May be a lot of you may have seen the works of this artist, but I found him recently while surfing the web and I thought I’ll share his channel here for others to enjoy or rediscover his works.
He creates videos of Deep Neural Networks trying to make sense of live camera inputs through what they’ve been trained on (mostly just fire, water and flowers) - really interesting to see.
When watching this I’m thinking it’s beautiful and amazing, then I had a thought.
Even though this is AI at work, of course it’s doing this based upon a learning process, but there is no meaning to its artistry, and the art is that of the creator of the AI.
It’s one of those very difficult questions to answer, What is Art? One might answer that it’s a form of communication of ideas and feelings in an expression that captures the essence of an individual’s experiences.
If so then art requires a conscious awareness that can interpret the communication of this kind of data and order it into some kind of context.
In this case the artist created and trained the neural net towards their own ideas.
Until a machine Intelligence is able to experience awareness and contextualise into abstract forms of communication, they will not be able to truly create art all by themselves.
I agree with you Tim that it is the artist who trained the algorithm and the art is his - the ideas and concepts were his and the algorithm is the tool he used for his art.
Nevertheless when watching the video it gave me the feeling that ‘something’ was there - an intelligence, trying to make sense of what it was being shown through what it had seen before.
Kind of like when we watch a magician perform on stage - there is no real magic, but one is still left with the awe of witnessing something magical.
Similarly, I know the AI is not conscious and if we’ll go into the depths of things - it may just be an algorithm recreating the inputs received into forms of visual outputs it was trained on or something like that.
But there is that sense of magic though. It feels as if the algorithm has an intelligence, it is perceiving the inputs and trying to make sense of things from the limited knowledge it has. Feels as if it may not be conscious but it is there. Learning and then recreating things.
I love it. This, to me, is a perfect example of the slices of controllable alternative perspectives that AI is enabling. While there may not be a part of that algorithm that explicitly has the machinery to make the decision on the direction to take in the imagery it produces, nor the ego rewards of having the view counter go up on that video, I see very little difference between this narrow AI and the visual system of another human that has some serious issues with how their visual cortex is functioning. The artist is the one with the agency to transfer these hallucinations onto a physical medium and control their style base - but how much credit do you give Van Gogh’s uniquely tuned visual network of his brain for his art? At what point does the visual part of your brain start being “you” and not just a light wave encoder?
There is no agency built into that video encoder network (that we know of), so I suppose it was created by the human and not the AI. What if it had an LSTMish network to decide when to make a new piece, a RNN type networked hooked up to be rewarded by views, and an unsupervised image crawler to supply the training data? I feel like that is pretty darn close to agency enough to be able to claim collaboration (the human was holding the camera and moving stuff around, plus presumably created it).
I’m of the opinion that art is defined by interpretation - it being viewed - coupled with a vague awareness that the concept of “art” exists, which seems to come naturally hard wired into human brains with our obsession with beauty and novelty. Probably had to be present for reproduction originally, then we got some bit of wetware that totally went off the rails.