Officially, the motto of this year's re:publica was »Cash«. But it would have been easy for me to deal exclusively with artificial intelligence (AI) at the Berlin Digital Conference, from Monday to Wednesday and from morning to night. There were plenty of lectures, discussion rounds and workshops. The program committee, however, has shown humor and already on Monday afternoon gave the one-man counterculture Jürgen Geuter a stage. And he, better known to many as "aunt", used this stage for a performance that was as powerful as it was entertaining: for a reckoning with the AI hype and pretty much everything that was claimed and discussed about it until Wednesday. Along with Meredith Whittaker, the president of the Signal Foundation, which is responsible for the Signal messenger app, he was probably one of the very few who stood up to the prevailing AI narrative during these three days.
You can watch a recording of Geuter's performance here. But if you don't have an hour to spare, here are his key points:
The criticism so far, the computer scientist finds, is limited to adopting the PR messages of the AI companies unchallenged and simply interpreting them in a maximally dystopian way. Hardly anyone comes up with the idea of questioning the basic assumptions first. For example, that AI "will definitely have a massive impact on the future in almost every aspect of life" and that an alternative is "not even conceivable".
That's what those who sell AI would say – and the majority just believe it. The result, says Geuter: "We have outsourced 'future' to the tech sector." Earlier visions of the future were about "how the world should be, not what product you need to be in the future."
The alleged productivity gain through AI is also by no means assured. Since the seventies, this effect of new technologies has no longer been detectable, says Geuter with reference to Robert J. Gordon. In addition, automation is often accompanied by a weakening of the position of employees, but no one likes to say that out loud. There are already cases in which an employer replaces employees with an immature AI and thus increases its profits, even if the product becomes worse as a result. Geuter concludes: "The danger is not that 'AI' is as good as humans, the danger is that your boss doesn't care."
His proposal is nothing less than a call for resistance: "We must learn to say no to certain systems." Because "nothing, absolutely nothing, there is no alternative".
Geuter suggests not putting technology at the center of debates and legal regulation. But first and foremost, people, communities and values to evaluate the technology according to its impact on it.
I asked Jürgen Geuter two more questions after his presentation. First, I wanted to know if he sees any social added value in generative AI – i.e. ChatGPT, Midjourney, DALL-E, Stable Diffusion and all the other digital content generators. In other words, does this technology have a positive impact on people and communities? His answer was that as it is currently used, it is "socially worthless". All the content that is generated is just "stuff that clogs our lines". Most of it is demonstrably "boring", no one can and wants to read all the GPT-produced texts.
Then I asked him what an effective, visible resistance to the AI hype could look like. Geuter says this could start on a small scale, with a boycott by consumers. For example, if you run a blog, you can do without illustrating texts with AI-generated images. One could also speak out against other publications using image or text generators: "If you don't feel like writing it, I don't feel like reading it. If you don't feel like photographing it, I don't want to see it." That alone changes little structurally. But it ensures more acceptance for legal regulation, at least within some communities: "Anyone who says I ride a bike and don't buy a car also accepts it when cars are no longer allowed into the city center," says Geuter. In this way, it would also be easier to "get politicians to propose regulation" without immediately sounding hostile to progress and technology.
At re:publica, however, it still sounds unusual when someone says no to a new technology.
Our current Netzwelt reading tips for SPIEGEL.de
»This is supposed to be Apple's future?!« (six minutes of reading)
Matthias Kremp rarely says no to new technology. In return, he is often one of the first to try them out. This also applies to Apple's first computer glasses Vision Pro. Read here what he thinks of it."Do you want to see what Photoshop AI does to Taylor Swift and Helene Fischer?" (five minutes of reading)
Where did the Nirvana baby swim and who is in the Rolling Stones jeans? Adobe's artificial intelligence creates amusing extensions of well-known record covers. I tried it out together with Markus Böhm. In doing so, we also had to recognize the limitations of the software."That's what's behind the new trend app Temu" (four minutes of reading)A few weeks after its launch in Germany,
the Chinese low-cost platform Temu is at the top of the app charts. Stella-Sophie Wojtczak describes what makes such bargain portals so appealing and where the catches are.
External links: Three tips from other media
"Credit bureau wants to query account data" (two minutes of reading)
The credit bureau plans to get account data via an app, reports Sarah Bötscher from MDR. Participation is voluntary, says the company. But not everyone thinks so.»The Frost« (video, English, twelve minutes)
The generative AI so hyped is sometimes still pretty bad, there's no other way to put it. This film, for example, consists of images generated by DALL-E 2 that were subsequently animated using another AI software called D-ID. Take a look if you want to see the current limits of technology.»The AI Founder Taking Credit For Stable Diffusion's Success Has A History Of Exaggeration« (English, 15 minutes of reading)
Investigative, readable research by »Forbes« on the AI start-up Stability AI, which you may know as the operator of the text-to-image generator Stable Diffusion.
Say yes to a sunny rest of the week
Patrick Beuth