Future Magic (II. Artistic Struggle)

(Note: this essay is part of a series.)

One of the main objections – and it is very understandable – people have to AI media is that it “will” eliminate human endeavour and creative struggle. We tend to value these things, whether for moral reasons or simply because we don’t like feeling we have been “ripped off”. AI imagery, being generated not by suffering humans but by impervious software, feels like a trick, especially at this rudimentary stage where human involvement in the process is still minimal. So, in case my quasi defence of AI imaging seems flippant or shallow, I want to state from the outset that I do appreciate the moral value of creating something with one’s own hands, painstakingly.

When people watch my videos, they are seeing the product of one man to a degree that none of them really know, because I have never described the processes involved in my work. They are seeing images I created myself, often using software that I wrote myself. They are hearing sounds I synthesised myself, often using software that I wrote myself. They are hearing sounds filtered, volumed, panned and mixed using software that I wrote myself. My code handles audio at the level of individual bytes. I wrote that, and it took time and energy to learn what to do and then to do it, and keep working on it until it functioned properly, and then to improve it to make it more streamlined and efficient. The point is… I really do understand the value of being intimately and exhaustively involved with the totality of something that one creates - to have shaped the macro by governing the micro. I can honestly say that I understand this better than anyone else I have ever known who worked in media creation.

I could describe “how far down” the technical ladder my knowledge reaches, but it would be extremely tedious for you and ultimately boastful - and I’m not even sure it’s something one should boast about. It’s like a craftsman boasting that he made his own tools, and made the tools with which he made the tools, and gathered the raw materials with which he made those tools, etc. You have to admire his dedication, versatility and self-sufficiency, but you can’t help wondering whether he should have just gone to a shop and bought the tools he needed so that he could concentrate on the challenge at hand. Why make it all so much more onerous for himself?

The answer is, of course, that he values his dedication, versatility and self-sufficiency. He wants to know that he can overcome every little challenge, that he is developed enough as a man to be able, entirely on his own, to deal with the raw elements of the world and reconfigure them into some novel thing of function and beauty. I admire the man who is like that, who literally needs nobody else because of his knowledge and ingenuity.

But at the same time… in our age it is insane to operate in this way. Sophisticated tools exist for a reason, and it would be unfair to describe that reason as “laziness”. There is a limit to how much effort a human can be expected to put in to getting a task done. If Vangelis had eschewed the Yamaha CS80 (perhaps the most sophisticated synthesiser available in 1982) and ascetically confined himself to the primitive tools available to Delia Derbyshire in 1963, we would simply not have the Blade Runner soundtrack. We can should appreciate Derbyshire’s thoroughness and understanding “at a low level”, but we should also appreciate the majesty and complexity of what Vangelis achieved “at a high level”.

(Note also, Derbyshire did not compose the Doctor Who theme. Ron Grainer composed it and someone else had to actually “perform” it since the tools of the time were so primitive as to require expertise that only a highly-skilled technician would have. Two decades later, because of much more sophisticated tools, Vangelis could both compose and perform the music himself.)

It is the same with any task and any creator. In truth, though I wrote the code, I did not create the programming language in which I wrote the code. I did not design or build the computer that ran it. I did not mine the raw materials with which the computer’s parts were manufactured, nor conceive the processes or design the machinery that fashioned those raw materials into a CPU, a RAM stick, a hard drive, etc. I don’t even know the principles of metallurgy or electricity. In fact, I don’t really know how a computer works at all. I only know how to use it. For the jobs I set myself, it is the right tool to use.

“Use the right tool for the job” is a way, not merely to avoid labour, but to free yourself to focus on the labour that matters, which is the labour that only you can do. Other people can manufacture the oils that enable an artist to paint a masterpiece; he does not need to manufacture the oils himself. He does not need to weave together the canvas and grow the hemp for it, nor construct his own easel and chop the wood for it. If he does all of these things, that surely adds to his achievement in a technical and perhaps a moral sense, but in taking the time to do this unnecessary work he is depriving himself of time and energy to produce a second masterpiece. Other people could always have manufactured those oils for him; what the world required of him was not the oils but the painting, however it be produced.

At the same time, we would feel absolute disgust if it turned out that he did not really paint the masterpiece himself but had a robot do it on his behalf according to some scant instructions from him. We would feel incensed by his claim to artistry if he offered the defence that, when the robot had finished, he had intervened and said to change some particular detail of the image. That he would consider this to be adequate involvement in the creative process, and thus proof that he is a genuine artist and not a loathsome fraud, would incite the rage of any decent person.

