Machine learning creates professional level photographs

[Norman 202]

oh-kay, supposing art is a manifestation of the soul and as our art has expanded (we've moved on from caveman art ) does that mean the soul has expanded. and is so, how far can it expand?

[Norman 202]

can it expand sufficiently to recognise that art and soul and humans are not joined at the hip? does the soul have free will, to associate itself with beasts, machines, germans?

[Norman 202]

You can't just look at art and remove it from all of this or reduce mankind's artistic achievement to chemical reactions in the brain.

I can, given my experience with chemical reactions in the brain :-) and I can't believe there isn't a school of philosophy that agrees with me. But I can offer no proof so it's back to wikipedia for me.

[Norma Desmond]

Norman, there is. Philosophy of Mind. A good starting point is Nagel's short paper, What It's Like to Be a Bat, which lays out a well-conceived and well-written defense of there being something more than what scientific materialism can address. Then, there's a collection of essays edited by Ted Honderich, I think called The Philosophy of Mind that presents both materialist and dualist positions. I studied this stuff about 15 years ago but imagine these articles are still relevant and may have been updated. Not much talk of souls, even in the non-materialist papers. Bible's always good for that. Read it in my twenties and some portions since then.

[Norma Desmond]

Here's how I'd put it. Soul may be required for some people but not others to theorize about art. It's not necessary to make art, to experience art, to appreciate art, to hate it or to love it. Soul probably suits religious or faux religious dogmas and maybe some academic idealists. It is a concept that's been dealt with with in art and other endeavors a lot. So understanding it when it's addressed specifically might be helpful. But Monet can paint a water lily even though he has no soul and John and Jane can swoon over that water lily, discuss it, throw blood on it, or pay millions of dollars to own it even though they don't have souls. I can read The Divine Comedy and will understand it better if I know how the word "soul" has been used and what it means to Dante and others. I can do that even though I don't have a soul and even though I don't buy into the idea of souls as a reality. Plenty of people do it all the time, just as plenty of people of faith are bright enough to understand what agnosticism or atheism is as well as being feeling enough to feel something for art made by an atheist or agnostic if they're open enough even though they believe in souls.

[Supriyo]

Wait a minute, I thought the word soul as it is used in this context means essence of something (in this case artistic artifacts), not the immortal spirit. Thats how I was following the discussion so far.

[Norma Desmond]

That would only be true if you want it to be true. There's a difference between saying that Monet doesn't have a soul and saying his painting has soul. I think Aretha has soul and Helen Reddy doesn't have much. I think neither one of them has A soul. Having a soul is a historically religious or spiritual concept incorporating the idea of inviolability and immortality. Having soul is a much more recent and informal way of talking about having feeling and often being looser rather than tighter in those feelings. I can honestly say until now I've never met anyone who didn't know the difference.

[Supriyo]

Which to me would be the same as saying Monet's painting has no soul.

 

But how can one detect looking at a soulful and soulless artifact side by side, which one is which, and if its not possible, does it matter whether it has soul? Isn't that the core of this argument, that AI machines can create something that will intrigue or fascinate us, just as a human artist can, even if the machine is not self aware. If there is no clear way to detect whether an artifact has soul in it, does it matter?

 

My own take is, the soul (I treat it as essence) is provided by a human by picking out a machine's work that appeals to him (I know Fred disagrees). Soul is subjective too.

[Supriyo]

This is quote from gotquestions.org:

 

The soul and the spirit are connected, but separable (Hebrews 4:12). The soul is the essence of humanity’s being; it is who we are. The spirit is the aspect of humanity that connects with God.

In that sense, spirit is more connected to religious beliefs than soul.

[Wouter Willemse]

Obviously man's capacity for making art has to do with our biology and with the way our brains evolved.

 

Sorry, but how is this obvious? Because to me it isn't obvious, would it that our brains changed causing a move from classical music to jazz, to pop music? Or that we can no longer make caveman-paintings because we started to walk more upright? Would it mean that our increased capacity to program a machine may enable us to share our ability to make art, and have a machine make art?

And since you're convinced man is the only species capable of making arts, isn't the constant evolution of all species a potential enabler for other species to start creating art too?

