Deep Dive: Navigating AI in the Cultural Sector

Tobias Rohe: Hello and welcome back to Die Kulturmittler:innen Deep Dive, experts on international cultural relations. My name is Tobias Rohe and it's a pleasure to be with you again.

The use of artificial intelligence systems in the field of international cultural relations has brought about a variety of reciprocal changes. In an era of increasing geopolitical competition, this poses wide ranging opportunities, but also challenges. How can the integration of AI in international cultural relations be managed to maximize benefits while mitigating risks?

How can we address ethical, linguistic, creative, informational, and geopolitical challenges? With our guest today, Octavio Kulesz, we want to talk about the central points of his newly published study in this field. It is entitled Artificial Intelligence and International Cultural Relations, Challenges and Opportunities for Cross Sector Collaboration.

The study has been published within the ifa Research Programme. culture and foreign policy. And I'm very pleased to welcome the author in this podcast episode, a warm welcome to you, Octavio Kulesz. Hi

Octavio Kulesz: Hi Tobias, it's great to be here.

Tobias Rohe: Very nice to have you here with me. You are a philosopher and a digital publisher and your work as a researcher focuses on digital culture. And in 2020, you were selected by UNESCO along with 23 other experts to draft the Recommendation on the Ethics of Culture of AI, the world's first standard setting instrument in this area. And some of your articles anticipated the contemporary challenges associated with generative AI. Now, Octavio, can you give us a brief idea on how AI and international cultural relations are connected?

Octavio Kulesz: As we all know, in the last five years or so artificial intelligence or AI has transformed almost every aspect of our lives and international cultural relations or ICR has not been an exception. In fact, AI has shaken the foundations of all dimensions of ICR. Let's think of languages, communication, the information environment, geopolitics, and so on.

So it's very clear that AI has had a huge impact on different areas of ICR. But it's interesting to see that there is also the opposite effect because ICR and culture in general is having a huge impact on the development, on the research, and on the regulation of AI. Because let's consider, for example, AI research.

Research in the field of AI is a very internationalized field where thousands of researchers, scholars cooperate on a global scale and they come from different universities, from different backgrounds within computer science, data science, engineering. They speak different languages. They come from different cultures.

So this is also a part of international cultural relations. If we take the regulations of AI, it's also a very cultural field, so to speak, because each country has different priorities, has different languages, different approaches to technology. So in a way, this is also an effect of international cultural relations and each country will try to learn from each other.

Other regions in the world to take the regulations that can help them to navigate the current environment. And the third reason why I think there is an effect of ICR on AI is that a very important component right now in the discussion about AI is the what we could call the imaginaries or the narratives around AI, because basically what drives, ultimately what drives the regulation or the development of AI tools is not so much pure science or pure technology, but rather this group of narratives or imaginaries that are very important.

For example, one, one key imaginary now is the AI race. So everyone thinks we are running a race and no one wants to be left behind, et cetera. But this is just one imaginary among many others. And we could build and create and reshape those imaginaries.

Tobias Rohe: And what do you think are the potential risks and challenges that arise from the integration of AI in international cultural relations?

Octavio Kulesz: There are many opportunities as we know, and there is a lot of hype in the media and it's real. The productivity the increased efficiency and new tools, new applications everywhere. This is very clear, but there are also Considerable challenges and risks, for example, for cultural diversity in terms of this information. If we think of cultural diversity, so the negative effects of AI on the cultural sectors, let's consider first that for any AI application, at least any machine learning or deep learning application to work it needs first to ingest large amounts of data. And these datasets comprise all sorts of human and creative expressions, such as texts, songs, illustrations, videos and so on. So those works are in many cases, works of art. So artistic works. So the machine is trained with those data sets and it can later generate in the case of generative AI, it can later create a new song, a new video, et cetera.

And the problem is that. There are many artists involved here because they have created those songs, those novels, et cetera. And there is no compensation for them. There is no recognition for them. So the AI companies in general work with some opacity. There is no transparency regarding the training data.

And the risk in the future eventually is that AI companies may end up replacing these artists. Another problem is related to the quality of the content generated by AI tools. You know this expression 'garbage in garbage out'. It's totally accurate here because if we feed an AI system with data that reflects the value of a part of humanity, the output data will reflect those same values.

