AI needs us

Jonathan Maas ,

The German artist Mario Klingemann uses artificial intelligence in his work, and is considered a pioneer in the field. ‘AI art is made for people, not for other computers,’ he says.

Memories of Passersby I

‘The best thing I can do is get out of bed every year.’ This remark was dreamt up not by an ironic joker, but by GPT-2, a text generator based on AI and fed with 60,000 famous and appealing quotes. The system scrambled and processed these itself, discovered patterns in them, and thus generated new material. 

This recent work by the German AI artist Mario Klingemann – exhibited in Madrid (which is in lockdown at the time of writing because of the coronavirus) – takes the form of all-new, generated phrases on an analogue letterboard of the type once used in railway stations and airports. Kneel on the kneeler in front of it, and you’ll see a sentence appear. 

‘The visitor usually experiences this as a personal message,' Klingemann explains when we speak to him via Skype. ‘They give the sentence meaning: that's exactly what I care about'. The work, named Appropriate Response, is Klingemann's reply to the attention economy, in which people, sometimes rather hysterically, attach meaning to posts on Twitter or other online platforms.

Portraits

Klingemann was born in 1970, is considered a pioneer in the use of AI in the arts. He is self taught, and became interested both in art and technology at a young age. When he chose which educational path to follow at the end of the eighties, there was no study in which both his passions came together. He first started working as a graphic designer, then became a full-time artist, always using computers and algorithms.

Appropriate response

After the rise of artificial intelligence. Klingemann created his best-known work, Memories of Passersby I, an autonomous AI installation auctioned at Sotheby's for $42,000 in 2018. It consists of two screens showing distorted portraits of a woman and a man, produced without human interference, apart from the training software used. The work employs a system of neural networks and self-learning software, trained using thousands of digitised portrait paintings from the seventeenth, eighteenth and nineteenth centuries.

The composition and sequence of the new portraits it generates are entirely determined by the computer using machine learning. ‘With ordinary computer-generated art, you need a program to get the computer to do something,' Klingemann explains. ‘You give it rules within which it carries out its tasks. You can enter a bit of randomness so it surprises you, but that's it. With machine learning, you don't give the computer rules: it searches for patterns based on the data you feed it and learns from that.’

Clouds

What moves an artist to hand over the creation of artistic work to a computer? ‘I'm addicted to surprise,’ says Klingemann. ‘I want to see new things all the time. My own imagination is too limited, so I work with an extra power source: computers and AI. They generate new images and possibilities that I’d otherwise have no idea of. As humans, we can only imagine things based on what we’ve seen before.

'People need certainties and want to categorize everything. Computers are more fluid.'

Mario Klingemann

We can’t conceive of anything beyond our own system and perception. You recognize faces based on static characteristics, such as eyes, noses, and ears. But sometimes you also see faces in the clouds, because you encounter the same patterns. Computers are even better at that: recognizing patterns based on information that you feed them. They make connections that you will never be able to make.’

AI may be a source of inspiration, but Memories of Passersby I takes this one step further: AI generates not only new ideas, but also the end product itself, without a human artist as curator. Is man made redundant here? No, Klingemann says reassuringly, it takes a person to provide the work with meaning. ‘The machine itself doesn’t know what it’s doing. It needs a receiving party to determine whether it has meaning or not. I see the output of AI as a seed that needs soil to become a flower. 

Mario Klingemann

‘AI art is made for people, not other computers. You can imagine a closed system of only AI, but what it produces will be incomprehensible to us. You can already see that in music: AI music without human rules such as harmony, melody and rhythm produces random noise. That's a different universe.’

Although Memories of Passersby I continuously generates new portraits, the system is not infinite. Without new input from outside, it will at some point stop making new combinations. ‘AI needs us,’ Klingemann says. ‘It feeds on us.’

Mortality

Yet people fear that technology will make them unnecessary. This is not an entirely unfounded notion, because computers are in many cases smarter, faster and more efficient – a lot of human labour has already been, or will be, made redundant by tech.

Creative and artistic work, on the other hand, is a human domain par excellence, because it’s original and inspired, not mechanical. Klingemann proves that AI can be creative and original, and can therefore do something that people would never have thought of. But it doesn’t know what it has made, and cannot judge its own work without consciousness. Right?

‘Well, how do I judge my own work?’ Klingemann sighs. ‘I determine standards myself and decide whether something is interesting or original. But if I’m to do this, what I’ve created must differ from what I already know. It’s a personal consideration, where everyone will come to a different conclusion. A computer has no consciousness, and we believe that people do. But computers can find the patterns on the basis of which we form judgments. That’s why they often know better than we do what we like and what we think is beautiful. And they’re also increasingly able to recognize our emotions.

But people are still superior when it comes to integrating different and complex information flows simultaneously, including sensory data that a computer doesn't have. On the other hand, what I learn from AI is that everything in that universe is multidimensional. There are no hard definitions there anymore, and that’s something we humans need. We need certainties, and we want to categorize everything. Computers are more fluid, and tie different patterns and categories together. In the art world, AI is opening a door to new possibilities. 

Some take that door, others don't. I see AI as an impulse to change things within the art world, and for the time being it’s a tool I work with. Art must remain human. If something is completely machine based, it might be interesting as novelty, but it doesn't appeal to us in a sustainable way.’

In Klingemann's view, the biggest difference between man and machine is that people have a certain motivation to do something. Machines don’t; they’re slaves. ‘Our greatest motivation is our mortality. The only motivation that a machine might have is finding a way to power itself without needing us. I don't see that happening anytime soon.’

This was the first interview in a four-part series about AI and art. Next week, part 2: writer Ross Goodwin, who uses AI to produce books and poems.