Architects, A.I., Creativity & Copyright
How A.I. Will Steal All of Our Ideas, and Why it Probably Doesn’t Matter
In This Post:
How A.I. Will Steal All of Our Ideas, and Why it Probably Doesn’t Matter
A Word from the Government
Architecture as a Private Service/Public Art
The Mixed Blessing of Creating in Public
Design as Process, Not Product
As long as I've been covering the AI Space, Architects have wanted to talk about copyright issues. If you're one of those architects, then this post is for you. But if you're one of those architects who has been hoping that the U.S. Govt might have issued some kind of blanket protection over creative work, spoiler alert: they didn't. So maybe it's not for you. But you should read anyway because the decisions being made now about AI will affect all creatives for the rest of history.
The US Copyright office just released its first relatively meaningful statement on AI & Copyright. Just so I'm properly managing expectations, you should know that it's relatively meaningful. Meaning, they put some effort into it, and it doesn't say nothing. But I think a perusal is better for asking questions than finding answers. I’ll give you the highlights:
A Word from the Government
To be granted a copyright, there must be meaningful human creative input, and protection cannot be granted to works that are exclusively AI generated.
LOGIC: If this weren't the case, someone would inevitably flood the copyright office with AI generated works. Someone could train an AI to produce 10 million different house designs, and then copyright all of them, and then make a legitimate case that each one was 'original' in the sense that it had never been seen in the history of humankind before.
PROS: This keeps humans 'in the loop' and prevents opportunistic jackassery.
CONS: Leaves open the definition of what constitutes 'meaningful' human creative input - a concept that would be nigh impossible to figure out at present, and will certainly be evolving rapidly as AI plays a larger and larger role in our lives.
The mere act of writing prompts is usually not enough to claim copyright, regardless of what content it eventually produces.
LOGIC: A portion of the USCO's report was on the issue of so-called 'hallucinations', and the fact that even the most skilled prompters can generate results that they didn't intend. To argue 'I wrote the prompt, the AI created the thing, and therefore I created the thing' is only half right. The AI utilizes other inputs, besides your prompt, to create the thing. Therefore, however ‘creative’ the content, you can’t meaningfully claim that it is your creative content.
PROS: Maintains a connection between the skill of the creator and the validity of the copyright. If someone is good at writing prompts, they might be able to create good art, but the quality of art derives from their skill at the prompting, not the art.
CONS: Ignores the fact that all creative tools have their own kind of rebelliousness. The grain of a particular piece of wood determines how the sculptor interacts with it. The sculptor controls the chisel, but in a sense, so does the wood. It's more accurate to say that great art occurs when a master is able to negotiate the interplay between tool and medium. Through that light, AI seems no different than a paint brush, or a cello.
Using AI as part of a larger process shouldn't preclude copyright protection - you can achieve copyright protection on human-created works that uses some AI elements.
LOGIC: This one seems obvious, and imminently necessary. But also more controversial than it would seem at first blush. There are advocates for the idea that because AI is trained on the work of others, that it should be forbidden to copyright work produced with the assistance of AI. If it's not produced by the artist's own hand, exclusively, then it is somehow an act of plagiarism.
PROS: Demonstrates that the USCO recognizes that AI is an inevitable part of all human activity, including creative work, and falls into line with reality.
CONS: Again, the Devil will be in the details of what constitutes 'some' AI elements.
A fourth major conclusion took me a bit by surprise: that no new laws were needed at the moment in order to address AI copyright issues. I had initially thought that the USCO would leap at the chance to try and legislate our AI future. I'm glad they didn't though, as it would have been a farce. It also occurs to me that this is just a better way to legislate our AI future, because it will require the USCO to rule on individual challenges on a case by case basis, affording them better control over what ultimately becomes copyrightable, and what doesn't.
Architecture as a Private Service/Public Art
Unfortunately for architects, much of what we do has already been ruled 'uncopyrightable' in one way or another, which is why I've always been a bit perplexed at the architects who get antsy that AI is going to steal their creative work. It certainly will, and indeed already has. But that's not so much a function of AI, as it is the unfortunate case law that governs Architecture, specifically, as a creative act.
For one thing, architectural designs skitter along the boundary between copyrights and patents.
Copyrights concern “Artistic, literary, or intellectually created works, such as novels, music, movies, software code, photographs, and paintings that are original and exist in a tangible medium.” While patents “can include new and useful processes, machines, manufactures, compositions of matter as well as improvements to them.” It’s easy to imagine how architectural works can be both (and BTW, something can be both, as in, you can file a patent and a copyright on the same creation, it’s just rare). For this reason, we’re going to be using the terms a bit interchangeably from here on in, despite the fact that they are legally distinct. We’ll lump both patents and copyrights under the general umbrella of ‘legally protecting an original creative work.’
In both senses, protecting architectural designs is legally hard to do, for four reasons:
Reason No. 1: You Can't Protect Generically Sensible Approaches To Work, Even If They're Novel And Unique.
Every architectural design is, in some sense, unique, because every design occurs at a unique point in space and time. I could sit down right now and probably draw a house floor plan that has never been done in human history. But it would likely contain the kind of generic design decisions that all architects do, and make sense to non-architects regardless. The dining room would be adjacent to the kitchen. The powder room would be near the front door. The closets would be tall and vertical, rather than short and horizontal. So even if the specific dimensions of the plan are wholly unique, I can't claim legal protection over the plan, because it's built upon logics that are, well, generic. The possible degree of protection depends on how any given idea can be expressed.
