The ethics of AI art

Post Reply New Topic

There are some great ethical conversations to be had about AI art in the future, but what we have today is not made by an artificial intelligence, but through machine learning. The "AI" involved can only spit out styles and types of art that it was trained on, and any so-called creativity comes from blending the work it already has been trained on. That is where copyright infringement cases will focus in the near future... is consuming specific art without paying those artists for their work something we want to encourage or discourage with legislation?

I hope we will discourage it, but we'll see... if nothing else, these machine learning programs should be trained on the widest possible spectrum of art to avoid the chance of ripping off a particular artist with its work. But even then, all its sources should be made public to avoid any confusion or misinterpretation of its work.

However, I think there is an interesting aside to discuss for how technology is used alongside humans in the art world. We can look at how chess evolved so quickly since the mid-90s because of the advancement of computers that are great at things like tactics, while humans focused on their strengths of strategy and creativity. The types of tournaments involving hybrid teams -- humans & computers (eg: Advanced Chess, or cyborg/centaur chess) -- has pushed ratings higher than ever, proven some weird openings actually viable, and answered centuries old questions about how good a certain games and positions really are or were for their time.

Art is much more subjective and personal, so I don't know how/where machine influence can help us (except in the production of quick art assets), but until "artificial intelligence" is actually here I think that's the main conversation to have aside from the legal implications.

^^^Excellent and thoughtful post.

I think I disagree on a couple of things. Regarding copyright, I think this is going to be interesting because it raises the question of who owns the rights to AI art. I would say whoever created the ML model owns the copyright since generative art is based on what the artist or developer trained the ML on based on personal taste. Once a human being is in charge of the direction machines take in generating art, that person is responsible for that art whether that human had a direct role in creating it or not.

I also think that ML generative art challenges the present-day notion that art is strictly subjective. Art ONLY being objective means that art can only be defined within the human mind. ML proves this is clearly NOT the case since ML depends on being trained on objective data.

Kind of reminds me in some ways of when photography was new in the mid-1800s, and there was a push by traditional classic artists to demonstrate the superiority of the portrait over a photograph. But, in the end, photography overtook painting.

I don't disagree that the person who trained the machine learning model would have ownership of the generated art, that seems to make sense (they did the work after all). However, unless they train it solely on their own art, I think there will be conflict between them and the people who made/own the original artwork.

Right now, most ML art generators have been trained on "stolen" artwork, similar to how face scanning systems have been trained on our social profiles and photos. As long as that's used for non-commercial work, I think they skirt the law in most regions. However, some states and countries are fining companies who trained their ML systems on profiles without giving users a notification or chance to opt-out. The EU is leading that charge at the moment, and those cases will likely inform how everything with "AI" generated art can proceed from a commercial standpoint.

But I would definitely argue that the person training the system has less ownership of the output than the artists work it was trained from. After all, it seems like the success of the system has more to do with the quality of the inputs.

Going back to the chess example in my other post, if you trained a chess computer on amateur games from a local chess club, no matter how good that "director" is at picking the games and tactics to focus on, it's going to lose when put against a system trained on grand-masters games. Similarly, an art system trained on student work will create very different work than something trained on the masters which people pay millions to own. Our copyright systems aren't made for this sort of world, and how things change/advance (and the ethics of how it should) will be super interesting.

You seem to vastly misunderstand Levin's work. What Levin describes as mind being made up of a collective of intelligences is about emergent phenomena on different scales. It's related to Conway's game of life, but mereley giving an example but actually explaining it with real organisms and researching how communication between cells and parts actually work.

Jung however was deeply esoteric and didn't explain anything, he just felt things had a deep connection, based on his visions whenever he had epileptic seizures.

They are diametrically opposed: Jung is elevating everything to the level of having a mystic spark of divine creation to it.
Levin is trying to map that spark in terms of electrochemical signals to demystify it.


My guess is that the advent of AI will cause some serious mental issues for mankind. If a computer, of which we know it doesn't "think" can hold a conversation, create music and create images, where does that put humans?we have assumed humans were somehow special among the animals, because they were either made by god, or at least had some mystic spark no one could explain. But now this doesn't just make humans less special by putting it amongst the animals, it puts animals and humans amongst dead objects.

The secular humanism western societies have subscribed to is merely Christianity where we don't say "because we're made in god's image". ... I'm genuinely worried by what will come after.
I don't see why the ultra wealthy should keep other people around when machines can do everything, including entertainment and scientific progress.
It's a mistake to think AI art only creates repetitive and derivative work - it just doesn't yet have the ability to train itself on new images, once it has been trained by a user.
And art generators by themselves do not judge quality.
But one could easily imagine GANs to create and judge art and build a sort of pyramid of algorithms - you know, like there are hierarchies of judgement for artistic creation out there right now.
Most people don't care much for the raw, random artistic output of individual humans. Even for art people,there is a minimum of one curator sorting the generated artworks.
For most people, this is still way to extreme, and several steps of selection and judgement later, you have the smallest common denominator of a blockbuster movie about heroes and bad guys, a musical about a rock musician from the seventies, a series of books about a child who discovers he's actually got super powers, etc.

So... I think there will be computer generated art of all kinds flooding every communication channel, while people generate their own. ... I think art will no longer be a product or used for interpersonal communication. I think there will be no more commercial artists, musicians, and soon filmmakers, photographers, actors, or models.
There will of course be celebrities, but they will rent out their likeness so you can watch the film you want with their face in it.
It will be 'Brave new world', but there are bigger problems coming at us in the next decades than the end of commercial artists -

Human made art as its own category of course already exists and will keep existing, we call it fine art today and it will be as it is now: status symbols and items for speculation for the wealthy.
Because whoever cared about a person peeing at a canvas in the first place won't care about a computer more efficiently creating jpgs of pee stains on canvas.
Fine art hasn't been about the artwork itself for a long time, so a more efficient way to produce artworks doesn't matter.

Humans appreciate art. But in the future AI will be used to create art and other AIs will be used to appreciate the art. In this way, humans can be cut out of the loop entirely.

That assumes AI is programmed to imitate human sensory stimulation by receiving signals from the environment. The level of complexity involved in top down processing of environmental stimuli (as happens in our brains) would be really hard to integrate into a computer chip.

Why imitate such limited senses? AI may be able to create and appreciate art in a spectrum ranging from deep infrared to ultraviolet.

Harder than it looks