Most of the thread is opinions not related to my argument, additionally many opinions expessed I’ve openly agreed with in this thread.
AI is a tool & Art is an expression of human creativity.
Art is subjective, but the subjectivity of art doesnt change meanings because of ones feelings.
Id say Blood Red Mirror “lacks” creativity, but I wouldnt say its not art. It sold for 1.1M.
Its still falls under the definition of art & speaks to the moral state of the art industry.
My personal view is that ai is in itself a tool but people misunderstand its output and function leading to misuse. If the input is bad then the outcome is also bad. If the input is stolen and unoriginal then the outcome is stolen even if it is original.
Dumb analogy: It operates like Nintendos Kirby, sucks in originals and makes something that is based on what it sucked in, most of the time in a new way. At some times and with skillful planning it can lead to creative and unique outcomes, but if it sucks in generic stuff the the outcome is also generic. But like Kirby, just getting an outcome from the input doesnt do anything. You need to use it at the correct time or add more personality to it to really make it its own thing.
If a data set was more directed and filterable, then I could see it having more nuanced use. As it is, it is an amalgamation of interpretations without direction and is unfinished at its outcome. If artists went the next step and based art from its interpretations then it would solve many of the issues. Ai should be the skeleton of a piece while art is the fleshy details, the specific highlights and shadows, the detailed fingerprints in the piece.
I think the discussion of art theft may be distracting from the heart of the issue one may or may not have with AI “art.” The art theft aspect is built into pretty much all current AI training models, so it can’t practically be decoupled. But as a thought experiment, suppose hundreds of thousands of artists consented to use of their work to train an AI model such that theft of art was not an issue. Would you still have a problem with the outputs of this AI model? How about if an artist utilizes AI that’s trained only on their own work. For the latter, would straight outputs from prompts still be looked down on? What if they touched up and slightly modified the AI output?
I think it may be important to try to look past the obvious unsavory parts of contemporary AI art (e.g. use of art without the owner’s permission, utilization of AI art to outright replace artists, etc.) to see if there is really a space for it as a tool or in some other arena.
Question, is ai art reproducible in exactness? Is there an randomness to it? Assuming all inputs exactly the same and you can reset it to a save point, will it produce the same or even similar images?
Sorry, meant to reply to the thread, not to you directly.
I’m not sure. I think each output is generated from scratch upon issuing of the prompt, so I think each one is technically unique (although they can look similar). I’m not sure if the currently available models are set so that they aren’t trained further on anything until the next version or whatever though, or if they “learn” with each prompt generation too.
Maybe @pfm or @odds can speak more to this, as they seem more familiar with it.
I think the royalties would make any generation prohibitive given the main use is corporate cost cutting and right now they just pay once and get presumably as many rolls as they want for the month. Each of those should results should be paid for whether or not they use them.
In fact, there would be no end to royalties being paid. It’ll become like Spotify but a lot more expensive because what music artist would consent to Spotify using their work and then generating millions of songs based on it in an attempt to replace them outright?
But yeah, if we ignore all that as well, then I still do have a problem with an artist using it to generate pieces of their “own” without actually doing any of the work. At that point, why bother? Might as well just sell your style and let others key in prompts and that final product would neither credit you nor the prompter, but once again the soulless algorithm that churned it out.
This is what I think: If AI makes art, I believe the only art that can be presented and admitted as art is the AI itself and be credited by whomever programed it. But no one can claim the art the AI created as THEIR art. It is instead a product of the artwork that is its programming.
At what point do we consider a tool’s product as art? We can use tools to make art but that is because we are present with them. We are a part of the process at integral steps. We are there at the beginning and we are there at the end.
With AI, it doesn’t require us to be present at all. We are only needed as a directive. If anything AI makes US the tool because it is doing more work than we are.
You can make the model deterministic (same input = same output) but that’s not a desirable trait.
For a LLM like chatGPT, there’s a bit of randomness when it chooses the next best token in a string of text being put together. You can actually adjust the “temperature” to control the randomness of the output.
Personally I think people can call AI generated images “art”, but its very lazy, and it really is unethical when it comes down to it. Without any of the actual artists AI would have no way to produce its own images. Its handy for automation, but it is kind of an unfortunate future for the art industry in general as I could see alot of people in companies turning to this as time goes on. Reading through the posts im sure most of us want whats morally right for artists even though its an uphill struggle because of how far AI has come.
