On (not) Being an Artist
As I didn't go to art school (shame, I wanted to, but did medicine instead and then worked on the origin of life and now do artificial intelligence {natural path if you think about it}) I thought I should read some books about it to learn how to be one. Here is what they said... I'm interested in this for two reasons 1. I want to be an artist. 2 I want to automate the artist (as much as possible) and understand how AI could help with each of the things that make one an artist, so I don't have to bother doing it all myself which is such an effort, all that painting and drawing, etc... although I enjoy it in moderation. Things to maybe do are in italics.
Think Like an Artist by Will Gompertz
Be enterprising and business minded, lie if you have to, like Theaster Gates in order to be valued and noticed. A depressing thought, not great at this, and don't know how to do it.
Have a positive learning attitude w.r.t. failures, like Bridget Riley who took ages to find her style, by experimenting and getting negative results, but then discovering skills and knowledge along the way which she was able to synthesise later. The same applies to science of course. I'm ok here I think. Right now Arnheim is an imitator, e.g. the Fall of the Damned was pure and limited imitation. Combination can be done with prompts but I'm not sure it can be done with understanding.
Be curious like Maria Abramovitch and Caravaggio, allowing invention, discovery, a questioning of biases and assumptions behind what you are doing in other words. Yes, very important. I have lots of artistic biases that I am not curious about, could improve here totally. [be Socratic thus questioning assumptions and solving problems consciously that would otherwise be unconscious] Get a collaborator who can help me to take myself seriously. Andrew Mania and I were like this for a while at school, it was a very powerful artistic experience.
Steal rather than copy, like Picasso did with the blue period (and Christopher Wood unfortunately didn't because he died too young, by the way Wood is all copying, but very good copying, I'm said he killed himself). Steal means understand the other things and simplify and integrate, compress, understand concepts underlying the medium. Understand artistic concepts at a deep level, and use the learned processes with understanding. OK, makes sense, must try to understand art better. I enjoyed the example of Picasso and the simplifications of the Bull, such abstraction is compression in Schmidhuber's sense, it displays understanding by effective compression. I'd like to get Arnheim 3 to do Bulls with fewer and fewer primitives and see how it does. Idea for a generative art piece. Like Piela de la Francesco see if Arnheim can learn how to generate the feeling of space when asked to draw over an empty architectural form, or even drawing from scratch. With a prompt like "A Specious room" how will it convey the feeling of space?"
Simultaneously do big picture and detail thinking, thinking of the context of the work, and the details of the work itself, so have a hierarchical attention system w.r.t. to the execution and implications of the work. It must have meaning and integrity at multiple levels. Makes sense, like the opposite of a plane crash. Thats why I'm interested in plane crashes I think, because it shows usually a hierarchical compound failure of a complex system [by the way Gompertz is a bit uncritical in his thinking about the role of the computer and the superiority of the human mind, which is a bit tedious but I'll ignore that bias]. In this sense, a good artwork fits into its niche well, it is well adapted to its niche, and displays adaptations at multiple levels, to use an ecological and evolutionary metaphor. His example of a small pink dot on the Girl with the Pearl Earing's mouth seemed somewhat trite, that a painting must have a 'way in' but I think this makes sense in terms of attentional blindness, we are very little aware of what we're actually seeing, think of the change blindness examples like the Gorilla and the basketball video. Subtle unconscious effects are massive in painting. A good artist learns to exploit these unconscious perceptual effects. Can CLIP help us to discover such effects and principles!? How might we automate the discovery of such effects if they exist?
Have a point of view, e.g. Peter Doig needed to discover he was interested in the subject of memories being collage (mmmm) what separates the real and remembered sounds relevant to CLIP actually as duel-encoders are in a way dealing with the remembered. What is my point of view? Race, I have colonial mentality, am a bit of a coconut, don't know much about Sri-Lanka, feel an outsider and not really included, but marginally English. AI, I suppose I'm not a fan of humanity much, and for that reason I have some hope in AI, but I'm not terribly interested in ethics as such. Borderline personality disorder for which I've received DBT at the priory, after a girlfriend told me I had it, yes, maybe. Although I'm not sure I do, I think its more an environmental issue related to suddenly leaving my nanny in Sri-Lanka at the age of 3 and never hearing her mentioned again by my parents! But how to do art about the almost delusional fear of abandonment one can have based on such an event, and the terrible feelings of being alone? Also, living in psychiatric hospitals until the age of 9 probably didn't help! If you could modify a piece of art to express something about yourself what would it be?
Courage to express ones feelings and ideas in public to a potentially hostile audience.
Pause and be self-critical sounds similar to the Socratic point above, but modify ones work in response to this valuation. Have a value function and exercise your policy on that value function in other words. We totally need to add the possibility of individual patch modification to Arnheim 3.0 in order to allow the users value function to be effectively exercised.
I think thats about it.
Silbury Hill [Version 1] inspired by the Ruralists, David Inshaw's beautiful paintings in particular.