THEFUTURE
No time for Schadenfreude
In this issue: The left thinks AI is a scam, the right is building with it, and Europe is asking the better question. Plus: the human in the loop might just be slowing things down, Google wants to make you a musician, and which AI should you actually be using right now.
What we’re talking about: Is the left missing out on AI? Ranting on Bluesky about intellectual theft, while everyone else and especially the right is prompting, building, achieving?
That’s what writer Dan Kagan-Kans argues in a post on Transformer. The right, accelerationists, industrialists, the Trump administration, is setting the tone. They don’t have the better policies (they really don’t), but they’re first to the debate. The left, he writes, is blind to what AI could do: make the state more capable in delivering a good life.
If you want to read a takedown, Gita Jackson, co-owner of the collective news site Aftermath, has a thorough one. She states that the left doesn’t hate tech per se, only this one, because it enriches the few. We are promised liberation and get exploitation. (She makes good points while he has more of a vibe and fewer footnotes.)
What’s missing from this debate is Europe, where the tone is very different. Brussels has taken on AI and big tech from the point of regulation, and whatever you think of the results, it’s asking the right question: What is this technology actually good for, and do we want it that way?
Extended overthinking: Designer and illustrator Pablo Stanley writes about the growing disconnect he feels between himself and his creative work as AI tools take over. He compares the experience to a slot machine you always win at: addictive, instant, and hollow. Then there is this:
“The ‘human in the loop’ feels less like the guide and more like the thing slowing it all down. Some people call this slowing of the machine… this fine-tuning… ‘taste.’ The ultimate human moat. Guys, stop it with your ‘taste.’ I’ve seen how you dress. You don’t have any. And taste is really just doing things the way the top 10% of creators do instead of the average. That’s a pattern. Patterns are exactly what these things are built to learn… just give it time (a month?).”
Read the whole thing. It won’t make you feel better, but it will make you feel seen.
What else I’ve been reading:
And now: It happened just after Germany’s AI for Media Network met in Hamburg to talk about AI deepfakes and the importance of rigorous verification. Public broadcaster ZDF showed an AI-generated video In a news segment on immigration enforcement in the U.S. The error was quickly caught because of the visible OpenAI Sora logo in the video.
What followed was mockery, consequences, think pieces, all of it. But there’s really no time for Schadenfreude. And while I’m not too optimistic we’ll be able to detect every deepfake in the future, Anika Gruner is actually working on a solution with her start-up Neuramancer.
Three Questions with Anika Gruner
Anika Gruner co-founder of Neuramancer, a deepfake detection start-up.
What's the most important question right now?
I believe it is important not to be afraid to design change “on the drawing board.” When I started at public broadcaster BR in the new position of “trimedial production coordinator,” (way back in 2017) there was certainly a lot of uncertainty among my long-established colleagues. You have to get everyone on board so that they are open to new ideas. Processes need to become more agile and you have to adapt to new things more quickly. Ultimately after two years, the position was considered successful by everyone, editors and management alike, and was also continued at other BR locations. If the attempt had not been made and the confidence to simply try it out had not been there, it would not have been possible to dispel the concerns.
The recent ZDF scandal also shows that certain processes need to be adapted more quickly to current requirements. Perhaps an “innovation scout” would be a good start, not only in the editorial office but also in smaller local editorial offices. Someone who attends AI conferences or media innovation camps and takes away the concerns of all the journalists and media employees out there.
What's one fact about AI that everyone should know?
What it can do, but above all what it cannot do.
There are so many dynamics that users are still unaware of. For example, catastrophic forgetting. If you look at how AI learns, you understand much better where its limits are – and in my opinion, knowing these limits is much more important than finding the perfect prompt.
What's a good hobby to pick up?
My co-founder Anatol Maier and I have developed our new hobby ‘AI Slop Hunting.’ Whenever we are in a furniture store or department store, we ‘hunt’ for objects that have visible AI slop printed on them instead of neat graphics. Sometimes it’s hard to believe, but AI slop is already everywhere. As a deepfake detection start-up, it is very important to us that images from generative AI do not simply flood everything, but are only used sensibly and qualitatively. By the way, you can test your own AI recognition skills on our website wefightfakes.org (for non-German speakers, simply click on ‘play’). We can’t wait to collect more AI Slop photos from everywhere.
Hands on: Raise your hand if you’ve sat through an AI-generated song in a meeting and wanted to leave by the second chorus. It was maybe funny for ten seconds, it isn’t anymore, and it’s not going away. At least this time there’s a 30-second limit: “Create music” is a new option in Gemini that uses Google’s Lyria 3 music generator.
let's make a song - it's not pop, more artsy, jazz, experimental. first, static from a radio. someone searching for a frequency. then there is one - a monotone voice repeating numbers, in russian, calm, but urgent. nothing else.Somehow I think Lyria was not trained on Russian number stations. Just maybe it listened to itself during the training process?
let's do an ambient noise track, a data center humming, no beats, no buildup, no drama, just hummingIt’s beats and no humming. One more, this time with instructions straight out of Wikipedia:
let's do a new track. experimental electronic music in the style of Ryoji Ikeda and Raster-Noton label. sound in a variety of "raw" states, such as sine tones and noise, often using frequencies at the edges of the range of human hearing. Rhythmically, Ikeda's music is highly imaginative, exploiting beat patterns and, at times, using a variety of discrete tones and noise to create the semblance of a drum machine.With slop like this, looks like experimental music and niche nerds are safe. At least for now.
You can just build things: I relaunched my website, and I can’t begin to describe how much fun it was. Dark mode, a slider plugin, a bookshelf, easter eggs. All things I’d been putting off for years because I feared the time it would take. Now I just willed them into existence, going back and forth with Google Antigravity, Claude Code, and OpenAI Codex. Could the code be cleaner? A thousand times yes. But it works, it’s mine, and I love it.
One more thing: AntiRender is a website to upload an architectural render, the brightly lit, green tree filled imaginations of what is about to come. Get back what it’ll actually look like on a random Tuesday in November. No sunshine. No happy families. No impossibly green trees. Just cold, honest, depressing reality.
This is THEFUTURE.