[Norman 202]

Or that we can no longer make caveman-paintings because we started to walk more upright? Would it mean that our increased capacity to program a machine may enable us to share our ability to make art, and have a machine make art?

the thing is we can still make caveman paintings, we haven't lost that ability which suggests to me our creative core (soul, brain, call it what you like) is flexible(as i suggested earlier). it can grow and shrink but can it jump between species: man->machine?

[Wouter Willemse]

but can it jump between species: man->machine?

 

That is the question, and since we absolutely have no idea about the answer (as it is for now pure Sci-Fi), I find it baffling so many are inclined to say "no" or reduce the definition to "human only" already. I find that excluding future possibilities, for no good reason.

[Wouter Willemse]

Phil, you keep repeating that art is by definition human as a reply whenever I contest this assumption that art is by definition human thing, without further explanations - it's getting a bit tiresome. It is not a fact that art is intrinsically human, it's is just your opinion (or the definition of art you adhere to). And your opinion isn't fact nor truth, nor does it give me any good reason to put limitations on the definition of art. In fact, it gives no reasons at all, it's just a statement. It does not convince, and throughout this thread it should have been painfully clear I don't agree to your statements, so continuing to repeat them is either an insult to my intellegence or a blind spot on your behalf - neither especially welcome, I'd say.

 

So for what arguments is art intrinsically human? What is useful about restricting the definition of art, which reasons do you have to prefer this limit?

[movingfinger]

The interesting discussion and comments being made in this thread are caused primarily by the drift into the definition of the word 'art'. The most general definition of 'art' in the Oxford Dictionary (on-line) is "works produced by human creative skill and imagination." As I read all the comments here I think it's fair to say everyone agrees with that definition and has been working from that premise. But that definition, like the word it tries to define is itself vague and therein lies the rub. We all draw the line differently on where human creative skill and imagination stops and other forces take over. Take tools for example, functional implements to perform human-desired tasks otherwise impossible, unpleasant or difficult. The design and construction of a tool can be considered a work of art in itself. So based on where I draw the line the Leica film camera, certain Japanese hand saws, a 1953 Chevy Corvette, to name just a few, are all works of art. Other tools, in fact most, even from the categories I illustrated (camera, saw, car) do not make it to the category of 'art' in my opinion - your opinion may vary.

 

Leo_p.. opined that computer programmers are artists and to an extent, I agree. There are computer programmers who are true artists at the craft. In fact, being a mathematician/computer programmer myself (alas not an artist in the field), one of the most comprehensive series of text books in computer science is a 5+ volume series titled "The Art of Computer Programming" written by Stanford professor Donald Knuth. This 5+ volume series is itself a work of art (and despite my open mindedness I will accept no counter argument that it is not :)).

 

Now we move to the next level of abstraction. Humans make tools and then humans use the tools. Some users employ a tool mechanistically to peform a function the tool was intended to perform and others bring their creative skill and imagination in the usage and thus create art with that tool. Where this line between mechanistic and creative is drawn is again subjective and in the eye of the beholder. Someone can use a handsaw and create furniture that is to me and many others a work of art, and I can use it to cut a board in two and cobble together a functional table but the result is, alas, not art (by anyone's standard, even my parents :(). Someone can use a Leica camera and create art, someone else can point it somewhere and produce a stale representation of what it was pointing at.

 

Now with Artifical Intelligence (AI) we move up to a third level of abstraction. In AI we are creating tools not to then be used directly by humans in order to facilitate some effort and in which the human user's creativity can be directly brought to bear in the resulting product. Rather with AI we have tools intended to go off on their own using the independent assimilation of data and then to evolve in their behavior based on that data. Yes, that evolution progresses based on the human designer's algorithms but once set in motion the human is removed from the effort. So while I would clearly agree that an AI's algorithms and design might be considered works of art since they are the result of a human's creativity and imagination, the products produced by an operational AI process being removed from subsequent human input and intervention, for me, DO NOT meet the threshold definition of "work produced by human creative skill and imagination". In fact I would argue that an AI program creator should oppose the calling of work produced by his/her AI program 'art'. Given the definition of art (see above), since the program creator is the last human in the chain, to call the resulting product 'art' means that the programmer has direct influence on that product. If the AI machine is really AI then the human builder should not have influence on the product of that AI machine. It would be akin to calling the parents of child X producers of 'art' based on that fact that their offspring X has produced art. Here the mathematician in me comes out, for 'X' you can substitute Michaelangelo, Cartier-Bresson, or whomever you consider to be a producer of art. I see that earlier in this thread it was asserted that one's children are 'art', I disagree with that, but I suspect that most all would agree that one cannot claim the art produced by our children as proof of our own artistry. Thus an AI developer, who truly believes his/her machine is AI should not claim the output of his/her AI machine as his/her own creativity and imagination.