In my opinion, this is the real problem with AI “art” today: not that it does a lot of labour for the creator, but that it limits his involvement in the creative process to a brief instruction which it then responds to with a huge degree of randomness. The creator is no longer in control; in fact he has become little more than a muse for the software, which has become the creator and relies on the human to give it ideas to work with. The result is not art. It is entertainment, and no doubt the masses will derive great amusement from using it, but the artist should not degrade his role by thinking “this is it”.

I believe that a necessary component of art is some degree of struggle. The act of creation should not be cheap or painless. It should cost. I don’t know exactly why this is so. But if you learned that Leonardo da Vinci painted the Mona Lisa in an hour, and didn’t really care how it turned out, or if you heard he got a robot or AI software to do it for him… would you not feel cheated? If you wouldn’t, I think you should. I am absolutely certain of it. Art should be the result of struggle. If there is no struggle, it is not art. This is an intuition of mine - as I say, I cannot explain it except with reference to the above descriptions of skill-gathering, talent-honing, time, energy and commitment - but I am absolutely sure that the intuition is correct. Part of being human is caring about what we do; if the thing comes too easily, we have not proved that we care about it, so the thing does not embody a volume of time and care that went into its creation, so the thing is valueless - at least, as an embodiment of human achievement, because it simply is not that.

The threat of AI is that it degrades humanity by devaluing all future art by negating all struggle by making all effort unnecessary by doing everything for “the artist”.

I see two possible solutions to this.

The first solution is that we simply don’t use AI. Frankly, that is just not realistic. People are going to use it, regardless of the arguments. I think, soon, the products of AI imaging will be more ubiquitous than those of Photoshop. Eventually, they will be more ubiquitous than the products of digital cameras, including all of those embedded in smartphones. The default assumed origin of any image, photorealistic or not, will be AI. It doesn’t matter whether you or I like this. It is simply unstoppable, short of a global government ban on AI visualisation software.

The second solution is for us, as humans, to rise to the occasion that AI makes possible.

Those who abhor AI art do so because they believe that artistic achievement, to be real, must involve challenges even though those challenges have been technologically overcome and need no longer be undertaken. Imagine a great artist who refused to paint on the basis that he - whether through lack of knowledge or opportunity - could not manufacture his own oils or his own canvas or his own easel. Imagine a great programmer who refused to program unless he could build his own computer. How would we regard such obstinacy? Perhaps as self-defeating, indulgent, misguided, even insane. Charitably, we could see it as naive idealism. But the result is the same: the program is not written, the painting is not painted. The artist goes unaccomplished, because of a moral objection that the work, in not being arduous enough, would be a lie.

Of course, you could say that in such a scenario the painting can still be painted: we must simply allow the artist to manufacture his own oils. But in truth, few if any artists in history ever made such a stipulation. They simply obtained the materials they needed for the challenge that was theirs, and then did what the world required of them, what their own nature demanded of them. They could do so because other people took care of the mundane tasks for them.

In our time, the question is whether to allow AI to take care of mundane tasks involved in creating a work because it enables us to create that work. If we choose not to, then the work will remain uncreated. This means the world will be without that work, and the artist will remain without accomplishment, and he will suffer the pain of having never expressed that thing which is within him. Of course, this is not the end of the world. Throughout history, countless talented men must have lived with that pain. The point is, AI means that a great number of them need no longer do so.

Because of the Yamaha CS80, Vangelis could create the Blade Runner soundtrack. Because of AI, many films could now exist that otherwise would remain pipe dreams, unseen by the entire world.

I think what people find demonic about AI imagery today is the deception that is inherent in the way it currently operates. The resulting image looks like it was created painstakingly by humans, but it wasn’t. It is a result of a million choices made by software. It made those choices mostly randomly and certainly without the guidance of experience, genetic tendencies, feelings, deeply-valued memories, a culture and the artist’s unique relationship with that culture, a mind shaped by a life. In short, AI imagery is a deception because it is not an expression of a unique human soul. The degradation of the human viewer results from the degradation of the human artist: he did nothing, therefore the viewer is being tricked. It is not a “trick” in the manner of a film tricking the viewer 24 times a second. It is a trick at a much more fundamental level.

The solution, if we want to keep using AI imaging, is for the artist to become much more involved in what it does. A thousand times more involved.

If that can happen, then “AI art” can be actual art.

