This library is the future of the world

If you’ve done any image prompting with AI, it’s clear what this is. It was undoubtedly imagined by ChatGPT and then some expert craftsman were brought in to turn the computer’s vision into real life.

In theory, I like this kind of thing. It’s a killer app for improving your own home and a devastating turn of events for interior designers.

But seeing it in real life is bothersome. The aesthetic is a mess of some kind of machine’s sense of harmony, not a human’s normal use of space. It’s a prompt of ‘balanced timelessness’ with the kind of soft lighting is too cinematic and typical of AI images.

But what frightens me is that it doesn’t work. It’s as if Sam Altman produced the image, gave it to the workers and said ‘do it exactly like this’ without anyone pushing back. For one, the lighting doesn’t work for a library. There is modest natural light from the window but at night it won’t be bright enough to read anything in most of that room, particularly those comfy-looking black sofas. Secondly, the books look like props. It doesn’t look like a real library where the shelves are full and people actually read the books and instead looks like they were chosen for their sizes and colors of their bindings.

And critically, notice the plants on the top shelves. Now that aesthetic may look good in a photo but in real life, plants need light. There are no plants on earth that can survive in those spots. Moreover, the person who posted the image said they were “mostly cacti but it’s true they did look a lil wilted last night”. So while a cactus might need less climbing up that ladder to water, it would need hours of direct sunlight daily. Even the snake plant on the floor (a notoriously indestructible house plant) probably won’t make it.

Why did no one look at those cacti and say ‘Wait, won’t these die?”

This is a preview of the world to come. It looks pretty but the information in it is hollow. The image is beautiful, but it won’t sustain the life it was designed for. In short, it’s a big sales job, not something that fixes a real-world problem.

The machine will tell us what’s best and we have to live with the consequences. It amazes and scares me that the people at OpenAI are such slaves to these machines that they built it without seeing the obvious flaws.

The person who posted this image works in ‘human data’ at OpenAI. These are the people who should be most aware of its pitfalls, like its tendency to be “confidently wrong” or to optimize for appearance over function, yet they replicate those exact pitfalls in the real world. What does it say about their ability to control or guide it?

This isn’t what alignment looks like.