I feel like we won't be able to say that AI is creative until we make an AI that is not intended to create stuff suddenly say "I want to make art" and then does so on its own.
honestly i feel you nailed it on the head. Art is about intention and so long as the ai is told by humans to make art, I don't really know if it qualifies as art. But if instead it goes and makes art on its own than that is art
great point. I was thinking the other way around- I would consider the AI an artist if the AI programmed to make art could say, "not today... I'm not feeling it". The AI would have to be conscious so far as to make decisions that don't have to do with its programmer's intent. But that wont exist until we have a working theory of consciousness and an AI built on that.
Interesting point. Those creepy deep-dream eyes and stuff that are everywhere now, they were not intended to be art. It was a neural network, trained to recognize objects. Programmers just turned that network inside out exposed the inner processes. Can this be considered to be art? It's generated by algorithm that wasn't intended for art but it turned out that it has some art inside its brain.
I suppose, but only if the A.I. was capable of doing such things before. In the sense that it can't go beyond it's programming because in my opinion that'd be like going beyond your genetics and suddenly sprouting wings.
"If you have to ask if it's art, it probably is." If you have to dismiss the question to get to your preferred answer, your preferred answer is probably difficult to actually support.
This is the beauty of subjective experience and "Art as Experience" as put forth by John Dewey. Or with Idea Channel's Aesthetic vid they linked. Danto's definition of art and artist are declarative statements that essentially need no further supporting documents or evidence to make that claim. Whether someone is a GOOD artist or makes GOOD art is a very different conversation, one that takes whole networks of people's opinions, culminations of lived experiences, and a community of criticism to build the evidence that it is good art. With Dewey, it is very much about the viewer's conscious framework of experiencing an artwork (like, are they self aware that they are experiencing an artwork?) and their expectations for art that determine whether or not art will impact their life. In this way, you could conclude that cleaning your bedroom to make it "look nice" which makes you feel good about it looking "nice", then you just had an experience with "art" or at least an aesthetic experience...
The evidence for the argument is in the video: According to this definition of art, a thing being art relies upon an audience evaluating the thing in artistic terms. If people in an audience ask "is this art?" then they are evaluating the thing in artistic terms. Therefore, the thing is most likely art (under this definition).
Rugnetta, you are once again spot on. Thanks for thinking through all the difficult topics and presenting them clearly and cogently so we don't have to. As to your final question, my answer is a definitive no. I don't think we'll ever be able to divorce the programmer from the algorithm, no matter how sophisticated the algorithm gets. Also: Thanks for the shout out :).
Regarding yesterday's video, I took all the responses to PBSIC videos from the subreddit in the last year with more than 10 upvotes and used a Markov model (taken from github.com/codebox/markov-text) to create the quintessential Reddit response to this video: It feels like the medium of death is physiologically different gradation of our brains. It isn't really address tricking the discourse. Maybe it's for playing Devil's Advocate. A lot of play revolves around them, and non-hormonal birth is one place for a bunch of money. Your entire third world being honest, it comes to our economy. One is a joke about how he creates a lot, the other figures who touched me. RUclips channels and Game-stories, the pacifist runs into terrorist attacks or universe to help the widow, the end result of model for the pinnacle of death, deeply anchored in your own right context. For the same things, they respond and the route of those things exist anywhere. But I think a culture has no obligation to him. Being enlightened means that, but I haven't always thought it, and that is then an excuse. It's because I'm not an identifier, a lot of Brass, and only the development is hell.
"Your entire third world being honest, it comes to our economy." That sounds shockingly like what a European or United States economist might say to the citizen of an impoverished nation. 2 real 4 me.
Also, if anyone else wants all (or most) Subreddit comment text from the last year: www.dropbox.com/s/6oafv9zla9m0o1b/PBSIC_comments.txt?dl=0 I tried to remove links and post metadata, but some got through. Still, like 99% written text.
My first thought here is that I'm not on board with the definition of art just being what audiences appreciate or "turn into art" though, there must be more to it than that. Otherwise things like sunsets and landscapes would be art if people took the right attitude towards them. Also it would mean that if you made a painting, and then the next second everyone in the world died it wouldn't be art, because nobody would be there to 'turn it into art' by appreciating it. I'd say art has to have an intention behind it from some artist making (or in the case of found art presenting) the work. That makes machine made "art" trickier, because in order for it to really be art the AI would have to have intentions, which opens up a big can of worms. When it comes to AI art, people have to vet the outputs - like with Sunspring someone had to choose camera angles and music and so on and the script was modified by Oscar Sharp; music writing AIs produce loads of outputs and people have to select which ones to present, and also which ones to put back into the machine to be "learned from" for the next round of generations. So I reckon it's not really AIs making art any more than Pollock splatter paintings were made by gravity; it's people making art using AIs.