[Norma Desmond]

Now with Artifical Intelligence (AI) we move up to a third level of abstraction. In AI we are creating tools not to then be used directly by humans in order to facilitate some effort and in which the human user's creativity can be directly brought to bear in the resulting product. Rather with AI we have tools intended to go off on their own using the independent assimilation of data and then to evolve in their behavior based on that data. Yes, that evolution progresses based on the human designer's algorithms but once set in motion the human is removed from the effort. So while I would clearly agree that an AI's algorithms and design might be considered works of art since they are the result of a human's creativity and imagination, the products produced by an operational AI process being removed from subsequent human input and intervention, for me, DO NOT meet the threshold definition of "work produced by human creative skill and imagination".

I agree very much with your reasoning and you've articulated it well. I disagree with your conclusion, because it is a case of begging the question, which is a logical fallacy of assuming the conclusion in the premise of the argument. It's what Phil has done, and some others, throughout this discussion. Please read the Wikipedia entry on "begging the question." While it's interesting, the HISTORY section of the Wiki entry is dense, so you can skip down to the DEFINITION section.

 

In short, what you've done is this. You asked the question whether an AI machine can produce art, a good question. Then you've shown how removed the end product would be from human hands. But you began by assuming your conclusion (that it doesn't meet the threshold definition) in your main premise (which defined art as "a work produced by human creative skill and imagination.") In other words, you really didn't ask the question at all, because you defined away the possibility of your question being answered any other way from the beginning. You can't (in a logically valid way) ask if machines can create art, then immediately answer the question by defining art as something a machine can't make, then go through a long explanation of what a machine does that sounds independent and a whole lot like it's independently creating something that could be called art, except it can't be called art because of the very first premise you already put forth, which is that art is made by humans.

 

What you were actually asking is, can the old definition of art be expanded (which is something that pretty much has happened with the definition of art throughout history)* to include something that an AI machine makes. That was a good question. But you immediately disallowed a valid approach to that question by, instead of first exploring whether a machine could make art, you simply defined away the possibility before ever exploring the question.

 

*The Stanford Encyclopedia has a good entry on the definitions of art throughout history, the difficulty of doing so, and issues around words like art whose definitions seem to have to allow for expansion in the future and problems that ensue from that.

[Norma Desmond]

Just want to add something. I said:

 

"What you were actually asking is, can the old definition of art be expanded."

 

If you were not asking this, my apologies. But I don't see any other reasonable thing you could have been asking. If you weren't asking whether art could be expanded to include AI (machine) creations, then you were simply asking, "Can machines create art?" Now, let's just substitute in your question the definition of "art" your dictionary provided and you agreed to (a little longer way of saying art), and we get, "Can machines create something which only humans can ever create?" You've already answered the question in your question. That's begging the question. The question has already been answered in the question itself and no discussion can change that.

[Supriyo]

Fred, I like this last series of arguments. Although I didn't have time to think about it too much, movingfinger's argument doesn't fit the criterion of begging the question for this reason: begging the question here would need an additional premise that all machines are autonomous and they can do things pretty much independently of humans. Only if you assume this, then only the argument becomes fallacious. However, I don't think movingfinger included that additional premise. He later on argued that since AI machines are almost autonomous, their product doesn't meet the criteria of art.

 

BTW, I am not agreeing or disagreeing with movingfinger's standpoint, but I like to understand and think about logical fallacies, hence the comment.

Edited July 28, 2017 by Supriyo

[Norma Desmond]

Supriyo, I suggest you consider it further. The argument wouldn't need the premise you're adding.

 

He asked if machines could produce art and defined art as something machines could not produce (only humans can produce art). That is begging the question.

 

I think we all agree that humans and machines together can produce art. But, look at his argument. It's when he shows that the machine is removed enough (independent enough) from the human that he determines his conclusion, based on his original definition of art. That's circular and it begs the question.

