AI should have sent a poet

AI should have sent a poet: another go at talking synthetic art

So I went to a Comica talk on AI by Dave McKean (with sidebars from Iain Sinclair). 

McKean is always thoughtful and talked a lot of sense on how Machine Learning impacts the creation of images and how we think of Art and Creativity. 

To paraphrase: “There are valid concerns about permissions and ownership and the detrimental effects on employment, which I share, but tonight is about what AI means for how we see ourselves and what we make.”

The talk itself was excellent and sparked many further thoughts on the subject (although, possibly fewer than it would have if it weren’t for the bloke two rows behind me indulging in a running commentary/argument in a stage whisper, and the usual flooding of Q&As with ‘not really a question, but I just wanted to hear my own voice fill the room’). It was hard not to come away thinking how nice it would be to hang out with Dave or to bat this stuff around with him on a country walk. 

If you’ve read any of my early musings on AI [part 1. part 2. part 3], you’ll not be surprised to hear I’ve adjusted my thinking again thanks to this new input. I’m not going to reiterate the evening (you should have bought a ticket and /or his book Prompt!) ~ but here’s where it sent me…

Paul Gravett introduced the evening in a lovely shiny suit, but left for the shadows of the front row about the same time fould my pen…

 

Alongside the seismic changes it’s bringing to industry and daily life, AI makes increasingly impressive images and text that are becoming harder and harder to distinguish from the work of skilled humans. 

The output is remarkable. The speed and flexibility, game-changing. If it takes 1000 prompts to generate 1 commercial blockbuster of an image, that’s a morning’s work. 

This is AI as Tool ~ and like digital colouring, FTP, acrylic paint etc etc, it’s unstoppable in its spread and uncaring of the professions it decimates along the way. It is replacing those skilled humans on a number of projects, and making us wonder if skills that are valued because they are difficult to master should hold their value when they can be mimicked with the push of a button?

But just as the machine cannot care, it also cannot create Art. The best it can do is filter options for a third party to decide something has merit, either on its own (where the viewer is the one giving it meaning) or as a component in a work of art (where the maker uses it to express meaning). 

We are going to be forced into redefining words like Art and Creativity. We are going to have to let go of the idea that facility has even a passing relevance to artistic merit. 

It really shouldn’t be that hard: we know that a Lowry tableau is ‘better’ than most photo realistic street scenes, but thanks to decades of ‘creative professions’ we’ve gotten used to calling craft ‘Art’. As if doing something exceptionally well ‘elevated’ that thing to a work of art.

Pfft!

Deciding Art was a qualitative distinction rather than a human process fucked our values and made us forget where the art sits in the things we love. 

As McKean said, AI doesn’t collage a banana from existing images ~ it looks through the whole of the internet for tags that show or describe what a banana is, and then runs through a massive series of refined approximations until it generates an image that satisfies the same criteria as a number of those tags. 

So…

my thinking is that one way to use AI artistically is to ask a deliberate, creative mind to pre-filter the dataset the app refers to. To write new tags about the ineffable banananess of bananas.

Set the machine different parameters and it will offer you different, more surprising, results ~ things that are not just a distillation of everyone else’s idea of a banana, but are generated in response to your decisions.

Isn’t that the space where we’ve found our creative voices with new media throughout history?

It strikes me that this is a positive way to look for something new from AI. Currently we are seduced by novelty but that wears thin very fast. The initial, indiscriminate roll out we are seeing is going to hit a dead end pretty soon: the day after everyone can make impressive Content is the day that Content becomes uninteresting. 

More worrying is the nattering monkey voice in the back of my head that reminds me “the Greeks didn’t have a word for the colour blue, so they couldn’t see it”. It’s a mangled version of what perception is, for sure, but if all that Content is cobbled together by scraping existing tags, where does anything new come form? If we haven’t coined the tag yet, can AI imagine the thing?

We’re already in a world where shitty cover versions of great songs proliferate; where facsimiles and lesser quality reissues crowd the shelves ~ it’s just going to get worse as the snake eats it’s tail.

The Content we are served by writers who always wanted to make a graphic novel but can’t draw, by creatives who want to make a movie without the cost and logistics of physically filming it? That Content will doubtless look fabulous. But by summer 2024, even the best of it will just be a blip in a sea of fabulous Content. 

Humans always find value in scarcity or peculiarity. Perfectly rendered … stuff … will always appeal to an audience, but most people will be enjoying it in the way they enjoy a quilted toilet roll.

We may never agree on what Art is, but can we now scratch ‘convincing reproduction of the real world’ and ‘looks almost real’ off the list?

I think it’s good if we can. 

Yes, still, there are a shed load of jobs and careers under threat. I reckon 90% of my paid work last year could be redundant already. 

Yes, the ‘democratisation’ of creativity is a lie (of course). Tech is owned and datasets are controlled. If you thought we had a homogenised culture last year, you’ll all be agreeing it was nothing like next year. 

None of this pleases me. At the same time, I’m not filled with dread, or righteous anger. It’s shit in how it impacts individuals. It’s crap in what it will flood the marketplace with. But you already sit through Strictly Come Dancing and hatewatch Dr Who. You buy Spider-man despite it being convoluted retreads of the old days and you give the Waterstones best seller as a Christmas present, knowing it’s going to find its way to Oxfam by Easter. 

We’ve been celebrating and enabling a culture of Product for so long that we are defending pap. 

Art has always been available to all, but it’s also had a price far above rubies. We’ve studiously ignored that in our co-opted, commercialised, commodified, collaboration with Capitalism. If we have lost sight of what Art is and does, AI is going to remind us that many things we once put on pedestals are not art. 

But we’ll still want to buy them.

Gilgamesh from Dave McKean’s Prompt is exactly the sort of thing I love… but can it satisfy as much as if he’d photographed real tablets he’d hand-crafted to tell the tale?

 

Which prompts (see what I did there?) a footnote that will inevitably sound like I’m having a dig at Mr McKean, but I’m really not… it’s just a useful illustration of the upheaval that artists and authors are going to have to work around for a little while:

You can (and really should ~ it is astonishingly good work) buy Dave’s previous book Raptor today and get 128 pages of Pure McKean that took him months to write and draw for just £23 (or £13 on Kindle). That’s a more than fair exchange, even if the author himself gets a pitifully small proportion of your money. 

The 112-page Prompt took him days to produce and is on sale for £18*. 

That’s 2p a page less you would be paying for Pseudo McKean with a gloss of The Real Thing over the top**. We all need to eat and it would be churlish to begrudge a book having a similar price point to others, just because it took less time to make. But all of 2020s capitalism is a race for the bottom and it’s not going to be long before customers start to believe they should pay a lot less for any Product that was made so quickly ~ especially one they feel they could make themselves if they just typed in the same requests.

Obviously, the handmade work  is far superior… for now

 

We’ve insisted for so long that we should be paid for the skills we use and the years we’ve taken to learn them… it’s going to be hard to keep that argument going when any wannabe can convince themself they just need the right prompt and they can make the same stuff.

(Remember when you thought that if only you could find the right pen you could draw like Dave McKean?)

We may have to train bespoke AI bots if we want to use these tools for making Art and to maintain our value to clients…

So, thank you Dave McKean and Iain Sinclair, and Paul Gravett and the Comica team for a fine talk ~ this one really did get my gears spinning!

footnote: 

*ironically there is not a digital edition of Prompt on Amazon

**I’m exaggerating for effect, of course. Prompt IS McKean using a new tool