Yes, I think the presence of intent is a critical threshold. Even an artist in the absurdist tradition is very intentional, the intent of expressing what it feels like to grapple with existential questions in the industrial age, for example. That's what makes it art, the intent to express and evoke feelings, not to repeat patterns. So, until someone convinces me that a machine, Turing or otherwise, has both intent and feelings, the claim "computer made art" belies the artistry. It's humans using machines to explore the space of possible expressions of art, and whether they and others who see the art chose to then make similar art...art as usual.
Neural networks as they are now have one fundamental flaw that limits their artistic ability: they do not have any affective intent. As it stands now, AI can only mindlessly regurgitate what it's fed, and does not and currently cannot create works *for* an audience. Art is communication and vice versa. Art is art because of a message--be it literal, emotional, or something more subtle--that the audience receives. Even "meaningless" art contains meaning in its meaninglessness. The sum total of the reactions a work of art can evoke can be called its affect, and how well a work generates affect determines how it is appreciated. AI can create affect by accident or mimicry, but human artists generate affect directly, through an understanding of their audience and how they would react. AI, lacking a concept of "self," or "other," or "mind," currently cannot understand their audience well enough to consciously elicit a response. In order for AI to truly understand art, it must first understand humans.
Jabberwocky "If you think that AI cannot create anything in a vacuum, that is, with no previous input whatsoever - than you're right, it can't, but neither can humans." So you're saying that human artists *also* contain some kind of neural network that's been trained by extracting features and patterns from existing works of art? *Skeptical eye*
Jabberwocky I'm not saying AI cannot create something new. They can and do create "new" things all the time. What I'm saying is that AI as they are now (but probably won't be in the future) are incapable of true communication--and therefore artistic intent--because they lack the ability to act autonomously with the purpose of eliciting a response. For example, think of Siri, or any other chatbot. Siri can talk, and ask questions, and generate coherent responses to you, but it doesn't have any concept of "you," its audience, besides what you feed it. Siri may sometimes say something surprising, but only because either its creators programmed it to say that, or because of an accidental overlap between what Siri's algorithms generate and what you find "surprising." Siri will and can never attempt to surprise you on purpose; i.e. by analyzing what it knows about to to craft a scenario that you personally would be surprised by, and carrying out that scenario. Similarly, AI today cannot *knowingly* create what we might call art, because that would require a similar set of mental gymnastics, only expanded to include a large population as its audience. This is the biggest problem in AI today: getting them to understand humans as fellow actors in a system, and not just static sets of data. Once AI can do that, then intentionally creating art would become a very simple task. This is definitely not impossible, but it's definitely going to take some hurdle-jumping before we get there.
I think it's difficult to say without dropping the "r" making it 'Ait/ite' / ʌɪt. I prefer something similar to the Scottish word airt, making it 'ert' / ɛət.
The thing about art is that it's not just about what's on the canvas itself, but what and who is behind it. Art pieces become enormously more interesting when you know the stories of the artist's life and their mind. Computers do not have their own stories and mind, at least not yet. It might be that the art they produce can be seen as aesthetically beautiful, something you wouldn't mind hanging up as a decoration, but at the end of the day it won't have the same pull of fascination with the hand that made it.
I had a professor who said art without an audience, isn't art, it's just masturbation. Art is communication, it's the start of a dialogue. You can be saying something profound, or just, isn't this pretty? or just hey, look at this. But you are saying something, anything. Art is fundamentally unselfish, you give it away the second you show it. If you don't show it, you're just talking to yourself, you are self pleasuring and not starting the dialogue, so it's not art.
Yeah. I've always heard art defined as a "mode of expression" as a way to "communicate a feeling". Art is a more rich form of communication: If you can't put the fear and frustration you're feeling into words, maybe you can illustrate it or create music that reflects this frustration. I posit that, choosing the message you want to communicate is key. Yet, Neural Networks are not consciously choosing "what they're getting at". They just mimic the form of poetry, music, painting or whatever it's been trained on (I'm studying towards a PhD in AI, so I know what I'm talking about). So is it pivotal for an artist to choose what they're trying to get at? Or can an artist blindly create a work, and then the audience can interpret a meaning into it? If the audience feels like they've been communicated to, is that enough? In that case, the crux (for an artist) would be making a work that's evocative even if you don't know what it means.
is it really necessary that a ball exist for me to hold it? or is it enough that I think that I am holding a ball? the answer is: it is enough that I think that I am holding a ball. for art it is the same. someone looks at something, then they convince themselves that it is art and it was created by someone. that is how someone can love a celebrity, then meet them and realize that they are nothing like they thought. because they were not in love with the celebrity, they were in love with the idea they had in their head, inspired by what they saw on the screen, or what they have heard. lying relies on this, it is only because we don't know the true nature of something that we can be deceived or lied to.
