibilities of compulsory licensing are being researched.

The Vice-Minister of Culture Vygintas Gasparavičius raised the issue of remuneration to creators at the International AI Ethics Forum which took place in February in Slovenia. At the end of this month, the topic is to be discussed with Dr. Angelika Schlunck, State Secretary at the Federal Ministry of Justice of Germany. I’ve also mentioned Poland’s perspective on the use of creation in AI training. It looks like this is still a very hot topic and we may see more unexpected turns.

Nevertheless, we should realise that international AI companies may not comply with solutions adopted at Lithuania’s level; thus, when dealing with the issue of remuneration in the context of AI, it is vital to find partners that support similar ideas in other EU countries and cooperate with them in order to present proposals regarding EU-level solutions.

Photo by G. Tvarkūnaitė / Archive of Kurk Lietuvai