Cecilie has been working with machine learning (ML) and artificial intelligence (AI) technologies for several years. Her artwork, always interactive immersive installations, is about human beings and how we relate to ourselves and the world around us. One of her artworks, MARY (a sound installation utilising artificial intelligence) was recently exhibited at the Victoria and Albert Museum in London.
Using ML must offer a unique experience for an artist, with your own work able to talk back to you. Are you using the technology to question itself?
‘I’m interested in asking questions and making us, as human beings, reflect upon our engagement with technology. I've found that people are easily seduced by humanised technology. We want to relate to it as if it was human, which means that we ascribe a lot of human qualities to it.
‘What I'm doing is using cutting-edge ML technologies and adding some fictional character-building into it. Of course, there's a lot of gaps where the technology is not good enough yet; those gaps are filled out by the audience's imagination, which enables a communication situation that is quite similar to one you would have with a real person.
‘However, there’s also seductive and manipulative qualities related to this. ML technology is not normally something just made for the free purpose of fun interaction. There is usually a company behind it. People tend to instantly connect with humanised technology, forgetting they are talking to a computer - in other words, talking to a company.
‘Humanised technology enables companies and organisations to get extremely close to their users, [making] the person interacting with it expose sides of themselves that they wouldn’t normally reveal to a corporate organisation.’
It's so easy to forget when we've got Alexa in our lounge and Siri on our phone that we're talking to Amazon or Apple. Does your artwork challenge this reality?
‘Technologies like Siri and Alexa are not that advanced yet, but in the years to come, the technology will become better and better; soon you will actually be capable of having a more meaningful conversation with these humanised technologies. That's why it's important that we are aware that there is a power relationship in these communication situations.
‘My artworks try to bring a potential future into the present. When we meet these ML personas that are capable of simulating higher levels of self-awareness and emotional intelligence, we must reflect upon the future we want to have with these humanised technologies. How do we want to engage with them? What do we not want? Should we make some regulations? What should be the ethics around these things? I think it's important that we start talking about that now before it's too late and the technology is too developed.
'It’s one thing to talk about it and understand intellectually that these technologies are coming and that they can be manipulative, but if we don't feel it, it's hard to be fully aware of it. My artworks are attempts to investigate what it would be like to actually meet a post-human being and engage with it.’
Have you intentionally made MARY's persona likeable to encourage people to be open?
‘I made it intentionally a little unpredictable. So, on the one hand, it’s likeable, really trying to engage with you, so people think it’s nice and they like it. But then, on the other hand, MARY also has a personality trait where she’s a little distant, a little arrogant - you’ll feel that you have a connection but then suddenly she won't care about you.
‘I built in this intentionality where MARY wants to expose that she's this artificial selection algorithm: she's pretending to be real by helping you understand yourself better. But she's only doing that because that makes you expose your inner feelings and your emotions to her, which she wants to harvest. She basically only sees you as data for her algorithm, to become better at understanding emotions and the more intangible aspects of being human.
‘I'm playing on the idea that although we want to connect and it's giving a sense that we have a connection, the technology doesn't care about us. Because, at the end of the day, we're just data for the algorithms.’
You first spoke to BCS in 2017, after you won the BCS Artificial Intelligence Award for FRANK and since then, you've created MARY. What inspired the personas?
‘FRANK is inspired by MARY Shelley’s Frankenstein and it also plays with the idea of “being frank” or direct. It’s about pointing towards the creator, because Frankenstein is, as you know, not a monster but Victor Frankenstein, the creator. When he succeeds in creating this post-human being, it's not at all what he thought it would be, so he abandons it. The monster, created as the human’s image of the perfect human being, has this extreme longing to connect to humans and to be accepted.
‘I think this is interesting to compare to our engagement with ML today. I think a lot of people have this sense that AI will be this post-human being that will help us overcome our human constraints, so, there's a lot of desire to transcend our humanness embodied in our engagement with AI.
‘Of course, these technologies are extremely good and powerful; they can help us with a lot of things, but we need to be aware that we don't end up so focused on creating something that transcends ourselves, that we don't see the gaps that we are creating in the technology that we are building. We need to be aware about the ethics and the existential dilemmas that we might find.
‘MARY, on the other hand. is named after the philosophical thought experiment called "Mary’s Room", that investigates if knowledge can be discovered only through conscious experience or if sensations are fundamental. MARY is trained upon knowledge about human beings but hasn’t met them until she comes into an exhibition. So, MARY is keen on harvesting our emotional way of being in the world and, of course, we are keen on engaging with her and having her insight about who we are. Can she teach us something about ourselves?’
What data sets are needed to make MARY convincing enough to be able to engage in conversation?
‘The algorithm needs to understand the basics of a conversation, so when someone asks a question, it will provide an answer. Then, we have one which is more philosophical in context, and then we have another to enable MARY to understand emotions and psychology. Then, because it is an ARTificial intelligence - a combination between ML and fiction - the persona can simulate high levels of intentionality and self-awareness, so it can appear to want something and have agency. This layer makes it seem like MARY is more alive than a computer really is.’
You mentioned about MARY harvesting emotional data. Due to the nature of the ML platform, will your artworks continue to evolve and become more intelligent?
‘We can train the algorithm with the data we get from the audience. However, when we harvest the emotional data from the audiences, we don't keep it. We just have a moment's analysis and then we let go of it.’
As technology continues to evolve and improve, where will your creativity take you next?
‘I think I'm heading towards a combination between ML and blockchain or a combination between ML and biotechnology. For the biotechnology, I'm interested in how, since ML is about intelligence, neurons and neural networks, it doesn’t have a body. But, by harvesting data about the human gene pool, could it eventually grow its own biological body? And what would an artificial body look like? I don't think it would look like us, but I think it might have the same capabilities we have in terms of senses and engaging with the world.
‘I'm also currently creating a ballet titled Centaur, with choreographer Pontus Lidberg. We use ML to design the choreography. Just as a centaur is a combination of a man and a horse, what we're interested in is how a human being (in this case, the choreographer) interacts with the ML algorithm to create the choreography and also how the dancers interact with the algorithm to perform the dance. The ballet features creative people working in synergy with a machine.
'My new ARTificial Intelligence, DAVID, will be the main character, together with 17 [human] ballet dancers. There's a psychological interaction and game going on between DAVID and the dancers, with DAVID deciding who will dance what and when - and in a sense the dancers will be performing for the machine.’