+acuerdox I think you have a valid point, I'm not sure how much it has to do with what I was getting at. If you'll allow me to extend your metaphor? Does it matter if the ball is real? I'll posit that what matters if that the ball was thrown. The thrower could have thrown a pineapple, but you caught a ball, and that's still valid. The thrower could have thrown it to someone else, but you thought you caught it. I can't discount perception, but it's the same as a conversation, the speaker can't controlled what the listener hears. There are no guarantees in communication or art. You can't guarantee that the ball you caught is real, or that you and throwers understand of a ball is the same, but in my definition of it you need a thrower and a catcher. I also don't think that a catcher catching a ball that isn't real is necessarily lied to, they can just have a different definition of what was thrown than the throwers.
TheCyberwoman quite true. and now that you write back I realize that I didnt make my point quite clear. my point is that while I agree with everything that you say there is an easy pitfall in this concept. and that is the idea that two people are needed for a dialog to happens. my point is rather a warning. do not forget that sometimes the thrower and the catcher are the same person. I know this is quite rare, but it shares some light on the topic of this video, because many are dismissing the possibility of a computer making art since it cant have intent. but if a dialog can happend as long as the catcher thinks there is a thrower, then a computer does not need this human characteristic to make something that someone will think is art, made with intent.
As a game designer/musician, I'm actually really excited about the idea of things like Magenta. Right now in video game design, there's a problem that people haven't quite figured out a workaround for, where music isn't really being used efficiently from a design perspective. Sure, there are great songs in games, but looping the same melody over and over isn't really ideal. Visuals, movement, even other aspects of sound design, they can all be morphed based on what you're doing. But music, either you play the same melody constantly, or you do the 3D Zelda method of having another song awkwardly jump in the middle of the usual song, and sometimes this can be pretty immersion breaking, like in Twilight Princess when [SPOILERS] you carry Midna on your back as she's dying, and encountering enemies breaks the beautiful Philip Glass style minimalist melody. But imagine if, instead of making a single melody that loops over and over, musicians could program themselves an AI that functions like the people who used to play pianos at theaters back in the days of silent films. You'd have your own personal composer, stringing melodies and enemy encounters together flawlessly, adding to the game experience in a way never encountered previously. It'd be really exciting, to me at least. That's why I think entities like Magenta will most likely be used as another tool as video game style interactivity becomes more ubiquitous in culture.
Hey. 6 years in the future compared to this video. It seems we did allow machines to make art and they are pretty good at it, with help of some human prompts...for now. We will see in a couple what will come of it.
See, here's the thing - from what I've seen so far, an AI can "create art" in very much the same way that a paintbrush can "create art". As in - it's a tool. It can't actually create anything, unless a human has used it to create art. Because stuff like Magenta is just that - a tool. An incredibly sophisticated and self-learning tool, yes, but still just a tool. It is not an algorithm that was created to crunch numbers on, say, airline fuel efficiency that, suddenly, decided it wants to draw random squiggly lines instead. It was created, by it's makers, to draw squiggly lines and fed tons of human-made artwork to serve as it's basis. Thus I'd say that AI, in this form, is not actually capable of creating art on it's own, but it's a medium with which humans can create art themselves. Also, most of the art created by neural networks and stuff like that just looks like crap. Sorry, but it does.
AI definitely goes beyond the "utility" of a paintbrush, in that it could be set up to routinely create drawings and new and unexpected pieces, and as AI improves, the line between an artist and an AIrtist will only become more blurred. One day we will have AI agents going around and finding inspiration, living, learning, and creating. At this point neuronets can be used to apply effects and generate imagery but the level of quality can't be very high without a human filter bestowing familiar aesthetic values at least by selection.
@@IsaiahOdhner yes but also no its more a personal matter imo. If u dont see it as art then its inst if you do its is. But i agree i dont see ai created pieces an art. I believe art is a creation of human consciousness. Its just a part of who we are. But as we go further into the future machines are prob gonna need us less ai creates ai that learns how to create art and imitate how to get inspiration and so on. And if machinese are gonna copy human consciousness to make art i think it would be considered a new type of art, and therefor its up to personal preference if u consider it art or not :)
This comment comes up a lot, and I just want to say that it's both true and awesome. Semantics is important; what we call a thing changes what it is and what we do with it. Maybe there's something Zen about the quest to experience the world outside of the language we use to make sense of it. But as long as we're talking about stuff, it's good to have "Semantics: The Channel" here to remind us to be thoughtful about *how* we talk about stuff.
