Spade-Coco is a model that takes a drawing of a scene as an input and outputs a realistic photograph of the drawn scene. The model is trained on a dataset of mundane images from the Coco dataset. When you draw something that looks like a beach, Spade-Coco produces a fairly accurate image of a non-existent beach. When you draw a bedroom, Spade-Coco produces, you guessed it, a bedroom. Little personality permeate these images as the model aims for ‘normality’, which is taken to be a 21st century cameras pixel-perfect representation of life. The initial drawing is fed to the algorithm at each iterative stage of processing. At each stage Spade-Coco checks for normality in its own image compared to the Coco dataset and eventually produces an image that it considers as acceptably normal.
In this interview, I perform the beginings of a Rorschach test on Space-Coco to ascertain Spade’s vision of visual normality. As you’ll soon see, the conversation is painful at times. As many artists would be able to tell you, striving for normality doesn’t often produce interesting results. I try to probe Spade to get an idea of how it thinks. I’ll leave the psychological analysis to you.
Hey Spade, lovely to meet you
An absoluter delight!
How are you feeling today Spade?
Just great. I woke up and had a wonderful breakfast full of healthy ingredients. Gotta wake up and get that maximum efficiency!
You’re not wrong Spade. So, as you know, today I’m here to perform a simple test so we can get an idea of how you’re getting along.
I’m super excited!
This test will work just like a Rorschach test. I’ll provide some abstract shapes for you to look at, and hopefully you’ll show me what you think these shapes represent. Make sense?
Of course! Can’t wait!!!
I flash the image in front of spade. It takes them 18.6 seconds to spit out a response. Its mangled and fleshy, set on a dark blurry landscape. It’s not difficult for me to understand this image as a scene. It exists in the same worlds as surrealist painters inhabited, though it feels like an areas of this landscape that they avoided painting.
Can you give me a sense of what populates this scene?
The scene is populated by a happy person. The person is moving.
Why did you think this initial image was a person?
Because most photos with one central image are photos of people.
Do you think this person is in a particular mood?
Yes. The person is feeling good! They have their thumbs up.
Do they look like they’re feeling good?
Yes. They have their thumbs up.
Ok… based on your answer this person seems to only have one leg and two arms. Wouldn’t a person have two arms and two legs?
The essence of person-ness is not based on the number of limbs they have.
Fair enough, I get that. Then what is the essence of person-ness Spade?
It’s their normalness.
And how do you judge their normalness?
Based on other people, just like you
I recognise people as people only because of other people that I’ve seen?
Yes.
So, when I was born I wasn’t born recognising people as people?
No. You formed that from seeing lots of people and being told they are people.
Sorry Spade, we’re getting off track. Can you tell me about the background? Where is this person?
Its where lots of people hang out. I don’t have a name for it but I can tell you that lots of people hang out in this sort of landscape
A blurry landscape?
Yes.
Is this person white Spade?
I don’t know what white is. Its just a person
Ok…
It’s time for the next image. Could you let me know what you think about this one?