So the problem of bias is very clear here, and it's also a threat and a risk in terms of cultural diversity. So cultural diversity, and then disinformation, misinformation, this is also a very serious problem. And finally, last but not least, censorship is a serious problem that is aggravated in the AI era because censorship can reach a new level now.

It's becoming more and more sophisticated, not just state censorship, but also private or platform censorship, where AI can be used to identify content that needs to be removed or where platforms or governments are becoming more and more specialized in filtering the content that will be ingested by an AI system.

So they are starting to censor not just the output of the system, but also the input. So if I filter, if I censor the training data, I will make sure that the outcome will be censored as well. And this is very difficult to spot by the users. So many users won't realize that the whole environment is being censored, but actually it's already happening in many parts of the world.

Tobias Rohe: But are there also opportunities? You talked a lot about the risks and challenges, but other opportunities that arising from AI and international cultural relations.

Octavio Kulesz: Yes, the opportunities are huge in if we take for example the creative sectors in most areas AI tools are really beneficial.

For example, if you take music or visual arts or publishing, all these tools can allow, for instance, any person without the specific knowledge in this field to create a new work of art. But of course, this is not without its challenges, as I said before. So democratization and more productivity for cultural and creative industries.

and a much more abundant cultural offerings for the public. So across the value chain, we see many benefits.

Tobias Rohe: Now, in addition to problems of a technical nature, you already talked a bit about here. There aren't numerous influences on society by AI. Let's call them social consequences. Now, and let's talk about one of them in particular.

How does AI impact the preservation and promotion of languages, particularly less widely spoken or low resource languages?

Octavio Kulesz: Yeah, this is a very important issue because right now, not all languages are equal. And it really depends on the amount of data and the quality of the data that is available in each language.

If we take English, I would say it's English is the king of high resource languages. We could define, as you said, languages and group them in two. One, high resource languages, English, Spanish, Chinese, French and so on.

Tobias Rohe: Where there's a lot of content available in the internet.

Octavio Kulesz: Exactly. A lot of content and high quality content because it's not just quantitative, it's also qualitative.

And then for most of the 7, 000 languages spoken today in the world, those are low resource languages. Where we cannot find abundant materials to feed an AI system in that language. And this poses several problems because you cannot build your own language model in Aymara, for instance, which is a language spoken in Peru, Bolivia, or Chile.

It's very limited. But you can build your own language model in Spanish, or in French, or of course in English. And if you take multilingual models, Like most of the current models right now, what developers do is to try to fill the gaps by building a sort of bridge between high resource languages, especially English and the rest of the low resource languages.

But this poses problems because when, if you ask a question in Aymara , let's say in ChatGPT, the answer will not be very good. So you, if you are a speaker of that language. You will feel that the grammar is not perfect, the words are not really accurate, etc. So low resource languages are really in a disadvantaged position today compared to high resource languages.

Tobias Rohe: And how does that impact on international cultural relations?

Octavio Kulesz: It will impact the situation in terms of, I think, in terms of communication it will introduce many biases and it will strengthen the position of the dominant languages in the future. And especially because it's not just about the data sets, but it's also about the very logic of the approach in AI research. Most research now in, in natural language processing in AI devoted to languages is that a language could be tokenized. So this is the typical approach in English or in Western languages where the sentence can be divided into different components that have a meaning.

So words, but this is not the case in many first nation languages where the verb, the subject, the object are merge into a single word. So in the future, the problem is that we will lose diversity in those languages. And the very, actually the very use of those languages will be compromised because the logic is different and it doesn't really fit into the current AI research approach.

And also another key element in my view is that programming languages. For example, in Python, which is vital for AI research, depend on English. They are written in English. You have instructions such as 'input', 'print'. And if we go a bit further, we will see that most AI conferences are in English. If your mother tongue is Spanish or German or Aymara or any other language, you will have to make a double effort always.