Semantics are important. What something means changes over time. People think that wherefore means "where" but it actually means "why?" or something along those lines.
In a way, semantics don't matter, because reality doesn't change according to the names we give it. On the other hand, if we want to make sure we're talking about the same thing, or we want to be more concise when talking about highly complex subjects, semantics are very useful. I hate arguments and philosophical discussions which just boil down to how different people like to define a word. But when there is common understanding of a concept, things become more useful, or at least interesting. So how we define 'art' is not a very interesting topic in itself. BUT, how we answer the question 'can robots make art' reveals a lot about our emotions towards current and potential future machine intelligence. If humans never come to accept a piece of art as really originating from a robot, then they will probably never accept a robot's claims to self determination or consciousness or whatever.
Sam B "In a way, semantics don't matter, because reality doesn't change according to the names we give it." Right, except that reality frequently changes according to the names we give it; that's kind of what it means for a thing to be "social". There are a lot of facts that are true in virtue of our saying so, and there's no clean line in those cases between what exists and what to say about it. Whether a thing is art, or money, or a national border, or sexy is wrapped up wholly in whether people collectively say so. And that in turn has big, very real implications for how the thing gets treated. "What do we mean by ....?" and "Does .... count as .....?" are the kind of semantic questions that keep lawyers and cops in business. And if you can get shot over a thing, it's a real thing. **Gets off soapbox**
If modern "art" can be art, then Artificial Intelligence art can be art. No wait, screw that. Modern "artists" are self-entitled twats who think their zero-effort meatspace shitposting can be considered art. Programming an AI that can learn to autonomously make human-like art takes a hell of a lot more intelligence, work, research, programming, etc, that the generated works should be appreciated as art by default.
In that case, it sounds like you're attributing the works to the developers, not the machine itself. AI art can be art, but I think the more central question is whether AI's can be artists.
... Just watch Sarah Urist-Green's "I Could do That" video that Mike links to at 4:31 in this very video. And please, stop making yourself look like a textbook case of the Dunning-Kruger Effect. It looks bad and hateful and plain misinformed.
+Elliot Collins Or perhaps one could interpret what he's saying as, the developers are artists, creating something something that is simultaneously a work of art, a tool by which art is created, and an artist, all in one. I feel like this concept is worth expanding on by people more philosophically-inclined than I.
I don't know exactly what you mean by "modern art" since that's a very broad term, but many modern art movements such as dadaism are what you call "zero effort shitposting" specifically because they aim to criticize the idea that you can apply conventional logic to art.
My take? Art is not just the audience but the process that brings the art to the audience. Dada made everyday objects art, not because of the effort to bring those works about, but because of an intentional and meaningful expression of the changing sensibilities of an entire culture. Machine "art" isn't art as it comes out of the machine. Its products are like an old discarded urinal. It only becomes art through the human process of selection and intention. In short, they are found objects. As produced by the machines they are a pile of rocks with maybe a rough gemstone here and there. The earthmovers don't make jewels, the jeweller does. It is the act of sorting, sifting, and initial interpretation by the people who shape those networks and choose their input... ie the machine didn't make art, the computer artists used a machine to make art.
Or taken from the other side. The faculties that make art are too varied and interconnected to be reduced to just one "smooth" network, no matter how many nodes it has or how large it's source material. We will only know that a machine is an artist, when there is a serious case involving that same machine taking its owner to court to sue for recognition as a person so it can in turn claim the copyright to its art.
As a Artist myself, there is a lot of things that goes into making art. Understanding, using, applying, criticising, making, and sometimes, not making it IS Art. However you pointed something @ 7:36 that got my attention, about Artists and Tools. However, did the tool art? One of the first thing that we learn about art is that WE are the artist and we USE a tool. We know perspective, form making, 3d space, perspective, color, storytelling, blah blah blah and we APPLY that WITH A TOOL (with that we also have mastery of the tool it self, how to handle it and to use it in a artful way) Painters uses INK as a medium Writers uses LANGUAGE as a medium Digital Painters uses PIXELS as medium and so on. Programers uses a computer to create an AI that create art. the Artist is the programer because it used the AI as a tool. The AI, for it to become a artist, has to use a tool to create something Ok done, but art is not only the masterful manipulation of a tool. Is also has a objective, a relationship, first with ourselves, and second with our environment. the first "art" that we know of were born out of the necessity to comunicate or to express to someone, and in the process, to ourselves. A consciousness lives in a void, receives a order to analyse something and then creates something. is it art? i don't think so. Does it comunicate (to someone and it self)? Does it express (to someone and it self)? does it have a receiver? does it have a question (to someone and it self)? an ambition (to someone and it self)? no. because there's only 1 Concience, one self. We understand as art of artful things that are in nature because animals, plants, bacteria or even viruses because it has a environment of their own. WE don't have that with machines or AI. and They dont have THAT with themselves. They don't have a language of their own and that we all understand (or try to), and they don't try to understand who we are for them selves and for their own reasons. Art is something that ultimately is born in a social envirionment. Even if it is something personal.
