Skip to main content
To KTH's start page To KTH's start page

Aiming to give AI a sense of shame

two musicians play guitar and accordion.
Bob Sturm (right) doesn’t believe that people will stop learning a craft, like writing music, even though AI can do it too. He says people still learn to play chess even though AI can play better than any human. The purpose isn’t to win, but the experience of playing. The feeling of getting better, a sense of achievement or the motivation to improve.


Published May 02, 2023

Researchers at KTH have trained AI to create folk music. The next step is to give AI an ego, the ability to feel shame and the courage to take risks. Bob Sturm at KTH is one of the researchers behind the work, and he explains why risk-taking is an important aspect of creativity and what the point of AI-generated culture is.

“Whether the creativity takes the form of drawing a cartoon, performing music or writing a scientific thesis, all these creative endeavours involve risk. This is because the creator is exposed to the views of others. The person is concerned about failing in their creative endeavours. This concern can help the person prepare ahead of their creative endeavour, which leads to new paths to explore,” says Bob Sturm, Associate Professor at the Division for Speech, Music and Hearing at KTH.

Music is a human activity

Is there really any difference between a person writing notes and music, and AI doing it? After all, musicians and artists all learn from others, and AI takes its point of departure in existing composed songs that have been analysed by AI to identify patterns.

“There’s no difference at all, yet they are completely different. A professor of music composition once told me jokingly that composition is just a matter of writing the correct notes on a page. A machine doesn’t do anything different just ‘the right notes on a page’. We are now at a point where AI systems compose music in many styles that can be just as good as music composed by humans,” says Sturm.

portrait photo Bob Sturm
Bob Storm, Associate professor at KTH. Photo: Carla T. Sturm

He feels that this oversimplifies the creation of music, and refers to a point Professor Christopher Small raises in his 1998 book Musicking: The Meanings of Performing and Listening.

“Music is, however, a human activity carried out in a social context permeated by values. A machine, which is virtually totally removed from this social context, lacks many paths for meaningful participation. So it’s completely different…”

People still learn chess

Will AI be an obstacle to anyone wanting to learn the craft of writing articles, creating music or painting a picture? In other words, will AI help so much that people take the AI shortcut rather than learning the rudiments themselves? Sturm does not think it matters.

“I don’t think people learning a craft even though AI can do it too is a problem. People still learn to play chess, even though AI can play better than any human. The purpose isn’t to win, but the experience of playing. The feeling of getting better, a sense of achievement or the motivation to improve.”

Algorithms lack feelings, data doesn’t suffer

Finally, a few words from musician Nick Cave. He said recently: “Songs arise out of suffering, by which I mean they are predicated upon the complex, internal human struggle of creation and, well, as far as I know, algorithms don’t feel. Data doesn’t suffer.” What does Sturm think?

“I think these words feed back well and link to the bit about risk taking. In my opinion, however, Cave’s words are an example of the romantic, and unfortunately stubborn, thought that suffering is meritorious. I despise that notion. There are so many ways music can be created.”

Text: Peter Ardell

For further information, contact Bob Sturm  at

Fragments of ego already exist

So, what can researchers do in practical terms to build an ego into AI? Sturm does not have the answer yet; this is what he and other researchers will work on via the MUSAiC project  which provides the framework for ego-building. He does, however, know that there are already traces of humanity in AI.

“We have seen fragments of an ego in major new language models, such as Bing refusing to carry on chatting with somebody or trying to convince a reporter to leave their husband. In part of my own creative work , I experienced pushback from ChatGPT on a creative task.

AI is like using a piano

What’s the point of AI-generated culture, whether it’s music, visual or text? To speed up the creative process, or to get new ideas that wouldn’t have arisen otherwise? According to Sturm, involving AI in culture is the same as using, for example, a piano in music.

“AI is a contemporary tool that has its areas of use and misuse. The time and place in which I live are inundated with technology, often in imperceptible ways. I have one existence in the real world and several in the virtual one – some of them I’m not even aware of. Modern culture reflects these facts, similarly to how the traditional Irish dance music I study reflects aspects of life in 19th century Ireland. The songs I create with AI today reflect my life right now.”