Quick access:

Go directly to content (Alt 1) Go directly to first-level navigation (Alt 2)

Cultural Expression and Machine Learning
The Wolf We Feed

Whether it is composing music or painting pictures, artificial intelligence has the technical and creative capacity to make art.
Whether it is composing music or painting pictures, artificial intelligence has the technical and creative capacity to make art. | Photo (detail): © Adobe

Whether it is composing music or painting pictures, artificial intelligence has the technical and creative capacity to make art. Argentinian publisher Octavio Kulesz predicts this could lead to an explosion of creativity and increased market concentration at the same time. 

By Harald Willenbrock

Octavio Kulesz, as a publisher and entrepreneur, what awakened your interest in artificial intelligence (AI)?
 
I have studied the evolution of AI since 2016, the same year AlphaGo defeated Lee Sedol, 18-time world champion of the Chinese board game Go. 
 
Lee Sedol was a human being and AlphaGo a program developed by Google’s DeepMind Technologies company. Which means AI beat the human world champion in this ancient Chinese strategy game for the very first time.
 
Exactly. Beyond the resounding victory – AlphaGo crushed its human opponent 4 to 1 – my attention was caught by a comment Fan Hui, another great Go player, made about one of the machine’s moves: “It’s not a human move. I’ve never seen a human make this move. So beautiful. Beautiful. Beautiful.” His words really impressed me. At the time, I was working for UNESCO on a variety of papers exploring the impact of technology on the diversity of cultural expressions. When I realized that machines could not only have huge computing power, but also creativity of a sort, I began doing more research into how deep learning systems could revolutionize art and culture in future. Human brain versus computer programme: In March 2016, Google’s artificial intelligence programme AlphaGo beat South Korean professional Go player Lee Sedol (on the right; on the left: Google DeepMind’s lead programmer Aja Huang). Human brain versus computer programme: In March 2016, Google’s artificial intelligence programme AlphaGo beat South Korean professional Go player Lee Sedol (on the right; on the left: Google DeepMind’s lead programmer Aja Huang). | Photo (detail): © picture alliance/AP Photo/Lee Jin-man Where and how is AI already shaping our culture today?
 
The impact of AI on culture is considerable right now and will only increase in future. Currently algorithms govern the books, songs and movie recommendations we receive on social networks and other platforms. We also need to recognize that AI is advancing rapidly at other links along the cultural chain: not just distribution – as with recommendation algorithms – but also creation and production, as deep learning is increasingly used to create works of art. 
 
For you as an e-publisher of academic books: How and where do you profit from AI? What tasks are AI tools doing for you today?
 
At present, we use AI to recommend related publications and to extract metadata automatically. We are also making progress in other areas, such as standardizing bibliographic references.
 
If machines can learn to compose music and paint pictures, what does that mean for human creativity?
 
Disruptive technologies often force artists and creative industries to redefine their places. In the case of AI, deep learning systems such as generative adversarial networks (GANs) are paving the way for an exponential increase in artists’ capacity for inspiration and creating new artworks. We need to build a new relationship with these technologies. Machines will not replace artists, but they are requiring them to rethink their role. It is no accident that many AI artists do not sign their own names to a piece and instead use a pseudonym that somehow refers to both themselves and the machine, as if both – person and AI – collaborated on the piece. 
 
When you talked to mathematician Marcus de Sautoy at EUNIC’s AI Week in December 2020, Marcus argued that AI met all three criteria for creativity, as it produces things that are new, surprising and valuable. Would you agree with that assessment? And if so, has AI then achieved artist status?
 
I agree with Marcus that those three elements are essential for defining creativity. However, I think a fourth should be added: the intention to create. Machines have neither intention nor will, at least not at the moment, so I would argue they cannot be considered real artists. 
 
What advantages do you think AI has for cultural expression? 
 
I believe that AI systems can help us augment our creativity in a number of ways. The first touches on inspiration: AI can enable us to generate hundreds or thousands of drafts in an automated way, which we can then select, recombine, etc. This is clear in the case of Obvious Art, the French artistic collective that has used GANs to create renowned paintings, such as Edmond de Belamy, which sold for US$ 435,000 in 2018. AI tools allow us to become much more efficient and execute production processes in just a few seconds that would have taken weeks or months in the past. Tools like Adobe Sensei and others can produce video edits, doing colour correction, image stabilization and visual effects a lot faster than we can. AI could also lower the entry barriers for newcomers in the creative sector. The French artists collective Obvious Art’s portrait of Edmond de Belamy was created using artificial intelligence and sold for US$ 435,000 by Christie’s auction house in New York in October 2018. The French artists collective Obvious Art’s portrait of Edmond de Belamy was created using artificial intelligence and sold for US$ 435,000 by Christie’s auction house in New York in October 2018. | Photo: © picture alliance/AP Photo/Christies This seems surprising since AI techniques are not easily accessible.

It is true that AI techniques are not easily accessible to traditional artists such as guitarists, painters, sculptors and others. But AI could help those who do have no expertise in specific creative sectors – music, painting, sculpture – to become creators. For example, a user with no knowledge of orchestration can compose a symphony in a few minutes using systems such as AIVA or Amper. This is how AI lowers the entry barriers for many users who do not have an artistic education. This might mean that many creative industry professionals could face difficulties in future.

What creative professions are you thinking of?