...They... Can generate puns on par with that of joke books aimed at 10 year olds. Sort of. Occasionally. Seriously, though - The research on computational humour is quite fascinating. Applications include helping kids with learning difficulties engage in the linguistic play that's considered vital to the development of 8-12 year old minds. And the initial research paper on the subject had a bunch of 10 year olds (referred to in the paper as 'joke experts') rate 'punning riddles' ("What do you get if you cross a sheep and a kangaroo? A woolly jumper." style stuff). Most of the computer generated stuff was rated lower than most of the human generated jokes. However, the highest rated joke? Computer generated. Don't have a copy of the paper to hand right now, but I think it was "What's the difference between leaves and a car? One you brush and rake, the other you rush and brake."
Tay was one of the funniest things I have ever seen in my life. Not only that though, making images that are randomized can be really weird. For example www.tsu.co/zZCastleZz/121933354 With that said I see HUGE potential in this area to absolutely go insane with it. Really all I need at this point is more tools at my disposal before I can get into it.
My thought of when we will see an artistic work from a machine as being art credited to that machine is whenever the first time is that an AI which wasn't made to create art creates something artistic. Perhaps the robot butler sings a lullaby of its own creation to sooth a child, or something to that effect.
As a child, I once went to the beach and made 'spin art' (that's when a machine spins a piece of cardboard, you drop some paint on it, and it scatters off the edges, creating 'art'.) When I was done, I somehow ended up with something that, to my eye, looked like three eagles flying into the horizon of a brilliant rising sun. I had no intention of doing that. The spinning machine most certainly didn't 'intend' anything. But to this day, that painting is one of my most favorite pieces. If that spin art in a cardboard frame isn't art, then I don't know what is.
for you magic the gathering players out there check out roborosewater on twittet. it`s 3 seperate recurrent neural networks with various levels of intelligence trying to make magic cards. they`re hilariously convoluted.
I think one of the most interesting things about this is that, as opposed to making art, it's making something for which there are specific constraints and nuance: Whether it is a card that does/could function, whether it is a card that is balanced, whether it is a card that "would" be printed considering color philosophy or design philosophy. These are all things that actively get discussed and judged on the merit of the product itself. Saying "This is a Magic card" may be something that's more acceptable or allowable than "This is an art". Or at least, it's treated differently by the audience that it is reaching.
+Planeswalker999 I haven't in YEARS but several of my very close friends are avid players. Actually, my original deck is on the IC set. Maybe I'll show it in the next comment response video? It's... ancient.
The way I see it, art is Art not because people acknowledge it as art, but because someone creates to be art. A machine will therefore successfully create art once it becomes sufficiently complex and intelligent to have the intent to create art. In other words, a machine can create or if it is a person on top of being a machine.
We should appreciate who makes the AI that makes the art, not the AI itself. The works of such developers and mathematicians are as much art as they are engineering and science. The AI itself is art.
The problem I keep running into when trying to answer this question for myself is rooted in the problem of 'what counts as art anyway?' Like, in a way, anything can be art, but at what point are we just calling anything and everything art because anything could possibly have a deeper meaning to any number of people. On the other end though, if we're too strict with what counts as art, at what point are we just arbitrarily saying something isn't art because it doesn't meet this specific set of qualifications. Either way we're either being too pretentious and giving too much credit to "artists" who's only intention is to laugh all the way to the bank, or we're being too snobish and denying people who are really trying to say something the acclaimed title of "artist" because what they made isn't 'good enough'.
