Ghibli, demand and supply
If something can be produced easily, it ceases to be art. And the art world moves on to create something unique.
The Current Thing
Once in a while I like to comment on the “current thing”. And for the last week or so, the “current thing” has been ChatGPT’s new image generation feature that is leading everyone to upload their pictures and get it in a Studio Ghibli-like format, and then flooding social media with it.
I’ve been off Twitter for the large part since this happened, since most of my feed has been taken up by such pictures, which is annoying. Instagram has started becoming infested with those, as well. LinkedIn, as expected, has more posts outraging about this current thing, than posts with this kind of images.
Some of that outrage is valid, in terms of potential copyright issues and “theft of art”. Other outrage is plain bizarre, like people claiming about the “impact of AI on climate” while outraging about the Ghibli images. It’s like their attempt to shoehorn the current thing into their single item of agenda.
Yet others have gone into the Marxist concept of “labour theory of value”, arguing about how Miyazaki and his team like to hand draw all their animations, and how it is an insult to them that an algorithm can now, upon demand, produce images of their style in a matter of minutes.
DallE Images
It’s been a couple of years since the likes of Dall-E came about. I remember during Diwali 2023, using (the then rudimentary) Dall-E to create an image of Krishna defeating Narakasura in an aerial battle, to share with whoever sent me a Happy Diwali greeting. Here is that image (I had to dig deep into Twitter to find it).
Notice that Krishna has three legs in this. Also notice that this doesn’t have any text - back then Dall-E couldn’t generate any legible text - it was always garbled.
Also, back then this kind of stuff was novel, and cool. I remember getting a lot of nice comments and replies to this image that I had generated then. Basically this kind of “art” was scarce, and so had some value.
Soon everyone and their uncles started using Dall-E to generate images. Soon they lost their value. Soon I started loathing this “Dall-E blue” that all images used to have. Even if people considered images like the one above to be “art” there was no value to this art because it became all commonplace.
Relative valuation
As an asset class, art is one class that is almost impossible to value. I remember coming across this somewhere recently, but can’t remember where, where valuation guru Aswath Damodaran spoke about how art cannot be “valued” (because there is no intrinsic value), but can only be “priced”, since other people ascribe positive value to it.
One way to think about art valuation is in terms of scarcity - think about the situation where when some dead artist’s lost painting gets found, and then is auctioned at what seems like an obscene price. The price is high because the piece is now scarce, since its creator cannot create more art any more.
On a related note, you can go to any museum and get “prints” of any art piece they have displayed there for a few dollars. There are lots of copies of the “prints” available (and they can be produced on demand, at a negligible marginal cost), and only one copy of the original. And so the valuation differs.
Even if we don’t think of valuation of art in monetary terms, we instinctively value it (with our “attention”) based on its availability. For example, when the first couple of Studio Ghibli style pictures popped up on my Twitter feed, I was intrigued, and spent a few minutes looking at those. Soon my twitter was so full of those pictures that the price I was willing to pay for them became negative - I stopped using twitter because I was bored of seeing these pictures. I saw other people tweeting about how they have muted words like “ghibli”.
AI and art
The general discourse with AI and art, be it the era of Stable Diffusion and DallE, or the current era of much more advanced image generation capabilities, has been that it is “easy for AI to create art”. However, my argument is that the very ease with which it is able to create “art” simply makes it not art. If everyone has access to the same code that can produce the pictures, then it is easy to produce similar pictures. And these pictures are now so commonplace that people stop seeing them as “art”.
Unless we are talking about art at the meta level where the prompts that go into producing these images are themselves art (suddenly reminded of a concept from my Business Law course in IIMB in 2005, where we learnt about only written text that can be copyrighted. So in a sense, prompts can be copyrighted), what AI produces is not art.
And the human ingenuity is such that it will leverage the tools that are now available to create what can actually be scarce, and new kinds of art will emerge.
There is this story of how in the 20th century, the coming of photography meant that portrait paintings, which were hitherto commonplace, suddenly lost value, and that’s when we saw the rise of “modern art forms” such as impressionism and cubism and dadaism and all such.
It is similar with AI. There is stuff that used to be “artful” that is now commonplace to create, and that loses value. And artists will find new ways to create art. And this will go on.
PS: I tried creating an image (same as above - Krishna killing Narakasura in an aerial battle) in the style of Studio Ghibli. ChatGPT said “I wasn’t able to generate the image you requested because it didn’t follow our content policy.”. Looks like Ghibli’s lawyers have started working!