Thoughts On DALL·E

I posted a little on Twitter about getting invited to the DALL·E beta, and over there I limited comments because I wanted a better avenue to expand on my complicated thoughts regarding this technology. DALL·E makes people angry and heated (particularly in my circles, for obvious and valid reasons) and I knew I could not properly entertain discourse about it there.

First, let’s get one thing out of the way: regardless of my complicated feelings, I am impressed: I do not know anyone who isn’t, even if they hate DALL·E. Frankly, if I see anyone just “pooh-pooh” the technology, I roll my eyes. It is obviously amazing:

And yet, my feelings about the technology existing at all are overwhelmingly negative. Let me elaborate.

Part of it is, naturally, existential saltiness: no one likes to feel impending replacement, and no matter what anyone else says, there are plenty, PLENTY of clients who would be perfectly satisfied with DALL·E as opposed to hiring someone and pay them a livable wage to make it for them. I still want to make a living at art. I legitimate feel that DALL·E endangers that. I don’t like that.

But my real biggest problem with DALL·E is its way of learning. I can almost guarantee that my many art pieces floating online have been part of what taught DALL·E, as has been the art of many of my colleagues, unwittingly, and without permission.

In such a case, and in one way of looking at this situation, you could conceivably argue that anyone who uses work from DALL·E commercially without paying any artist whose work was used some sort of revenue, is taking food from that artist’s mouth. And regardless, that artist should have been able to say no: do not use my work for this. I do not consent to this. Mosts artists I know would not want their work used in this manner by an AI.

AND YET…

There is a valid, different way of looking at this situation.

Every one of us who draws for a living learned by processing a large amount of artwork that was done by others. Sometimes referencing it, sometimes copying it directly for practice, and many more times by taking inspiration. We are literally taught to take elements we like from different styles and incorporate them into our own. Granted, we are not computers: I’ll get to that, so please, stay with me.

Ideally, the end result is something brand new and unique. But that end result did not appear from a vacuum: as conscious or unconscious inspiration, we used hundreds or thousands of different visual inputs to create our own style. And we did not ask any of the people who created those works we used for permission. If those artists demanded that of us, we would find it risible. They didn’t ask their own sources of inspiration. There might be the odd exception to this rule, but in most cases, we couldn’t ask for permission if we wanted to, especially since we take so much inspiration from the media we consume every day, put out by large companies.

It can rightly be said that DALL·E is learning in a way not unlike the way we learned, at least for an AI. I can see how someone could put forth the argument, “Why is DALL·E any different? Why should DALL·E not be able to learn in the same way, if other people do not ask their every source of influence over a lifetime for permission?” and I feel it should not be dismissed offhand.

At least in my opinion, the biggest reason is that no one person is going to replace a large percentage of an entire industry the way DALL·E could, nor can work or learn with the unfair speed or facility that DALL·E is capable of.

If even one person within my niche subcommunity copied my style down pat (something I’ve seen done to at least one of my friends) and started drawing stuff I would not want attached to my name/recognizable style, or severely undercut me in pricing, I would obviously have a serious problem with that. I do not think we are at all far from a future in which a client who finds me too expensive can input my work into DALL·E and ask it to create new works in my own style, even if they need some post processing. Who do I complain to then? I’m SOL.

There is an extra bitter, unfair edge to DALL·E: it learns very fast (as I mentioned above, I’m sure being able to emulate styles perfectly is in the near future, and since styles can’t be copyrighted, what happens then?) while using your own work without permission to do so. Even if DALL·E cannot create fully completed pieces ready to use, with a little bit of editing this is easily achieved.

Take as an example what I got with this prompt, “a cartoon baby dinosaur playing with blocks.” I’m sure the samples below will look very familiar to you, as it is a common clipart style that is available for sale, right now, and yet I got these for free, accidentally emulating said recognizable style:

You might tell me, ok, point the specific artist this DALL·E emulating here. And I can’t easily do that. But for a potential client, any of the above samples emulate that style well enough that they may want to just grab the above, slightly tweak it, and be done. They won’t pay a commercial artist if one click grants them an almost-ready image for free.

To me, any current wonkiness is completely moot. DALL·E will get past it, it is only a matter of time. For any person, artist or not, to feel that DALL·E will not revolutionize the industry, is shortsighted. It absolutely will. I feel that at the very least, DALL·E should have been an opt-in situation for artists. The ethics of the results of this technology where completely brushed aside. Once the technology exists, built upon the back of human work, done by artists who may not have wanted their work used to create their potential obsolescence, it is too late. And it already exists.

A part of me naturally gets excited by DALL·E and what I, as an artist, could do with it. I can use DALL·E to create things that I can build upon, so easily. I do not want to be shortsighted myself. Perhaps if this technology had been built by artists, or by people who aren’t infamous for being, shall we say, not very concerned about other human beings (such as Elon Musk) I would not feel so negative and hopeless about its existence.

I think it would have been possible for DALL·E to ethically exist. I also feel that ship has sailed. It is what it is. Anyone’s thoughts about DALL·E are, in my opinion, pointless at this stage. Whatever will be wrought by DALL·E is inevitable. I just hope it won’t devastate the industry as it becomes better and smarter.

Leave a Reply