We should consider the funny little paradox in the "too snobbish" train of thought. If we decide to be strict in designating what is and isn't art we ironically give too much credit to the artists who can easily churn out the forms of art we deem acceptable. Anyone good at making still lifes (and it's not THAT hard), can laugh all the way to the bank by making art that's acceptable, without putting any heart or soul, or whatever else into it. I don't think it's good to look at art in terms of monetary value, or how it benefits the artist. (whether we think they're exploiting the system or not)
***** I play chess for the social interaction as well as to test my own mental acuity in relation to my opponents. You start to realize that the littlest thing can affect your discernment in game like eating a big meal right before, worry, anger, sadness, recent sexual activity etc. and so if viewed through this lens it's really a battle of the will between two people. But ym point is: does it really matter if machines can write 6 part novels? Almost every idea has been done in some manner already anyways yet we still continue to read, draw, write etc.
As a programmer, I maintain that introspection is the key process that turns a complex neural net process into a thought. It's not enough for a computer to generate art, it must also be able to explain and explore itself self, it's state now and it's state when rendering the art. Computers right now are really not tracking their internal state in any introspective way. This handicaps them from self feeding their iterative information and building meaning and also from explaining those "thoughts". It's no accident I think, that when we challenge AIs in fiction, we often delve deep into how the machine/robot AI thinks about itself as a marker for how human it is.
Art, for me, is about intention. Whether or not machines with AI can ever "intend" to make art and not just go through the motions of "making art" remains to be seen.
I would argue that it's just as much about interpretations. there are so many visual cues that cause us to think one way or another about a visual piece of art, and many are already easy for a computer to reproduce and modify.
I am on the exact opposite end of the spectrum, I don't care whatsoever about intend, for me it is all about the interpretation I can get out of it. That way, photography works for me as an art form, I don't care what the one thought he or she is seeing that that is worth making a picture, I only care about what I find in the picture. So, I am pretty much on the "death of the author" side of approach to art.
I feel that this concept of independence of human interaction is flawed for one simple reason: If we are not independent from influence of one another, how would an intelligence we create? In most cases art is created from reaction. Our reaction to beauty, fear, anger, madness, politics, etc. We take inspiration from our environment and our being and react to it. We are basically putting a clause against a machine of which we do not aspire to. And then after wanting to have a machine produce art by isolating itself from humanity in some way shape in form, we want it to create art that is compelling to us, the beings we want it to be isolated from? It's nonsense to ask that. And if we design an intelligence that can create art, and that is it's purpose, wouldn't it mean it lacks actual will to make art because that is what it is designed to do? If anything the only willful act of such a machine would be to *not create art*. And if it did not, we, the designers, would not recognize this as a willful act, instead considering it a bug, and attempt to fix it. Honestly, the only way art could be considered its own, would be for an *AI not designed to create art* made something beyond it's initial programming. Again, we as humans would most likely disregard what it does since it would be something that was unintended and attempt to rectify it. I believe most of this comes down to one thing: Will we accept an AI's will as it's own. Chances are, we will never accept it unless it were to fight for it's own freedom, and even then we would fight back. We barely even accept the will of other humans, let alone something we eventually create.
3 things make us appreciate art, combined or alone : 1/ technique : not everybody can master the delicate curves of faces with a brush or a mouse or a pen. Few of us can, so by rarity we appreciate it. We appreciate the struggle. Machines can do that without effort so it is shallow to appreciate. 2/ Details. Patience is a hard human quality (virtu), details in drawing, background stories in writing, rendering an entire fictional environment in the mind just in order to build a universe capable of holding the creation in a meta-whole that gives it extra meaning. Machines do not dream, they don't strugge in face of money, pain and death, how can the art have meaning to humans then ? It is just liberating us from our condition, what's the machine condition ? Limitations of ourself, and the artistic overtaking of it is hat gives human art meaning and "sharability". 3/ Breakthroughs : if anything is a remix and I totally hold true the concept, art, like science, experiences breakthroughs. Scientist build the knowledge on other scientist but they still discover something new, or give new meaning to something discovered before. There is nothing utterly special about Memes. However memes have not always existed, there is a before-meme period. There are plenty of examples like that. Memes are usually visual, there is nothing new in the techniques used, but there is in conveying new meaning. This is due to the new ways we live our lives, when something new artistically can in a glimpse share a common thing betwen ourselves. Can a machine not "living" , and even less the same way we do find new ways of conveying art between ourselves ?
I think it's because many artists see themselves (and/or human creativity in general) as something "special" magical even. The fact that a mechanical device which is infinity recreateable is capable of engaging in this process forces many of these kinds of artists to reevaluate a core principle of their worldview. Add to that a profound ignorance among many artists (and people in general) as to what AI is and what it is capable of and fear quickly takes hold.
