A friend had the brilliant idea of asking one of those image AIs to draw a diagram of the anatomy of a human hand, since that seems to be the thing they have the most trouble drawing. The AI managed then produced these diagrams, which are frankly impressive in some ways, while 100 % confirming that they cannot draw human hands. Never mind getting the number of fingers right, look at these skeletal structures and tell me there aren't deeper issues at work.
I like the forearm ribs on the bottom left.
This is I have no mouth but I must scream sort of shit
let's see anyone get carpal tunnel now, with this new and improved beefy wrist the size and build of a horseshoe crab
The neck is nice, but unfortunately his hands seem determined to grow the wrong number of fingers in both directions to be able to crawl around on their own without being attached to any kind of wrist. And I can't support that.
Yeah, the imitation of the style of a medical diagram is actually very impressive. And as a human it is fascinating how it can be so good at imitating that particular style of drawing - which I think is very difficult - while not being able to draw the correct number of fingers - which I think is very easy - even on a single human hand drawn from straight above in the most neutral pose possible.
Yeah it's a fascinating demonstration of the limits of 'AI' in the fact that even with the full depth and breadth of humanity's knowledge, without clear and concise commands to direct it its about as useful as a paperweight.
I think they try to train these AI by having the people who requested the picture grade the results. That might help them learn to avoid making ugly mistakes like lines not meeting, but I think it will reinforce the problems in this picture. Because the only feedback this AI would get would be "FUCK YES! This is exactly what I wanted! Perfect drawing of a human hand!".
Yeah, like unless they program the bot parameters to interpret the vague language we use in common parlance into actual useful data that positively builds it's database then yet again it'll have the usefulness of a paperweight but with an interesting gimmick.
This is a cool result, and I enjoyed watching it, so it's not exactly useless. They just cannot improve accuracy from most people's feedback. I don't think I'm alone in giving the top right picture with seven fingers and a creepy spider-human hybrid skeleton top marks. And the AI will just absorb positive feedback as having done the right thing. This will not make the AI better at drawing human hands, but it might make it even better at drawing abominations. I'm there for that.
but it might make it even better at drawing abominations. I’m there for that.
Same, actually. Its like having an AI Zdzislaw Beksinski.
It's because the difficulty of highly detailed illustrations lies in the mechanical aspects that stand in between the idea of the thing in one's brain and the abstraction of that on the paper/tablet/whatever medium. For these AIs, there's no idea of the thing or messy systems of tendons and muscles getting in the way, there's just a feverish image formed through data association.
The real breakthrough I think would by layering the system with something that could create that sort of idea of a thing, even if it's just a collection of barebones 3d models and a neural network trained to animate them that then provides reference images into an AI that turns that into art, sort of like a very fancy automated rotoscoping system. That would probably be enough to make them get things like number and position of fingers and limbs right, at least.
Imagine the grip strength we'd have with seven fingers
Imagine the grip strength we'd have if instead of having legs our spines split into four fingers like some weird kind of spider legs, each the thickness of an arm.
Never mind that, imagine how good we’d all be at playing piano & guitar
The bottom left one almost seems like the ai is making a lot of circuitry-based assumptions about the anatomy of living things. It looks more like wiring.
It would be awesome if a powerful AI went rogue and tried to genetically re-engineer mankind to what it thought humans are supposed to look like. It would be practical effects horror all the time!
Yeah that's sort of the I have no mouth but I must scream plot
Also Man after Man
The AI can recognize what fingers look like. It can recreate its saved mathematical vectors of finger images. It knows that images of hands usually have multiple fingers connected to them.
The AI does not know how to count.
The AI does not know how to count.
It also does not have a sense of the image as a 3d object like we do. It can not and does not understand that a limb continues in a set direction when it passes behind an object, coming out the other side of the object as you would expect. This causes a lot of the problems with hands if a gesture causes fingers to pass behind other fingers, it has no idea which finger is what and where the fingers should be popping back into visibility. This leads to finger monstrosities.
These AI need to be able to process an image as an idea, they need to understand what a hand is and what makes up a hand, they need to understand and recognise what the human is in an image and what parts make up a human.
...And here is them understanding what makes up a hand.
It's all bones and fingers. Just add more as needed.
Bit idea: Medical AI that insists the problem causing your stomach ache is that you do not have enough fingers.
Same problem as you can get by asking OpenGPT "If it takes 5 machines 5 minutes to make 5 boxes, how many minutes does it take for 100 machines to make 100 boxes?". That one is so focused on understanding language that it manages to be bad at math by understanding the question well enough to answer in the form of a really wrong intuitive guess. This one is so focused on art styles that it can make an incredibly accurate imitation of a 19th century medical diagram while failing kindergarten tasks like counting to five or singing that song about which bone is connected to which bone.
This is fantastically creepy. I would like to imagine the top right skeleton looks like a walking hand with a head on top
Don't worry, it's not actually creepy. This wasn't made by a human, it's just what the computers we are developing to run much of the economy/government think our anatomy should be like.
Mimicking illustration like this is my favorite use of AI. Played around with Albrecht Durer woodcut prints in prompts a while back and it gave very cool results