What a title, eh?
Initially, this was supposed to be a short essay for my Gemini capsule (which has nothing to do with the product Gemini, which is, afaik, an LLM), but then I thought, well, why not write this on my after-hours website on tumblr (I need to create a sort of after-hours schedule for my posts, right after having written this).
My problem with AI is very simple: if you were around when people were looking for baby birds but what they got were not bird chicks, but some automatically generated horror, you already know what my problem with AI is.
I mean, Jesus, right?
When I am looking for something and using a search engine, an LLM, a link-list, whatever it is in 2026, I still expect what I find to fall into two categories: to be human-made and to be somewhat real.
Let me start with the second part, the somewhat real.
I am a graphic designer, and I have developed a bunch of photographs in my life. Took even more than I ended up developing. Storage is cheap, I rarely shoot in analog, so I basically submachinegun it.
Whatever happens to be close enough to a composition, gets cropped, thirded, golden meant, the works.
And then comes the editing: make this highlight less prominent, because it makes the eye jump between the highlight and the photo’s focal point, distort again whatever the lens distorted, use arcane midi-inputs to color-adjust the whole thing, and what I end up with is a derivative of the truth.
The output is somewhat real. Like, as if you have taken the reality of the light that day, the reality of the lens that day (some photographers swear that some lenses are more fickle than others regarding temperatures and humidity), and built something from that, as if it were clay. I think. It is hard to describe.
You develop a photo. You don’t shoot the outcome. You shoot the raw material you then end up using.
I think that this explains somewhat what I mean with somewhat real.
And I expect what I find to be human-made.
Chances are, that I have seen your work, that I am aware of it, and that I have questions, for later. Questions which are typical of my profession: how did you do it? What was the idea? Why did you publish this and not something else?
And if you are a human, I can ask those questions. And I most likely can get an answer. The answer might take months to arrive, but, you know, something is moving around somewhere.
But generative AI allows for something else; not only is there no process involved in the outcome which is in medias res with the outcome, it allows for this scenario, in theory:
to generate an output the moment it is requested, i.e. to make ready-made a reality of a search engine hit.
And when this becomes possible, you will never see a website again, other than the one you are currently using to search with.
Indeed, the search engine then becomes a finding engine: it finds what you are looking for, despite it not having existed before. Your search creates the outcome.
In German, this is almost a pun: the verbs finden for finding, and erfinden for making things up are very close to each other.
We are currently in a state where websites still host the ai slop because websites still run on the idea that some sort of trickle-down economy will certainly keep the lights on as long as you can show people ads and allow for their tracking by snoopy corporations, but I fear that this will be a thing of the past.
Why would a search engine pay someone else for showing their ads when they could conveniently cut out the middle man altogether, generate a user’s answers on the fly, generate better baby peacocks the moment they are requested, and still show them ads which other corporations then have to pay for for being hosted amidst a user’s found/generated content?
And that is my problem with AI.
Quelle: Original auf Code & Canvas