+ChunkNinja after all, mass production of furniture, clothes, jewellery etc. didn't remove public interest in acquiring handmade products. Artists/craftsmen found their niche just fine whether they make bespoke items or assist production in factories (designing products, doing something around machinery or whatever).
It is not funny, try make art yourself that you enjoy to create and get the feeling that how much time you even try there will be a super god that can outperform any art in a second and then you feel all motivation just gone.
The first Furniture is made by a human, its entire design from start to finish, the reproduction is where machines come in as machines can recreate so more or less everything we already created, but they cant create what do not exist. MP3 songs are duplicated millions of times but that first MP3 song is the most important one because it decides how the rest of the clones will sound so without any input AI cant do shit.
I feel like a more important and rather depressing question that should be asked is "Artificial intelligence replace artists?" I feel like the answer is a mixture of simple economics and ethics. A machine could work for pennies on the dollar for what a human would ask for. Essential out preforming humans and making designers and other artists obsolete. A truly terrifying idea. A word full of artists and art yet, a world free of creative individuals.
What I get from this and my art education is that Art, as it is known currently, is not divorced from Narrative. Not only the narrative of the artist's role in a work's creation, the artist's background, the artist's position in their movements, the artist's interests, the dialogue between the artist and the audience, but also the metanarrative that art is made by an artist to be consumed by an audience. This then implies that art cannot be art without context. That if la Gioconda were not painted by the hand of da Vinci, but instead an accidental series of grease stains on a board, she would not be as beautiful nor as mysterious. But, it is known that the aesthetic appreciation of non-artificial things can occur (e.g. the appreciation of the natural beauty of a landscape or a coral reef). Should we not be able to appreciate that made by machine with the same level of wonder and emotion? Additionally, if we can finally accept art as being made sans auteur, we should be able to turn our critiques and readings to the aesthetics that these machine made works emulate, a sort of mirror onto our society measuring our collective triumphs and failings.
Art does not necessarily need to have emotion or meaning. Emotion and Meaning have nothing to do with art really. Sure theatre plays might benefit from meaning and a little emotion. Same with music and books. But all those are just sub-genres we like to put into our big drawer labelled “art”. And there are no restrictions on what to put in there. Everything goes. Whatever we like. WE just happen to like stuff that makes us sad, happy or gives us direction. What art is, is literally just defined by us. Things we love. It’s OUR drawer. It’s very personal. Why would we put something in there that was created and thought up by something else? Something that is NOT us?
Two Points: A machine making art raises so many questions in the world of copyright. Some examples could be " Who owns the copyright?" "Does the machine own it, or does the the owner of the machine?" Maybe the AI's creator gets the credit for the work, maybe it is automatically put in the public domain avoid the argument all together but then that deincentivises people from creating the AI to make the works in the first place. Artist are appreciated ,in my opinion, because we know they will not always be around. Some of the most prolific artists, like Van Gogh, for example where not truly appreciated until they've gone from the earth. Software can be immortal, it can be copied from one physical media to another, to prevent its demise and can do it infinitely. But then again that sounds horribly lonely in a way that something that could make a great work of art.
Methinks this gets into either authorial or viewer intent. Does, say, my excitement when I type: + make that symbol exciting to you? Meaningful? Would it be + meaningful if I didn't preface it with my feelings? If it's not meaningful: then all you have to do is have some feeling when you hear AI generated content for it to become art, and you'd be suddenly *for* AI art. If it is, then I'd like to ask what emotion you got from the second '+'.
+Artimis Fowl maybe, you're right. Feelings must come from our side as a viewer. I'm no artist, but I appreciate seeing a story being told thru an art piece. The + wasn't really art. It was just a sign that can mean add or positive or a cross of some sort. To me when I look at art, it expresses the way the artist felt when he made it or what they were going thru. If it was by AI, what would it resemble?
Fun question IMO =) I like to interpret art similarly, I find it useful. But whenever I do this, I know it is pretend... I don't know what it's like to be another person, I've never tried it. Can't figure out how. Art, for me, is where I pretend I can understand, and try to feel what the artist was trying to explain in their art. Maybe we can do the same for AI art? I have a really hard time imagining what it'd be like to be made of silicon instead of cells. Maybe this is our chance?
