Considering AI art generators like Midjourney, how do we define 'originality' and 'authenticity' in art?

Question

Midjourney, creates art by learning from a vast collection of human-produced images. Some argue this process diminishes the authenticity or originality of the resultant art since it's derived from pre-existing human works. They say it's a form of plagiarism.

However, others say to this - human artists also learn and draw inspiration from the works of others, nothing can be created from nothing, suggesting that the AI's process is not fundamentally different. That humans are also neural networks, just biological ones.

This raises some philosophical questions.

• What distinguishes an AI's learned creation from a human artist's work, if both are informed by pre-existing art?

• Can we consider AI-generated art as 'original' or 'authentic'?

• Should we change our definitions when AI art becomes even better (it's already pretty damn good)?

• Why Art created by biological neural networks is Art, and electronic neural networks is not art? Are we just protecting our world from AI, cause we are biased?

Answer 1

You might like this

Artificial Metacognition Systems (AMS) is a field that aims to develop AI capable of monitoring, evaluating, and regulating its own cognitive processes. This innovative discipline combines insights from cognitive psychology, self-aware computing, and advanced machine learning to create AI systems that can think about their own thinking... Even in creative fields, an AI that reflects on its creations (say, evaluating the novelty or quality of its generated art or music) could iterate towards more innovative outcomes without human critique at every step.

I would think that creative meta cognition can be assessed as original, especially through iterative processes, though obviously they have no direct human agent.

Answer 2

Trying to figure out what kind of art AI art generators generate is a category mistake. They don't generate any kind of art if one uses the standard definition for art.

MW's definition is adequate: Aesthetic objects produced by the conscious use of skill and creative imagination. (Definition 4, rephrased for brevity.)

It is somewhat weak, because aesthetic object is underdefined. Objects or processes produced by the conscious use of skill and creative imagination for the purpose of evoking aesthetic experiences would be identical but more clear.

AIs do not have consciousness, skills, creativity, imagination, or intent. Therefore AI art generators do not generate art, they generate things that look like art. This is not a new phenomenon. Natural environments look like naturalistic art. People look like portraits. The Smile Nebula looks like a crude smiley face. Nebulae make an especially good example because most of them don't look like anything until you input the right inputs into an expensive, complicated machine (a space telescope and its image-enhancing, color-magnifying software).

Nor do humans using AI art generators generate art in the first instance: they discover things that look like art in their environment. They do create art if they do anything skillful, creative, and aesthetically intended to the generated picture (like cropping it and putting it in a frame), exactly as one can create art by drilling a hole in a pretty shell and putting a string through it to wear it as a necklace.

Imagine that we lived in a time and place far from the sea with no trade to the sea, where experience has taught us that pretty baubles that you drill a hole through and wear as a necklace must always be crafted by human artistry. Then making your shell necklace is still art, but telling people that you made your shell necklace yourself, without explaining that the shell is just something you searched for on a beach, is a lie.

Since AI art generators are labeled as they are and advertised as a way for the user to be an artist, I suspect that in the very near future, the definition of art will have changed to include "or machine-generated objects or processes evocative of the above". I predict that we will soon have a new word denoting art of conscious origin to distinguish it from new-definition "art" in general - almost all of which will be generated by machines. (As with music vs live music, in which the original stuff gets the compound word, while the recorded stuff swiftly usurped the original word because it's much more common.)

Answer 3

Defining "originality" and "authenticity" is quite straight forward. The first is concerned with the question of whether or not something is the origin of something or a reproduction or derivative of it or something else. While the latter is concerned with whether or not something is "real" or "genuine".

Now the much harder question is: What is art? For example:

• Does it refer to things that are aesthetically pleasing?
• Is it defined by uniqueness and originality?
• Is it defined by (high) skill?
• Does it serve a purpose?
• Is it a goal in itself?
• Is it decoration or exploration?
• Ought it to be "authentic"/natural?
• Is fiction and imagination required for art?
• Does art require an artist?
• Does art require an observer?
• When art is what an artist makes, what makes one an artist?
• All of the above? None? Some? More than that? Less?

Like depending on the definition beautiful things for example can exist in nature independent of a (human) artist. Originality, skill and exploration might also be found in science, engineering and other crafts which are often not given credit for their creativity. Technically uniqueness can also be achieved without skill, though few people give that the appropriate credit and just because something is new doesn't mean people consider it a piece of art. Similarly if something serves a purpose it often seizes to be art, but something completely random is usually also seen as pointless.

So without raising a claim to completeness, I'd argue that:

Art is the original attempt of an individual of perfecting something.

So in that sense art does follow an aesthetic ideal of the artist that they try to achieve as best as they could, but that doesn't necessarily mean that it has to be "conventionally beautiful". Also often times aiming for perfection does require a certain amount of skill and effort. But idk if your ideal is just to broaden people's perspective of how we look at things, then also idk the black square which requires very little skill or even "fountain" which is just an ordinary pissoir bought off the shelf and which requires no skill at all, can serve that purpose. Though in that case the art isn't so much the object itself and its inception, but the context in which it is presented and the statement that the artist intends to make with it.

So doing the same thing sounds deceptively simple, but might actually be impossible. Like installing a light bulb and a mirror in Plato's cave is levels more elaborate and skillful than to just untie the prisoner, but it would still fall short of having the same revelatory effect.