But let us keep in mind that AI imaging, right now, does not enable that. It is not art. It is essentially slop, conjured with just a few key words and showing nothing - literally nothing, let’s be honest here - of the soul and personality of the human who provided those key words. This, even leaving aside the issue of artificiality, is enough reason to loathe “AI art” today.

The equivalent in traditional film-making would be a film director who had hundreds of crewmembers working under him but chose to have no concern about what any of them did throughout the film shoot. He would end up with a film that might be technically proficient in every way, yet bland, lifeless, and lacking his unique “creative voice” - or any creative voice.

If AI imaging advances as I expect it will, there will be countless people who use it in the contemptibly lazy way I’ve just described. Garbage in, garbage out. Any user of AI imaging could choose to have no concern about what it produces (as long as it fits his initial, very brief, description). Well, he will end up with a drama that is technically proficient in every way, but dull and unmoving. In other words, AI will perform a huge number of functions for you, and it will perform them well and the results will even be beautiful, but unless you are closely engaged with those many functions and nudging them meticulously, the result will be as lifelessly generic as if a traditional film director was asleep on set. And that scant level of involvement, which a charlatan might exert, is all that AI imaging enables just now. Thus it turns creators into charlatans.

What is needed is a toolset for interacting with the AI in very precise ways. For example, I tell it to create a street in Edinburgh in 1987. It scours the Internet for photos of Edinburgh architecture and based on those it creates an appropriate street. The vehicles are accurately 1987. The shop frontages are apt for the time. But let’s say that, though this basis is adequate, it’s not really what I had in my mind’s eye. I need to change specific things. There is currently no way for me to do this. What we need is a way to tell the AI to keep the street pretty much as it is, but:

• make the street twice as long

• add a T-junction here

• change that shop to a takeaway

• more of the buses should be double-deckers

• change the camera angle to this one

• make the pavement on that side a bit wider - no, not that much

• the weather should be sunny

• and it’s early afternoon

• the shop ceilings should be higher - but only along this small section of the street

• a particular type of truck has to pass the camera at exactly 9 seconds, going from left to right. Its driver should be X character, wearing Y clothes, and there should be a bruise on his forehead at the exact spot where he was hit with a steel bar in scene 18

This is the level of control possible for a film director working with a large crew and a budget completely out of reach for an ordinary person. Said ordinary person could have the same level of control, but only if he was working, not in the real world, but digitally, using standard CGI software. The trouble is that it is painstaking work and the result will be nowhere near photorealistic. The unique promise of AI is photorealism achieved via verbal communication between the human and the software. This is already possible but there is no apparatus for fine-tuning at a level that would allow us to legitimately call ourselves “artists”.

Once that is achieved, I think it would be wrong to call the output of such an arrangement “AI art”. That would be to fetishise the role of AI when the artist, given his deep level of involvement, should rightly be claiming authorship. The fact of AI being involved should be incidental. The artist used AI, yes, and it might have been the main or even the only tool he used, but it should be the artist using a tool, with as much control as the painter uses a brush, the sculptor his hands, the author words, and the director his cast and crew.

At that point, it would seem absurd for somebody to say of the resulting film:

It’s not really a film, because making it didn’t involve hundreds of people and a huge budget.

That will seem like an arbitrary and ridiculous stipulation for someone to make, and self-defeating for them since they will be depriving themselves of potentially very good films. Again, we come back to the painter: if he doesn’t have to produce his oils and canvas and easel, why should a film-maker have to work with a crew of five hundred people in order for his film to be “real” or legitimate?

I think the answer is that we value the idea of people making something. This is very understandable, but that requirement is still fulfilled in the AI scenario; it’s just that every decision is made by one person and the labour is done by software rather than five hundred technicians. The labour involved is hundreds of hours of solitary toil by the person wrangling with the AI.

Then someone could say that they value the fact that the labour was done, not by one director, but by five hundred technicians - all specially-trained and instructed. I also value that idea, but the reality is very different. The carpenters who build a film set are anonymous, and they remain so no matter how knowledgeable a viewer might be about film set carpenters. Likewise, the sound designer, even the costume designer... can anyone really distinguish this person’s work from a colleague’s, when analysing a film? I very much doubt it. But still they like the idea of it, the idea that real humans struggled, that the labour was honest, and that the resulting film speaks of collective human endeavour.

I understand all of that. I share it to some extent. But I think the time might be coming when we should start appreciating individual human endeavour, because it is about to be massively catalysed by technology.