 

In other words, once he gets the human out of the AI picture (which I think is the valid part of his reasoning, because the machine is operating independently at that point, according to his own words), he then resorts to saying that, because art is defined as something where the human can't be out of the picture, the result can't be art. He needed no reasoning.

 

If he didn't get the human out of the picture when he showed how removed the human was from the final product, then we'd simply have a case of art where the human uses the machine to create art. But his conclusion says "it's NOT art" and his conclusion says t's NOT art because there's no human, which is the way he had already defined art.

 

Here's his argument, in his own words, boiled down:

 

"the products produced by an operational AI process being removed from subsequent human input and intervention, for me, DO NOT meet the threshold definition of "work produced by human creative skill and imagination"

 

To me, it's a textbook case of begging the question. Take it to your local philosophy teacher or logician and ask! ;-)

[leo_papandreou1]

"If the AI machine is really AI then the human builder should not have influence on the product of that AI machine."

 

Is "really AI" strong AI? Because strong AI is functionally human (at least) intelligence. Toddler intelligence would be mission accomplished, AFAIC.

[leo_papandreou1]

We wouldn't be able to distinguish a strong AI from a human based on either's work. That's the point of strong AI.

 

(Has anyone seen what Banksy looks like?)

[Supriyo]

Supriyo, I suggest you consider it further. The argument wouldn't need the premise you're adding.

 

He asked if machines could produce art and defined art as something machines could not produce (only humans can produce art). That is begging the question.

 

I think we all agree that humans and machines together can produce art. But, look at his argument. It's when he shows that the machine is removed enough (independent enough) from the human that he determines his conclusion, based on his original definition of art. That's circular and it begs the question.

 

In other words, once he gets the human out of the AI picture (which I think is the valid part of his reasoning, because the machine is operating independently at that point, according to his own words), he then resorts to saying that, because art is defined as something where the human can't be out of the picture, the result can't be art. He needed no reasoning.

 

If he didn't get the human out of the picture when he showed how removed the human was from the final product, then we'd simply have a case of art where the human uses the machine to create art. But his conclusion says "it's NOT art" and his conclusion says t's NOT art because there's no human, which is the way he had already defined art.

 

Here's his argument, in his own words, boiled down:

 

"the products produced by an operational AI process being removed from subsequent human input and intervention, for me, DO NOT meet the threshold definition of "work produced by human creative skill and imagination"

 

To me, it's a textbook case of begging the question. Take it to your local philosophy teacher or logician and ask! ;-)

 

Fred, I agree with you, I was wrong, and thanks for explaining further. Yes, his argument assumes that art is a human centric activity, but thats begging the question, why art is so. Is it only because dictionary or wikipedia states so?, which is assuming the conclusion.

 

I think people have tried to argue this point though in this thread, why art is human centric. Phil referred to soul and consciousness. My take is, consciousness seems to be relevant at some point in the creation of art. I disagree that only conscious beings can produce art, though. I wonder, if humans never existed, what happens to those things that are dependent on abstract thoughts and emotions. Do they still exist.

[movingfinger]

In short, what you've done is this. You asked the question whether an AI machine can produce art, a good question. Then you've shown how removed the end product would be from human hands. But you began by assuming your conclusion (that it doesn't meet the threshold definition) in your main premise (which defined art as "a work produced by human creative skill and imagination.")

 

Fred: I don't believe I began by assuming the conclusion. I did begin with a definition of art as a premise, or call my definition a postulate if definition is too rigorous. Consider this trivial example from geometry. We are debating whether or not a square is a triangle. I begin by defining a triangle as a plane figure with three straight line sides. That is my premise and it is made independent of the 'square' under contention. Then I proceed to explain that a square has four sides and why this seems to me to be the case. If I've successfully demonstrated to you that a square has four sides, AND IF you agree with my premise on what it takes to be a 'triangle' then I've demonstrated to you that a square is not a triangle. I don't believe this is begging the question. Also I believe this is how I proceeded in the question at hand. I put out my definition of 'art' independent of the issue under dispute and then tried to show why it seems to me that an AI generated product does not fit that definition. Yes, it is critical (vital, essential, etc, etc) that we agree on the definition of art in the first place. If we don't then of course we will disagree on what qualifies as art.