Where an idea is unique enough that it only has a limited number of expressions, copyright suits tend to prevail. Where they have multitudinous expressions, they don't. 'I'm going to cover the exterior with bologna sandwiches' only has one expression - covering the house in bologna sandwiches. 'My interior floor plan is laid out so that occupants can naturally flow from one room to another through the course of their day' has an infinite number of expressions, so there's no idea there that merits legal protection. This makes it very difficult to copyright architectural works, because for whatever idea you had, there are likely other ways that other architects would express the same logic.
Reason No. 2: You Can't Protect Something That's in the Public Domain (e.g. Buildings)
Copyright protection is also typically not upheld in cases involving design features that are plainly visible to the public. If you design a house in the shape of an octagon, you cannot then copyright all buildings in the shapes of octagons. People can see it, take inspiration from it, and design their own octagon-shaped building. The same applies to interior spaces generally open to the public.
This is one of those legal ideas that makes sense because the alternative would instantly give rise to absurdity. Every possible style of design and architectural feature would have been patented long ago, and no good idea would be able to be replicated, without paying the original author money. Architecture is necessarily a public art – that’s the whole point. And if you put your ideas out in public, people are going to copy them.
Reason No. 3: To Secure Legal Protection Over an Idea, the Idea Must Have Economic Value
Granted, all good design has value. And hopefully, you get paid for your designs. But you can't secure copyright protection merely because your creation has some demonstrable value to you. You have to prove that the reuse of your work caused you economic value loss. This isn't particularly specific to copyright law - it affects all lawsuits: you have to have standing and prove that you actually lost something that can be restituted by some action of the court. If another architect decides to design their house, for their client, using cladding made from bologna sandwiches, you'd have to prove that had the architect not stolen your idea, that client might have been yours. That client wasn't necessarily interested in working with that architect, they just wanted the bologna sandwich exterior. Had the other architect respected your copyright, that client would have come to you, because you were the only one capable of doing a bologna sandwich house. Again, this is very hard to do as an architect, because clients make their decision about who to work with based on myriad factors, such as cost, reputation, or personal relationships. If you had had copyright protection, that client may have just as well decided to work with their original architect anyway and forgo the bologna sandwich exterior, in favor of whatever exterior their architect thought might be best.
Reason No. 4: You Can Only Protect the Unique Parts of a System, Not the System Itself.
In Copyright law, you can typically only copyright the unique or original elements of a thing, as opposed to the thing itself. So, if Toyota invents a radical new carburetor and installs it in their latest model cars, they can't copyright the entire car on the basis of the fact that the carburetor is an original invention. It is true that that is the only car with that carburetor, and is therefore unique and novel. But regardless, the courts typically find that you can only copyright or patent the part of the whole which is original. For an architect, if you design a radical new lighting system that everyone is clamoring for, you can patent that system. You cannot, however, patent the design of the overall building in which the lighting system occurs, even though that design is made unique by virtue of the unique lighting system.
The Mixed Blessing of Creating in Public
So how should architects evaluate the 'plagiarism' threat brought by AI? Isn't it the case that AI could train on my designs, and then some talentless competitor could merely prompt 'Design like that talented architect I know . . .'? Yes, and that will certainly happen. However, there are a couple of reasons I wouldn't lose sleep over it.
Reason A: You Want to Be Copied.
You don't want to lose fees, obviously. But as discussed above, just because someone is copying your work doesn't mean you're losing fees. When you publish a built work in a magazine, other, lesser architects will study it, copy it, and bring your methods into their own work. This is how all architecture has always worked. Lesser architects imitate the better ones. If AI is 'adopting' your designs and utilizing them in its neural network, congratulations. It means your design signature will likely be on all designs for the rest of time. You basically won architecture.
Reason B: You Probably Don't Have A Copyright Anyway
I think any creative can understand any other creative's impulse to protect their original works. But the reality, for architects, is that the copyright protection was never really there in the first place. If you're freaking out about how AI is going to steal your work, perhaps it is some kind of perverse comfort to know that anyone could have 'stolen' your work at any time. The 'Age of AI' doesn't really present new challenges for an architect in how to realize value from their own creative works. It just leaves us with the same old challenges.
Design as Process, Not Product
The most daunting of those challenges has always been ‘how do I get fee-paying clients to recognize the value of my designs?’ The solutions to that are beyond the scope of this piece, but I recommend checking on Down Detour Road for a primer. I’m a firm believer in the idea that great architecture is much more about process than product. When a client pays for an architect’s services, it’s the process that they’re paying for, even though it’s the product that they eventually utilize. However, I believe that architecture has historically shot itself in its own foot by adopting, and promoting, the idea that the end product is the only thing that matters.
We persist with genius-hero myths, wherein architects are inspired by birds, clouds, dirty limericks, etc. and their resulting creations are just things for the rest of us to admire at a distance. Their product is the only thing that can have value, because their process is impregnable.
The age of A.I. makes this doubly stupid. When an A.I. generates a building design, it utilizes a completely different process than the creative human mind, even where it might come up with a similar end product. It is the differentiation between those two processes that may yet preserve the value of human designers. However, that can’t happen as long as we continue to represent to the public that an architect’s process is some kind of divinely inspired sorcery. It is in fact a structured, disciplined process of human creation. And, not for nothing, but that ‘human’ element is a necessary legal ingredient for any copyright or patent protection an architect wants to pursue.