What AI Thought a Coffee Table Looked Like

In the spring of 2025, while other people were using artificial intelligence to improve productivity, analyse markets or compose emails beginning "I hope this message finds you well," I asked it to make coffee tables.

Not merely one coffee table. That would have been eccentric but defensible. I asked for coffee tables in the style of the Renaissance, the Baroque, Impressionism, Cubism, Surrealism, Pop Art and several other movements through which civilisation had struggled before arriving at flat-pack furniture.

The experiment eventually became The AI Coffee Table Art Book: Explore the Creative "Mind" of AI. It remains one of my stranger projects, which is saying something.

The premise was simple. Each section introduced an artistic movement, explained some of its distinguishing features and showed a representative example. Then came four attempts by an AI image generator to create a coffee table in that style.

What would artificial intelligence do with Cubism? Would a Baroque coffee table require cherubs? Could Minimalism survive the addition of somewhere to leave the television remote?

The answers were frequently impressive, occasionally baffling and sometimes impressive precisely because they were baffling.

The Big Idea

Looking back from the summer of 2026, the project feels like a small archaeological artefact from an earlier stage of our relationship with AI.

In spring 2025, generating an elaborate image from a few sentences still felt faintly magical. You typed a request, waited, and watched something emerge that had not existed before. The hands might have had seven fingers, the furniture might have ignored several laws of physics, and any lettering usually looked as though the alphabet had suffered a minor electrical fault. But something had happened.

Now we are less easily astonished. AI-generated pictures have become part of the visual wallpaper. They appear in advertisements, presentations, social-media posts, book covers and places where somebody once might have hired an illustrator.

Familiarity has made us more demanding. That is probably healthy.

The question is no longer simply, "Can AI make an image?" It plainly can. The better questions are: Is the image interesting? Is it accurate? Does it communicate anything? Is it beautiful? And does it look as though a human being cared what happened?

My coffee-table experiment never proved that AI possessed an artistic mind. The quotation marks around the word "mind" in the title were doing some important work. What it did reveal was something about the relationship between machines, pictures and human expectations.

Ask AI for a Renaissance coffee table and it does not travel spiritually to Florence, study perspective under a master painter and develop a deep appreciation of humanism. It finds visual relationships. It associates Renaissance art with particular colours, compositions, interiors, costumes, textures and symbols. It then attempts to bring those associations together around an object usually used for magazines and neglected coasters.

Sometimes the result captures the style surprisingly well. Sometimes it produces the visual equivalent of someone attending a fancy-dress party having misunderstood the invitation.

That contrast became the most interesting part of the book.

The Confident Machine

AI image generators are extraordinarily good at producing confidence.

They rarely present a nervous first sketch and ask whether they are heading in the right direction. They arrive with something glossy, fully lit and apparently certain of itself.

This is also why they can be dangerous in less decorative areas of life. Plausibility and truth are not the same thing. A polished image can contain a nonsense object. A polished paragraph can contain a nonsense fact. A polished presentation can still be the work of somebody who began preparing it at 11.47 the previous evening.

AI is often most useful when treated neither as an oracle nor as an intern, but as a colleague with impossible confidence and no memory of where the mugs are kept.

The coffee tables made this visible. One could admire the colour, composition and ingenuity while still noticing that a drawer opened into another dimension.

That is a useful habit to cultivate beyond art. We can enjoy what AI produces without surrendering judgement to it. Curiosity and scepticism are not enemies. They are the two lenses through which emerging technology becomes genuinely useful.

But Was It Art?

The book also raised an older question: what makes something art?

Is art defined by the object, the intention behind it, the skill involved, the response of the viewer or the context in which it is displayed?

A generated picture has no childhood, no memories and no private longing to express itself through the medium of occasional furniture. Yet the human choosing the subject, refining the prompt, rejecting weaker results and arranging the finished sequence is making creative decisions.

Perhaps AI art is less like meeting a new painter and more like discovering a peculiar new instrument. The instrument can generate extraordinary sounds, but somebody still has to decide what is worth hearing.

This is one reason the book has found an appreciative audience. It is not simply a parade of unusual furniture. It gives people an accessible way into art history while allowing them to judge how convincingly a machine can reproduce the visible characteristics of different movements.

You can enjoy the experiment without becoming an expert. You may even emerge knowing why Cubism fractured objects into competing viewpoints—and why doing this to a coffee table might make it difficult to put down a cup.

A Slightly Weird Project, Reconsidered

In 2025, I thought I was testing what AI could do.

In retrospect, I was also testing what I did when faced with AI.

Would I accept the first result because it looked impressive? Would I notice the details? Could I explain why one image captured an artistic movement better than another? Was I responding to genuine visual qualities, or merely to the novelty of seeing a Rococo side table apparently designed for Marie Antoinette's broadband router?

The project encouraged looking rather than merely glancing.

That may be its most enduring value. We live among more images than any previous generation, but abundance does not automatically produce discernment. We scroll rapidly past pictures designed to seize our attention without asking how they work.

Art books slow the eye down. They teach us to notice composition, perspective, colour, symbolism and the choices concealed inside an apparently effortless image.

AI makes that education more important, not less.

The machine can now produce the picture. We still need to decide whether the picture deserves our attention.