“We are at the beginning of a New Cultural Epoch”.
An Interview with me by Maja Beckers, “Die Zeit” (2. Oktober 2023)
German national newspaper “Die Zeit” has a thought-provoking format: “What are you thinking about?”. In this series, the journalists ask leading scientists and public figures what they are thinking about right now. The questions are posed by Maja Beckers, Andrea Böhm, Christiane Grefe, Nils Markwardt, Peter Neumann, Elisabeth von Thadden, Lars Weisbrod, and Xifan Yang. In October, Maja Beckers from Die Zeit talked with Merzmensch (Vladimir Alexeev) about AI, Art, and Creativity. Read it in English.
ZEIT ONLINE: Vladimir Alexeev, what are you thinking about at the moment?
Vladimir Alexeev: At the moment I’m thinking that we are at the beginning of a new cultural epoch, and hardly anyone notices it.
ZEIT ONLINE: What do you mean by that? A new cultural epoch?
A paradigm shift is coming. And I don’t mean that in a positive or negative way, but rather as a neutral observation
Alexeev: I have worked on the intersection of culture and technology for many years. I now observe a very intense preoccupation with the question of creativity among artists, researchers, and developers, especially in relation to artificial intelligence. What is that, and what new avenues are opening up? We are still in the early years of this development, but I am firmly convinced that a paradigm shift is coming. And I don’t mean that positively or negatively, but rather as a neutral observation. At the same time, hardly anyone notices because we are distracted by the scandals, frictions, and hype that such a disruptive technology naturally generates.
ZEIT ONLINE: What exactly do you attribute to this new era?
Alexeev: I see parallels, for example, with the historical avant-garde at the beginning of the 20th century. After the First World War, the avant-garde questioned traditions, culture, and the whole civilization, which considered itself a “high civilization”, but could not prevent this terrible war. I see similar subversive tendencies among artists and researchers working with AI today. They are looking for entirely new forms of expression, ways of telling stories, and aesthetics.
ZEIT ONLINE: What is new about these forms of expression?
Alexeev: What is new is that more and more works are being created in cooperation between humans and machines. This requires entirely new skills and new craftsmanship. And the forms of expression are expanding; there are new ways to create specific texts, images, or music, even if you can’t paint or compose. These can be combined with traditional genres, or new things can be created. In my opinion, this will further expand our concept of art. In the 20th century, there were many new media, which is happening again: machines are being used, and algorithms and models are being created especially for this purpose. And they are also open to those who, in the past, would have said: “I’m not an artist because I can’t paint.”
ZEIT ONLINE: Does it now apply, as Joseph Beuys said: “Everyone is an artist”?
Alexeev: Absolutely. In many ways, the 20th century prepared us for the 21st. With its skeptical theories about concepts like originality and artistic genius, for example. But also with statements like Beuys’. With AI, suddenly, everyone can work artistically. Marcel Duchamp also often called on people to stop going to museums so devoutly and to start creating something themselves.
“Everyone is an artist” (Joseph Beuys)
ZEIT ONLINE: But if the call used to be to become creative, then it is not fulfilled if the AI does it for you.
Alexeev: I don’t think the AI is taking over. On the contrary, it expands the possibilities for creativity. Even training the machines is a human task: what do I put in? And even if the machine spits out something, that’s not the end. An artist, for example, could process AI-generated video sequences for an installation. We curate the result, process it further, and put it into a specific context.
ZEIT ONLINE: If there is no human authorship, as you write in your book AI Art, most people feel cheated when they learn that an image they are looking at is a pure AI production.
Alexeev: Yes, a study by the University of Amsterdam showed this. Many people still think very anthropocentrically that the domain of creativity is purely human. The idea that only humans can be creative is outdated.
AI art is not art at the push of a button.
ZEIT ONLINE: But doesn’t this anthropocentrism also make sense? For me, as a human being, what makes a work of art valuable is that another human has worked on it. That is not only an important aspect, it is also an essential point.
Alexeev: The editing by humans will also remain. AI art is not art at the touch of a button. I also find exciting the results a machine produces when dreaming or hallucinating. It’s like a mirror for us, and I find it exciting to see ourselves and our society filtered through these processes.
ZEIT ONLINE: You are so convinced of the creative potential of these technologies that you are an honorary ambassador for GPT-3 and 4 and DALL-E 2 and 3 of OpenAI. How can I imagine that?
Alexeev: I work with artists or writers interested in working with these programs. I advise them if they are wondering how to get DALL-E to produce a particular aesthetic or how to keep their style when working with AI-generated text. This is, as you say, voluntary. It interests me a lot, and I also get an insight into current developments in OpenAI.
ZEIT ONLINE: What do you think of the current lawsuits, for example, by authors against OpenAI or artists against Midjourney for copyright infringement? The co-creation of man and machine creates new problems, such as the authorship question. It is not just a question of who wrote a novel involving a human and an AI. But also, what about the authors whose material was used to train the AI?
Alexeev: This is indeed an important point. We need to take a step back historically and explain that most AI models we use creatively today were developed as research projects. It was often just a matter of using as much training material as possible. People wanted to see if it would work at all. It has been shown to work. Now, the data sets should be recompiled, and artists should be given the opportunity to say whether they want their work in there or not. If they do, they should be compensated for it. This will become a very normal process in the future.
ZEIT ONLINE: One reason an artist might decide not to appear in these datasets could be they don’t want the AI to plagiarise them all the time. In your book, you mention the case of the digital artist and game designer Greg Rutkowski, whose style was repeatedly copied by the AI. Now, when he searches for his name on specific art platforms, masses of images appear that could be his but are not.
Alexeev: Yes, there are good reasons and rights not to want your work in these training sets. And an analog world of art and culture that has nothing to do with AI will continue to exist. People will still go to museums, and there will still be works hanging there that have had no contact with AI. I suspect that there will be three strands to art in the future. One area that will remain completely untouched by AI. Another area is where artists will use AI-based tools to create and process AI-generated content. And then there’s an area in between where writers, for example, will ask the AI for possible plots but won’t adopt the suggestions and use them for inspiration. However, the word “plagiarise” in your question bothers me.
AI does not plagiarise.
ZEIT ONLINE: What bothers you?
Alexeev: An AI does not plagiarise.
ZEIT ONLINE: What else does it do? Create a remix? A collage?
Alexeev: None of the above. Because these models don’t create a collage out of whole parts of a novel, for example. That wouldn’t work because they don’t contain the contents with which such a model has been trained. It’s not like a pile of paper in a drawer that the model can access. The model has only semantic vectors. That’s why I think the term “stochastic parrot,” often used for language models, is wrong. There is nothing parroting here — there is much more behind the generation of texts than just probability calculations. So, I would instead compare AI to a child who goes to school and learns how to write, how to interpret, learn about different styles, and so on. When an art student paints a Rembrandt, it’s to learn how to play with light and shadow, for example.
Stochastic parrots is a misleading conception.
ZEIT ONLINE: But these models can still pick up details from novels, for example, or copy styles. Do you think that this changes the very idea of originality and authorship?
Alexeev: They are in a constant state of flux anyway. This new era will significantly influence concepts like originality and authorship. The idea may prevail that there is no such thing as an autonomous artistic genius but that every artistic creation draws on what already exists. Or the flood of AI-generated images will give human creativity even more significant meaning and will upvalue it. I think it’s essential to take into account the new possibilities and, at the same time, maintain control over AI as a human being. Copyright questions and which works can be used by AI also come to mind: We should not give AI free rein. At least not outside certain art forms. But we also talk about it more, explore the creative capabilities of AI — and use it for good.