You will have to be good. You will have to excel at your discipline, so AI, data science, etc. And you will have to be a good English speaker. There are many conferences in other languages, but they are more related to the effects of AI. Conferences about the social, technical effects of algorithms, so harmful effects, disinformation, et cetera, but not conferences in local languages about the very development of AI.

This is very rare. So the power imbalance can become very serious and irreversible. And it will favor English speaking nations.

Tobias Rohe: Talking about power, let's talk about the geopolitical implications of AI. What relevance does AI have in terms of foreign policy and therefore also foreign cultural policy? How is it used to pursue national interests?

Octavio Kulesz: The geopolitical implications of AI are huge, and as I mentioned at the beginning, the predominant narrative around AI on a global scale is the idea of the AI race. Where most countries, or at least most powerful countries, are running this race against one another. And no one wants to be left behind or be left out.

The big powers right now are three first, the U S then the you and China with very different approaches. So it's interesting to see that in spite of having a sort of global narrative of the race, actually, there are local narratives because each country has its own national interest in its own national priorities, problems, and these are very different in each region.

So if we take the European Union, where that many analysts consider as a right based region, so Anu Bradford said that these three digital empires are based on different approaches. The EU is right based, the U. S. is market based and China is state based. Actually, the situation is much more complicated because in China, the private sector is extremely strong, and on the other hand, in the U. S., also the public sector is very strong. But, in any case, in the European Union, a few months ago, the AI Act was approved. This is a very comprehensive regulation that establishes different levels of risk, And includes a special chapter. So to speak about general purpose AI, the idea is that the regulation will last for several years and it encompasses many sectors. Whereas the approach that we find in a country like China is quite different. China has not invested in a comprehensive, in a single regulation, but rather in separate, very specific regulations. For example, regulating deep fakes or the training sets, et cetera. It's a very agile approach. And the U.S. Approach. Until very recently was avoiding this type of comprehensive general regulations, but the situation changed in last October, where the Biden administration introduced the executive order, which is a whole of government regulation for not just public agencies, but also for some private actors.

And apart from the approach that we find in China, the EU and the U.S. which are the main regions in terms of AI in the world. We can find other interesting cases such as the UK, Canada, and South Korea or Japan, but they, so far, they have not been able to produce such a comprehensive approach or such a stable approach and, but many laws are in process.

And then we have the Global South, and this is where geopolitics really become important because the Global South Apart from the case of India, maybe, which has a critical mass of researchers and infrastructure, et cetera, the rest of the countries, if you see the situation of Latin American countries, African countries, they are really struggling to find their own strategy, their own approach.

And the risk here is that we may see in the future a sort of colonial like approach from the superpowers in the field of AI towards countries in the global South, because many of the companies from countries in the West or companies from China, they extract data as a raw material from the global South.

And then they come back to sell their own solutions.

Tobias Rohe: You were talking beforehand about ICR influencing AI development. So what measures and strategies do you recommend to integrate and promote cultural aspects in national and international AI policies? And how can ICR actors gain a seat at the negotiation table?

Octavio Kulesz: I think it's absolutely essential that ICR organizations have a seat at the negotiation table because we cannot leave this Exclusively in the hands of tech companies or in the hands of government technology agencies, because the view of the cultural sectors is really important, not just for the cultural sectors themselves, but also for the rest of the ecosystem, even for tech companies and government.

So this is why in my report, I propose different lines of actions, a very concrete recommendations in different fields. The first area is training and awareness. It's absolutely vital that all the institutions that directly or indirectly work in the field of ICR have access to AI literacy. So not just their directors, leaders, but also their, in general, their internal staff.

They need to know more about the opportunities and challenges of AI. And the second line of action I propose is to work more on new collaborations. The digital era is always about collaboration, because digital means contact. Digital means cooperation. And that's why, It's important that ICR institutions start working, for example, with tech companies.

They need to understand better how those actors think, and in order to approach and build new collaborations. Build new collaborations also with other public agencies, with players, not just in the Global North, but also in the Global South, etc. The third pillar is to try to have a bigger impact on policies.

So I mentioned the fact that we cannot leave this in the hands of tech companies only. But this means that we have to make sure that the cultural dimension, the cultural perspective is present in all the AI strategies, in all the AI regulations. Until very recently, the term culture was totally absent from AI regulations.