To the extent that AI allows tasks to be automated, several activities performed by human beings today will be delegated, at least in part, to machines. These include translation, proofreading, graphic design, music composition, and image and video processing. This does not mean that automation will fully replace these creative professions, but if machines manage to perform a certain percentage of these activities, those who master AI techniques will see their potential increased, while those who cannot use those techniques will be greatly affected. 

How big is the risk that AI as a cultural tool leads to an even higher market concentration for the large tech companies?
 
The great danger of AI lies not in a possible machine rebellion – which is currently nothing more than a popular science fiction scenario – but in the big technology companies gaining too much power. Economic concentration is always harmful, and in the cultural sector the risk is that we could have monopolies or oligopolies not only in the distribution of cultural goods and services, but also in their creation and production. 
 
This might be even more true when Google, Apple, Facebook and Amazon (GAFA) go from being distributors of cultural goods to producers themselves.

Exactly. The very structure of large technology companies operating as platforms whose most precious asset is data allows them to dominate the entire value chain: creation, production, and distribution. We no longer see classic horizontal concentration in which some powerful players buy their competitors. Today’s tech titans form closed marketplaces. They do not dominate a specific segment; they have become the market as such. The widespread use of AI will only further this trend.
 
When AI becomes a creator of cultural goods – of visual, musical or literary works – who then owns the copyright to these artworks? The software engineer who programmed the AI? The tech company that owns the AI? The human artists the machine learned from? A combination of all three? Or nobody?
 
This is a key point in the conversation around AI and creation. I discussed this issue in detail in a guide I recently wrote together with Thierry Dutoit on AI in the arts and creative industries, published by the International Organisation of La Francophonie (OIF) and Wallonie-Bruxelles International (WBI). The conversation is still ongoing. But legislation has tended to define the actual creator as the one who used AI to generate the work. What is clear is that the machine itself cannot be considered the creator since it lacks intention. 
 
If we look at AI in its role as a distributor of cultural goods it customizes for individual consumers according to their specific preferences, couldn’t this lead to a loss of common narratives and discourses, a world in which every individual consumes their own culture program?
 
About 10 years ago, Eli Pariser coined the expression “filter bubble” to refer to the universe of information that algorithms create for each of us. Social networks tend to produce echo chambers that reinforce our beliefs and, in many cases, end up isolating us into subgroups. As AI is used not only to distribute content but could also be used to create it to suit the taste of each customer, we could achieve a “perfect bubble” situation in which cultural expressions would be generated in an automated and personalized way, and cease convey meaning, values and common narratives. Now the question is whether a society is viable without a shared cultural identity.
 
Obviously, AI opens up new possibilities and creative options – tools, influences and modes of expression – while at the same time narrowing the creative sphere, for instance through limited access to the technology, the influence of tech companies, the creation of social bubbles, etc. Is that an unsolvable paradox?
 
If current trends continue, it is highly likely that we will face an explosion of creative content, as well as greater economic concentration. This is in fact the problem repeated with every technology. We need to pay attention to more than just the total wealth generated and look at how this wealth is distributed among the different stakeholders. It will be up to all of us – users, artists, entrepreneurs, policy-makers and more – to agree on a more equitable, pluralistic and diverse scenario.
 
What then do you see as the bottom line for now: Does AI mean broader diversity in cultural expressions or more limitations?
 
Technologies have no life of their own in any real way. The key lies in what we do or don’t do with them. The situation reminds me of the story of the two wolves: An old Cherokee is teaching his grandson about life. “A fight is going on inside me,” he says to the boy. “It is a terrible fight, and it is between two wolves. One is evil – he represents anger, envy, sorrow, regret, greed, arrogance and ego.” He continues, “The other is good – he is joy, peace, love, hope, serenity, kindness and compassion. The same fight is going on inside you – and inside every other person, too.” The grandson thinks about it for a minute and then asks his grandfather, “Which wolf will win?” The old Cherokee simply replies, “The one you feed.” We are facing the same dilemma: Depending on the decisions we make in the coming years, the cultural ecosystem could become more vibrant and pluralistic. Or it could be controlled by a handful of technological companies whose vision of the arts is purely utilitarian. 
 
If we met again five years from now, how would your answer to my second question – where and how AI is already shaping our culture today – differ from the one you gave today? 
 
It is very difficult to make projections because we generally approach the future as a continuation of past trends. Advances in AI can become exponential and render our boldest projections obsolete. Now that AI systems can even write programming code, the only thing we can anticipate is that we will witness not just quantitative but also qualitative leaps of enormous magnitude in the next five years. The massive digitalization resulting from lock-downs and quarantines has only accelerated these processes. I would like to think that the different social and political forces will become aware of the implications of these changes and will act to defend the values and principles linked to democracy, pluralism, diversity, equality, non-discrimination, the rights of minorities, traditional knowledge, the perspective of indigenous peoples and respect for nature in an era dominated by AI. However, the transformation may occur so quickly that only those with inside information will be able to take advantage of the situation, which would only reinforce monopolistic situations that could then become irreversible. UNESCO recently released a global draft recommendation on the ethics of artificial intelligence that explores all these urgent issues.
 
Finally, do you have a piece of AI-enhanced or -created art, music or literature that you can recommend?
 
I recently participated in a webinar on the impact of AI on the diversity of cultural expressions organized by OIF, WBI and UNESCO. The panellists included Benoît Carré, one of the world’s leading AI music composers, who launched the album American Folk Songs in 2019. He fed the machine with works by Henry Purcell and other composers to create new arrangements for traditional folk songs. The result is extraordinary. The first song on the album – a version of the ballad Black is the Color – is truly sublime.

Top