Id be curious to know if people had to pay a subscription for AI generated images (lets say $100 a month) with proceeds going to whatever artwork was used to help generate it , would it still be as popular? If yes then everyone wins, but if those utilizing the service are unhappy they have to now pay then they should probably think it over again as most AI driven images are used for personal gain to begin with both good and bad.
@odds you have said multiple times that art is just an expression of human creativity.
This is simply not true if we are discussing capital-A Art. I would say 1% or less of human creativity would be considered Art. Children’s drawings are not Art, and neither are the concerts put on by senior’s choirs or pieces made by church pottery classes. They ARE expressions of creativity but they are not Art.
What defines Art is the current culture, usually in the context of history and with an eye for craft and skill.
Earlier you made a point about painted portraits vs photo portraits. What you missed though is the context those portraits were painted under. None of them were made to be hung in galleries, and certainly no one at the time would have thought to travel around the world to see them. Only now, looking back on pieces that have survived 100, 200, 300 years do we see the painter’s skill and craft in the context it was made and deem it Art. All the while they were originally made because the painter was hired by a wealthy client to paint their portrait which was to be hung in the home (or government building or church) only to be seen by people who frequented those spaces.
Another example, Bach is arguably one of the most important composers of Western classical music, but this wasn’t always the case. He worked for a church and part of his job was to write new music for every Sunday service. He would write it, the church choir and musicians would perform it, and then it would be shelved. It was never intended to be performed by the world’s best orchestras, and you would have only heard these specific pieces if you happened to attend the church where Bach worked. By the time he died his music was considered old and out of fashion, so much so that it was “lost” from the regular canon until it was rediscovered 100 years later in the 19th century.
Bach and these portrait painters were just working, but from our point in history we look back and see the craft and genius in their work and deem it Art. I doubt anyone in 300 years is going to look at ai prompted images and do the same.
I think the key word here is “creativity”. As someone who has been known to have creative moments, but who has little artistic talent, this stands out to me. You can absolutely use your creativity to generate ideas or prompts to feed to AI. I would credit that whoever does this is “creative”. However art is an expression of a person themselves - it represents their emotions, their personality, their individuality. Using your own creativity to create art makes something that is uniquely you. Using your creativity to make art with AI removes the personal touches and takes away the life of the art. It’s like if I just said “oh, I have an idea for a book” then just plugged it into AI and published the first thing that came out. It might be creative, but it doesn’t have a pulse, especially not one unique to me.
If AI becomes normal in pokemon card art it would mark a change in era sufficient to have a before and after.
I would lose a level of interest in collecting personally. My wife is an artist/illustrator. It’s night and day as a skill set of drawing vs inputting the right words
Technically speaking each output is unique, despite an output having the ability to be almost identical.
A tempature parameter of an LLM is the ‘creative’ parameter. Which simply put is the amount of randomness allowed, which should result in a range of outputs for similar or same inputs.
AI goes through a training phase prior to inference. There are self-training AI but to my knowledge thats on much smaller scopes & doesn’t apply to the current public state of AI. CGPT as an example is only trained prior to new relases.
For a long discussion about whether or not AI art is “art”, there is very little attention given to the actual definition of “art” until here
I think there is a reflexive distaste for AI images and therefore a general rejection of them as art. But I can easily imagine definitions of art where AI art is included and definitions where it is not. I think it’s less about the question of what qualfies as “art” and more about what AI art lacks that other art doesn’t. For example, the primary reason why submitting AI art to an art contest is wrong is because the creative vision, composition, colours; all the artistic skill and merit was not performed by the submitter. Not because of a semantics game where AI art doesn’t qualify as art.
Again, it will depend on the definition of art and I have no strong stance either way because I don’t have a strongly held opinion on what the definition of art is.
The one thing I’ll mention is that history suggests this problem is not new. Lots of technological advances over time have moved tasks that were solely in the realm of artists to something more utilitarian. The printing press and digital printing allowed books to be created without needing to hand scribe and paint the pages. Cameras allow instantaneous image capturing of people and landscapes that used to only be possible by painters and portrait artists.