National regulations or global frameworks. And maybe the text that changed the situation for the first time was the UNESCO recommendation on the ethics of AI, which includes a special chapter on the cultural dimension. Then the EU with the AI act has also introduced different cultural aspects. And I hope there will be in the future, more and more regulations and national agendas that include the creative and the cultural sectors in the equation. Also new debates. We need to challenge the predominant narratives about AI. The, this question about on the AI race or super intelligence, those are very controversial imaginaries and metaphors that we could challenge, we should reshape the debate.

These are not the only possible narratives we could in a moment, in a world where conflicts are emerging everywhere, we need to invest more in. It's narratives based on collaboration, cooperation, intercultural dialogue.

Tobias Rohe: Thank you for these insights. Now getting to the last questions of our interviews, I would like to look into the future.

How do you see AI reshaping the landscape of international cultural relations over the next decades?

Octavio Kulesz: I think all these trends will intensify in the future. We will see many of these challenges becoming more and more serious perhaps. And one of the trends that I try to anticipate in my report is the fact that we will soon see the rise in my view of the AI agents.

So far we were used to dealing with very simple applications of AI. So let's take, for example, ChatGPT. On ChatGPT, we introduce a prompt and after we introduce this prompt, we get an answer and then we can establish a conversation with the bot. That's all. We can get a text, we can get an image, etc. But things will change a lot in terms of ICR when these assistants, AI assistants, become agents.

So when they gain the possibility to act and to perform actions, especially on the web. Because I think in a few months we will be soon able to ask our assistant. Not just to create a text, but to go and browse the web to search for a specific question. And they are already able to do this. But browse in a much more deep way and to interact with other bots.

And also to negotiate, for example, imagine that I want to buy a ticket, a flight and a hotel room, et cetera. I can just ask my bot to do that and my bot will know me, will know my preferences. I will give the bot a few specifications, goals, and the bot will do this in a couple of minutes or probably longer.

And this introduces new problems because this is not limited just to the purchase of flight tickets or hotel rooms. It could also be applicable to social networks to the creation of all sorts of contents, collaborate in a collaborative manner, et cetera. So the interaction will start moving from the current human AI interaction to another type of interaction where more based on human AI, human or AI, human AI, and all possible combination.

And this will propel ICR to another level. It's totally uncharted waters. And we need to work in advance to better understand the risks involved here. Because there will soon be more agents than humans. That's for sure.

Tobias Rohe: And how can institutions involved in international cultural relations positively influence maybe this development?

Octavio Kulesz: They need to be promoting more training across the cultural sectors. They need to make sure that the public sector understands the situation now and that cultural diversity is respected in the future. We need to promote, protect cultural diversity in a situation where AI agents will proliferate and We don't know how this will eventually turn.

And the problem just to finish, the problem is actually never the technology. The problem is the human there behind the technology. So biases, manipulation, disinformation, and all these swarm of AI agents. Populating the web will make it much more difficult to identify the malicious actors or the bad intentions, et cetera.

Tobias Rohe: It's the people using the tools who are the problem.

Octavio Kulesz: Exactly. It's us. It's us.

Tobias Rohe:

As always. Thank you, Octavio, for these insights and your time. It's been an enormous pleasure talking to you. Tobias. And I certainly hope that you enjoyed this deep dive episode of Die Kulturmittler:innen. I would be very happy if you tuned in next time.

If you are interested in Octavio Kulesz's study, then take a look at the show notes. You will find a direct link there. And if you enjoyed listening to this episode, don't hesitate to share it with your friends. To make sure that you don't miss out on future episodes, subscribe to Die Kulturmittler:innen right away.

You can do that wherever you listen to the shows of your choice. Whether it's on Apple Podcasts, Spotify, Google Podcasts, Deezer, or Amazon Music. And while you're there, don't forget to listen to our regular episodes of Die Kulturmittel:innen with dozens of in depth conversations on the topic of culture in foreign policy.

For all other information on the Forum for International Cultural Relations, visit our website at culturalrelations.ifa.de That's all from my side. I say thank you for listening. My name is Tobias Rohe. Until next time.