The Museum for Communication Berlin currently runs the New Realities. Stories of Art, AI & Work exhibition. I went to the opening on April, 25th.
The opening started with a chat with the curators Dr. Annabelle Hornung and Maren Burghard (the third curator, Stephanie Müller, could not attend the opening). The curators talked about how the exhibition came about and their motivations. I was very irritated by their decision to give the AI a name. To be fair I think it was the decision of the moderator, the other two quite often forgot to use the name. AI systems are tools. They are run by companies who give their systems product names. Why not use those? Giving the AI a human name suggests empathy which it does not have. It makes them approachable and likeable. They are merely tools that posess no intelligence. We already have an issue humanising our environment. We don't need to add tools that mimic human behaviour.
The exhibits were created in two phases: The first phase from the original exhibition in Nürnberg dating from nearly a year ago showed people and objects in the rainforest. For the second phase they used a text AI to generate prompts for an image generating AI to imagine the workplace. Some problems to do with labour rights and stereotypes were acknowledged. They asked the AI to come up with an AI union.
The pictures in the exhibition were very glossy, full of young and beautiful people. The rainforest pictures were impressive: nice light and saturated colours. I enjoyed the surreal look of the objects placed into the natural settings. The people placed into the settings did nothing for me.
I knew about the stereotypes amplified by the AI. Nevertheless, I found it shocking to see. CEOs are all male in suits, cleaning staff are all female, many non-white. The workplaces in the pictures are all very hip, the workers young and beautiful. In the end the pictures are all smooth. Clearly they are not real, but the world they come from is very much a Hollywood world. The AI is not programmed to be biased. The bias comes from the training material and reflects the hegemonic culture the material is taken from. It thus further amplifies and solidifies these biases.
However, I was more disturbed by the AI generated texts than the images. The sentences are smooth but totally meaningless. In the same way that the images are devoid of meaning. I suspect that, because text is an abstraction that needs to be decoded, we notice the lack of content more than with images that are directly perceived by a primary sense. While the generated images are uncanny the texts simply make no sense.
I really enjoyed the not quite right machines the AI generated. Apparently, it struggles with telephones with dials because the training material does not contain enough classic phones. Mobile phones are fine. The result is very surreal.
I found the AI generated digital worker union stuff perverse. Labour rights are fought for and won with blood by real people. The AI tools belong to the capitalists who use them to deskill work and concentrate power. I learned that click worker is no longer an acceptable term because it is demeaning the workers. They should be called digital workers instead. I reject that, it is a euphemism. All work is hard and takes skill. However, the AI machine takes away the skill of creating images, texts and any other creative output by utilising the work of many people who do not need to be skilled to analyse existing creative works. The AI owners extract the value. The process is the same as during the industrial revolution were the skilful weaver was replaced by machines and child labour to produce inferior cloth.
The struggle for a fairer world is not over. In fact it has entered a new hot phase. The technology is not inherently bad. However, the way it is forced upon us by the tech giants without any conversation of how we want to use it, what for, and under which conditions is the problem. The use of new technology is not inevitable and we can reject it.
The exhibition contains a section on concerns where various experts were asked for their opinions on AI. These voices are quite critical. However, I think this section was a little lost between all the images. At the end of the talk the curators were asked whether they thought AI was a threat or a chance. They both said it is a chance. Their answer comes from a position of privilege. Suggesting that the digital workers should just create a union to make working conditions better is rather naive. It disregards the power imbalance and the hard struggle required.
Two big issues I missed (I might not have noticed them) are to do with copyright violation and energy use.
Both training AI models and later on using them for inference takes huge amount of energy. All computer systems convert most of the electric energy into heat. This heat needs to be removed from the machines. Energy efficiency then depends on the location of the server farms and what is done with the waste heat. The ecological footprint needs to be considered when evaluating to use of AI methods.
The AI models are trained with huge amounts of material (tagged and annotated by the click/digital workers). The provenance of the training material is not clear. Much of the content is used without consent from the copyright holders. Suitable prompting can allow the reconstruction of the originals. What happens to the artist when an AI can faithfully reproduce the artist's style? Another related issue is plagiarism. AI generated material cannot be copyrighted. What happens when people who are expected to produce original materials, e.g. authors or researchers, use AIs to produce their work? Is that plagiarism? Is it OK use an AI to smooth out the language? What percentage of the work needs to be original? What about the ideas behind the writing?
What happens when we rely on AI systems to improve our own output or even create it? We have grown up without AI and had to work hard at school and later on to learn how to spell correctly, how to structure arguments and write them down or how to produce works of art. At the moment we recognise that something is not quite right with the AI output. But what happens in a generation or so when the use of AI as a tool is the norm? How will it shape our sense of aesthetics? I realise that this is probably the argument of an old man. We have already lost many skills to machines during the past few centuries. So maybe I indulge in some elitism when I fear for our intellectual endeavours.
We are already swamped by content and drowning in information. Who is going to consume all this auto-generated AI content? Maybe we will each get our own personalised content served in a virtual reality all of our own, or more traditionally as text, pictures or music. Personalised content will further isolate us and atomise our society since there would be fewer opportunities for shared experiences.
In the end, the exhibition was lacking. The curators showed off a new toy that produces shiny baubles. They touched on some of the issues but did not tackle the real issues. Maybe a critical investigation of the technology inevitably leads to the realisation that it is a very problematic tool.
The exhibition is running until the 15th September 2024 at the Museum for Communication Berlin.