 

What you were actually asking is, can the old definition of art be expanded

 

Actually it seems to me this is what you are doing, expanding on the definition of art that I provided. I am perfectly content with the "old definition". For me, the "old definition of art", is the definition of art, period. I will respond that what you are asking is can the definition of art be changed to include AI generated products. That is a valid response to our conundrum. Another approach, my preferred approach, is to create a new word to deal with such products. The information age has seen both responses to deal with the many new and exciting issues created by new technology, many new words have been created ('email' instead of 'mail') and many old words have had their definitions expanded. For something as fundamental and historic as 'art' I prefer a new word for AI generated products in order to keep distinct those products generated with a more direct link to human creativity and imagination from those generated by machines ingesting data and algorithmically spewing forth some product it 'thinks' I might also enjoy.

 

I can see it now, I go on Amazon and buy a book of photographs by Ansel Adams, or google sees me searching on the name Ansel Adams, and before I know it thumbnail images appear in my browser, machine generated from scans of millions of landscape photographs or perhaps even simply AI machine generated shapes, with the following teaser "Since you like Ansel Adam's you might also enjoy these fine art images". I will go out on a limb and state here and now for all to see, those images will not be 'fine art' ...but that's just me, your results may vary ;)

[Norma Desmond]

Movingfinger,

 

This is your example:

 

1) A triangle is a figure with three sides.

 

2) A square has four sides.

 

3) Therefore a square is not a triangle.

 

That makes sense.

 

But, then you're comparing that to this:

 

1) Art is something human-made.

 

2) AI is not human made.

 

3) AI is not art.

 

They're not comparable.

 

In the geometry example, the definition of square and triangle are INDEPENDENT of each other.

 

If we put your geometry example in the same sentence structure as we put your art example, we get this:

 

1) A triangle has three sides.

 

2) A square does not have three sides.

 

3) A square is not a triangle.

 

The way you know a square does not have three sides is because you know a square has four sides, not just because you know a square doesn't have three sides. In other words, you know something INDEPENDENT about the SQUARE and INDEPENDENT about the TRIANGLE that make them different.

 

Yet, in the art example, you don't suggest two different attributes that art and the AI object have that would make them different. You are simply saying that what makes them different is that one is man made and the other is NOT man made, so you're dependent on man-made or not-man-made in both cases.

 

You've said the triangle has three sides, which makes it different from a square that has four sides. What is it about art that makes it different from an AI object which is not man-made? You haven't said. All you've said is that art IS man made. So you are using the negation of your definition of art as key in the art example but you are not using the negation of your triangle definition as key in your geometry example. The key to you geometry example, and the reason it doesn't beg the question, is that you've supplied something else beside the negation of your definition of the square which proves that a square doesn't have three sides. The proof that a square doesn't have three sides is that it has four sides. You don't have a comparable independent proof for your definition of art.

 

In the geometry example, you aren't simply negating the definition of a square to reach your conclusion. You're affirming a definition of a triangle. One has three sides. One has four sides.

 

In your art example, you have not affirmed a definition of art independent of the definition of AI. All you've done is used one as the negation of the other.

 

Whereas a triangle has been defined independently from the definition of a square (they are two distinct postulates: a square has four sides; a triangle has three sides), AI and art have not been defined by independent characteristics. They've only been defined by the negation of a single characteristic.

[Norma Desmond]

Phil, I guess you missed the part where I was responding to Movingfinger, who used the example of triangles and squares. Neither of us reduced anything, but you were probably too busy laughing to realize that. We were analogizing the logic of the geometry example and the art example, not the substance, to see if the question had been begged. My point, which I don't expect you to understand or care about and the reason I didn't address you when making it is that I don't care wheter you understand or care of not, is that you and Movingfinger used only your provided definition of art and the negation of that definition in determining whether an AI-created product could be art. His example of geometry validly used two independent definitions, one of a triangle and one of a square, to show they were different. Because you and Movingfinger did not independently define the two things you were looking at but used the negation of a single definition/characteristic in defining the other, you begged the question.

 

As for intuition, you simply use it as an excuse against logic, not as an accompaniment to it.

[leo_papandreou1]

The genealogy of art as a human activity is almost surely the (not entirely correct) belief that it sets us apart in the animal kingdom. It's much more an assertion of our place in ye olde chain of being than it is an essential definition of art. Not that it matters, pending strong AI or visitors from outer space ...