So art usually isn't just about pleasing some aesthetic ideal of the masses, but rather seeks to go after a particular goal where no one else has gone before. Lots of art for example, is concerned with making emotions perceptible, finding the right notes, harmonies, words, rhythms, images and whatnot in order to create in the observer an effect similar to what you try to describe. Finding a language for which there are no words (yet).

So the problem with plagiarism isn't just that it takes away monetary value from the original artist. And yes you can have that discussion about whether or not something can be truly original in the first place. But even if it is a "perfect copy", it's possible that the change in context and intent of the statement just makes it inferior to the original as it lacks the non-object bound parts that have been replaced.

So idk, by now people probably already expect artists to be a bit off their rocker, but if a few hundred years ago someone deliberately decided to risk their income and reputation by wasting perfectly fine canvas in order to draw some butt-ugly nonsense, then people might have asked themselves "why"? And I'm not sure you get that same response for AI where it's essentially the definition of the algorithm that they aren't going for something in particular but just try to push the buttons that are statistically meant to be perceived by viewers as "positive".

So you're not thinking "what did the artist try to achieve with this", but rather "yea" or "nay". Because apart from tickling an undefined itch, there's no deeper meaning to be found. No intent, just statistics. And for the people interested in statistics it is interesting for other qualities than beauty.

Of course an "artist" could prompt the AI to generate art for them and thus trigger followup questions, but the problem is that unless they fine tune the result so much that it's a skill in and off itself, their "art" will just end up being mediocre and either confusing or just the aesthetic equivalent of "elevator noise" (nice inoffensive and forgettable the moment you step outside the door).

So "authenticity" in the art sense, might not just be about who's been doing it, but also about whether or not they have put some serious thought into it. Or whether they describe a real world observation and thus help out people in a similar situation to make heads or tails of it or whether they just imagined something like that and the reality is completely different from that imagination.

Also as said "originality" somewhat covers the societal problem of how to deal with artists, as that is either the best or worst job imaginable (and quite often is both, with high highs and low lows). As well as the problem that art doesn't just include the object or performance but also the conversation via context of the artist and the observer and especially the latter might be crucially missing with AI generated art even if they'd be capable of (re-producing) objects that are in some regard similar to what people like and learn to improve that.

Answer 4

Your question is entirely a matter of opinion. An appreciation of art is inherently subjective. There are many accepted artists who produce works through a variety of very technical methods, so a person might be inclined to consider a work produced by a bot to be art, and if they did you could argue until you were a fetching shade of blue in the face that they were wrong, but they would be quite entitled to ignore your pompous expostulations.

Answer 5

The following approach could help you understanding what creativity (ergo, originality and authenticity) really is. Before, some context is necessary.

What is creativity

Most jazz musicians aren't creative. Improvisation in jazz means essentially creating new melodies over an established harmony, kind of like making creative comments about an established subject. Well, although jazz musicians improvise their solos, the result is 99% of time just imitation.

Creativity in jazz is enormously difficult to achieve. What jazz musicians tend to do is just repeat complex recipes that they've mastered for decades. They are not new, they are just variations over well-established recipes. Not only an untrained ear but even a reputed musician could perceive that an improvised solo is really presenting an original idea. Most of the time, it is not. Jazz musicians that improvise in music tend to just repeat recipes.

And improvising by following known recipes is not even intentional. When a musician learns to play an instrument, he is essentially acquiring a number of motor memory facts. He is not leaving his fingers to act creatively, to move in chaos, but in a well-established order (without that, for example, a musician would not be able to perform a natural scale: all note combinations are just motor patterns).

Even when a musician tries intentionally to play "outside", to make his fingers behave in ways that are not established in his brain and motor memory, he might not be able to reach real freedom. Playing outside has the same complexity of making a software program to create a random number (See https://en.wikipedia.org/wiki/Random_number_generation).

Then, how would someone knows that a jazz improvisation is real, that is an original idea, that it is authentic? It is very difficult to know when a musician is really improvising, but there's an answer:

A well-trained musician is improvising creatively when he fails repeatedly.

It was when I found out I could make mistakes that I knew I was on to something.

Ornette Coleman

Failing shows that the musician is really exploring unknown territory. So, there's naturally a perceivable failing rate. The last time I saw Pat Metheny playing (Marciac, 2023), I've counted three creepy failures (I have a good knowledge on jazz, guitar playing and music theory). I knew the meaning of that: he is constantly exploring, trying new ways, breaking rules. A guy who has 20 Grammy awards can play without any error if he wants to. But he chose to break his ways in such particular event.

Are AI Art Generators Creative?

Following the notion we've just developed, definitely, NO.

On one hand, generative AI is just a probabilistic result. It is the equivalent of splitting billions of sentences in parts and then randomly mix them.

On the other, generative AI is not creative by definition, since it generates a product based on previously learned patterns.

That was simple. The complexity comes here: when will we know that a computer AI generator is really creative? I would say when it be able to proactively follow the process of exploration, trial and error, just like a human. The result should raise with an intention (proactivity), not as a request (reactivity), be the result of a desire (not a request) and involve exploration in multiple disciplines.