In these examples, what ends up happening is a delineation of art and utility. We produce books at scale but people still practice calligraphy. I would consider the photos on my phone to be more utilitarian than artistic but people still paint portraits and landscapes. For better or for worse (and I think many will say for worse) the production of digital images has become/will continue to become a utilitarian task. I’m not going to wade into the morality of how it happened (theft vs learning, there are active legal battles on this) and I really do feel bad for digital artists who’s work is now effectively devalued as a consequence. But I think over time there will be a delineation of doing digital art as an artform vs the utilitarian act of, say, quickly producing an image for a presentation you have to give. Basically filling in the role of clipart or google images.
Again, I really want to make it clear that I’m not making a value judgment. Just trying to get to the reality of the situation we find ourselves in. There’s an accessibility feature about AI images I like while simultaneously feeling really bad for graphic artists.
There is one assumption that I think we can make and that most of us agree on:
AI chatbots and LLMs creating “art” is never art. I think AI enhanced tools however are not AI in the same sense, and should be considered separately.
However, I’m generally agreeing with PFM, I’m not seeing @odds at odds with 10 other people. There’s not much agreement here at all on what art is. The thread is more about AI bot “work” in general, but neither a specific use case as odds describes, nor a general concept, leaving a lot of blanks for our emotions to fill-in…
@odds Says this, I think.
I personally need to know more about odds’ process.
So, @odds , how exactly do you use AI?
Assuming you are using the tool to fabricate your own work, I can’t agree with @dyl that you are replacing an artist. You could just as well, though not as easily, learn the skills to create the art yourself, and thus you’ve created an artist: You. Therefore, replacing an artist is not what you are doing, you are a new artist. But this is where the full context is important. If you’re collaborating with the living artist, then something in the creative collaborative process is lost. @Dyl offers this exception himself, I think.
I agree with @dyl, except I’ll say we are already sliding down the slope, and will have to claw our way back up.
So, @odds , how exactly do you use AI tools?
What do we mean by digital images?
Any photographed work becomes digital. Like Yuka Morii does clay sculptures but that is then photographed and digitized. AI should then be able to steal all of those and reproduce the style.
Also re: photographs and painters, I’m not sure it applies in the same way. Any painting meant merely as a recreation would never be as accurate as a photograph. I think @azulryu said it best, AI is not a tool, it is the author and we are the tool. Except the author would reproduce nothing if it wasn’t fed any stolen data to begin with.
So honestly, even with the thread title, skipping ahead to discussing what art is is already a win for techbros for getting you to accept/ignore that their model is built off of exploitation and theft. We shouldn’t even reach that because fundamentally it should not exist in its current state.
And let’s not pretend there are no other externalities. Each query/image generated uses up resources somewhere in the world. Like many other products and services today, the actual costs are not reflected in the price. AI has its place in actual useful products and services, mainly in research, data, medicine but anything else is unnecessary.
And, yes, of course everything from coffee to phones are exploitative as well but AI, like NFTs and crypto before it, pretending to be miracles bestowed upon us is what rubs me the wrong way.
What a digital image is is pretty straightfoward. Any image file you can store as bits. Yes, Yuki Morii’s photos are not excluded but the clay is.
It’s not about being a 1:1 recreation. The purpose of painting a scene or a portrait is to capture the likeness of something in real life, similar to photography. I don’t have to hire a portrait painter because my phone has a camera on the front.
I purposely didn’t get into the theft argument because it’s mostly irrelevant to the topic at hand. These model can still exist using 100% non-copyrighted materials and/or pay to use copyrighted material. The “theft” is a choice companies make, not something inherent in the technology.
This is a silly argument. Every Google search you do takes up resources. This site takes up resources. You take up resources. Hiring a graphic designer takes up resources. Based on what I could find, ChatGPT has a real cost of about 36c per query and uses 15x the power a Google search does. Just like all things, sometimes the cost is worth it sometimes it’s not. All this also assumes this nascent technology will stay this price forever despite the incredible incentives for companies to minimize the cost over time.
That’s your own personal feelings. I can attest that LLM have been extraordinarily helpful to me. Lumping it in with crypto and NFTs is just an association fallacy.
