One of the interesting things to pop up online recently has been the subject of art generated by so-called artificial intelligence (AI). It’s a subject for contentious debate: can something really be considered a piece of art if it has been “created” by a computer program, rather than a human being?
My interest was piqued by Lee McAuley of the Cuckoo Club Archives, who mentioned it in a recent piece, and to whom I give all credit for spotting it – I’d no idea it was so advanced. In order to explore the question, is it art? I’ve been playing around with a version called Deep AI – available to try here, and I fed it the following text:
An old pier running out to sea, sun setting, people walking towards the sunset, blue skies and tobacco coloured clouds, light rays, romantic, impressionistic style.
The result was the header picture. Then again, same input:
And again:
So, each image is unique: same words, different output. There’s also a remarkable alignment with the textual prompt, whilst maintaining the look of something definitely painterly, rather than a pastiche of images brutally cut and pasted from around the Internet. There’s something interesting here and, though there’s a temptation – as a human being who likes to think of himself as “creative” – to be dismissive of it, I don’t think we should be too hasty.
The freebie images are a modest 1024×512 pixels, but useable, say for blog illustration, or, with a bit of Photoshop enhancement, as e-book covers, or simply for pondering. I find them quite haunting and, in spite of their unique nature, strangely familiar in that they combine elements I feel I have seen before, but which are just out of reach of memory.
There are other online generators, free to try, but they all have some kind of limiter, or a token system, to prevent over-use of the servers. I also like Nightcafe Studio, which I fed the following prompt:
A young woman wearing a long, red dress. She is reclining on a chaise lounge. Victorian and romantic in style.
To which it responded:
The result is somewhat lush and stylised, though not unpleasing, and nicely lit. She has an oddly shaped thigh, strange hands and what appears to be the stump of a third arm, but for all of that it would not look out of place on a gallery wall, given a suitably pretentious blurb. It’s also unique – sort of. No image will ever come out quite like this again. However, once you’ve got the image, you can copy and paste it as many times as you like, of course, which, like all digital art, renders it nothing more than a worthless and disposable curiosity, right?
Well, that brings us to non-fungible tokens (NFTs), which I looked at last year. NFTs and digital art go hand in hand. Digital art, whether it be by human or AI, is – by our normal calculations, based on supply and demand – of no value at all, because we can copy and paste it as many times as we like, and the result will be indistinguishable from the original. However, AI generated art can come with a unique digital token, which proclaims you as the owner of the original file, which is something that, in our topsy-turvy world, can then be traded. And, though it might sound unlikely, it being essentially the value of nothing, some tokens are trading for millions of dollars – or at least those that receive the most hype.
Here’s another one. Input: Man writing at a desk, background of bookcases. Lamplight. Studious, romantic atmosphere. Impressionistic.
The result:
The debate over AI generated art also throws up the old chestnut about the nature of human consciousness, and the belief among the so-called “hard AI” scientists, that it’s just a question of time, and a critical mass of artificial neural complexity, before we create a sentient computer. But this kind of thinking is bourne out of a strictly materialist paradigm, and goes too far for me. Our machines are breathtakingly intelligent, but that’s not the same thing as saying they might ever become sentient. Like a chess playing computer, it does not arrive at its moves by thinking about them like a human player, but its moves are always good ones. It does the same job, but better. Like an electric saw, it’s better than a handsaw in certain applications, but only because we have made it so. And even then, we wouldn’t use it everywhere.
AI sentience also rather presupposes the brain is what generates consciousness, and I do not subscribe to that view either. I’m deeply impressed by A.I. generated artwork, but feel there’s a danger here of setting off down the wrong path in our appreciation of what it means and that, like all A.I., we should not be tempted to make the retrograde leap from master to servant. A.I. serves a purpose. It can protect, it can run complex services on our behalf better than we can ourselves, and it can entertain, but it cannot be allowed to control and delimit, either our actions as free beings, nor supplant our imaginations.
Another one: Input: A young woman in a long red dress, fantasy forest setting, backlit, lush greenery, light rays. Output:
A human artist invests time learning how to paint. Then, having mastered the art, a large painting might take months, or even years of the artist’s time to complete, and the end result is always going to be fragile. It’s likely then, a very old painting by a recognisably competent artist will have survived any number of potential calamities, and is worth all the more for its rarity, and the simple fact of its survival. By comparison, a computer generated artwork takes seconds to make, and the result can be backed up digitally so many times as to be virtually immortal. NFTs not withstanding, I know which artwork possesses the greater intangible value, the greater allure, to my own taste and I would care nothing for who owned the digital title to an AI generated artwork. All of which is to say, while AI can produce some stunningly beautiful and provocative images, let’s not lose our heads over what it means.
Is it truly art? Well, yes, I think it is, but certainly not like anything we have known before.
The Rivendale Review 