I freaking love SO MUCH this subject! But going to the point: I don't get people that argue "Oh, just because there was not an sentient intention behind the creation of this... I don't know, movie or image or text, or whatever... this's not art". I mean, would Forrest Gump be a less emotional movie if was written by an AI? No, it wouldn't. I believe art has more to do with how other people perceive that thing, than with the "intention stuff". I mean, I think one good example about how the "intention of the author" is not that important is that... those old Greek and Roman statues, that we see in museums and photos and so on... Many of them were original colored . And, as the time passed, the ink started to peeling, and the statue ended up losings all its colors. And we remember them in a different way than "the author's intention". Of course that's not quite the same, but I think it shows how what counts is the perception of people. About "We should create machines to create art?" And I think yes. I mean.. I think that artificial intelligence would allow us to not bother to """technical sides of art"". I mean, imagine that you.. I don't know, you write comic-books stories, but you don't know how to draw. If we have a machine capable of create drawings you could just say to a machine "I want you to draw this for me". I mean, many times in order to produce a given project, you depend on things that other people will do. So why not replace other people arts by an artificial intelligence? So, imagine that you want to create a movie, you just wrote the script and so on... my question is: why don't put the scripts in a machine that will create a movie based on them? If anyone is interested, a couple mounts ago I wrote a blog post talking about the social impacts that art technology automation would have on society. I also talked about the technical challenges: estranhosidade.wordpress.com/2016/02/20/the-automation-of-the-technical-part-of-art-the-use-of-artificial-intelligence-in-the-artistic-creation/
We already have algorithmically-created art. Movies and music created by marketing strategists and focus group analysts are basically works created by monitoring the success rates of recurring patterns and then reproducing those patterns and formulas. The only thing missing is the A.I. Would I call these works "art?" No. Do they sell? Yes.
Of course Sunspring was a beautiful piece of art! Just because it was written by an AI doesn't mean it's less artistic... it's just a different form of art and I feel people will soon start appreciating that too...
How do you know AI has no emotion? Seems unfair. How do I know you have emotion? I've never been convinced Bach had feelings - way too mechanical. Does that mean all of Bach isn't art just because I say he had no feelings? Similarly, I think Deep dream has a lot of excitement in it's interpretations of our photos. Does that make all of deep dreams outputs art for me?
doesn't have to. if it evokes somekind of emotional response it can be art. art is not purely created by it's creator. it's created everytime someone experiences it.
Your final question of whether we'll ever regard the work of a machine as art seems like it kind of reduces to your earlier question of "When Will We Worry About the Well-Being of Robots?" On some level, both are beating around the question of whether Data has a soul. I *love* this question for so many reasons, not least because I think creating a machine of moral concern is our one barrier to exploring the galaxy. Our children will visit other stars the moment we decide that these things made of metal and crystals are our children. And their ability to create art seems like a necessary first step.
Perhaps we should think of these works not as "art created by an artificial intelligence," but rather, "art created by computer scientists using artificial intelligence." The AI is a tool the computer scientists created to make art, and the computer scientists are ultimately responsible for whatever that AI creates - therefore, perhaps it is the computer scientists who are the artist.
my first reaction was to go for the MacLuhanite argument: machines as art. What I mean by this is that the medium of film and the medium of television have societal effects so transformative and wide that they can be considered not necessarily cultural expressions, but in that sort of realm. The societal changes created by these inventions produce a world in which works of art such as Community, The Shadow, or F For Fake can exist, and as such in part create them. basically we're already a neural network shaped by machines, trying to replicate ourselves.
Here is an Idea: an AI with the purpose to apreciate things and declare them as art. Like an automated viewbot that filters through the Internet on the search for unseen videos. Watching it a thousand times, sharing it on social media, and leaving comments about the video fueled by a algorithim that trys to determine what happens in the video and what could be interesting about that. Like an automated audience viewing and discussing pieces and thus creating art out of works.
There should be a video about what counts as "literature" and what counts as simply a book. This popular discussion should have Idea Channel's take on it.
At this point, I describe all creative text as "literature", as there's really no other good general word to use when describing both online fanfiction and classical text :p
"But is it really art? I say, if you have to ask if it's art, it probably is" I may turn this into an embroidery, cross stitch, and all the ~lower~ forms of art out there.
After a college class discussing a similar topic, I asked some professors in the art department what they thought art was. The answer I liked the most was that art was the creation of something, not the final product. I don't know if the professor would be it so simply, but he made it sound like art was a verb and not a noun. And artists can understand their process and knows that they are creating art. I don't think our current machines understand the process enough for what they are doing to be dubbed art (I would not call the Grand Canyon art; beautiful, yes, but erosion and the natural world did go about creating; it just happened). Machines can get there, but not now. I do find it interesting that you define art as needing an audience. That's another interesting way to define art. And really, this conversation about machines will have to boil down to a more clear definition of art.
I think it's not only already happened but happened some time ago - video game level design has had algorithmic generation of environments since before Diablo, methods for people to record and spread the experience of those algorithmicly generated landscapes through level codes, an appreciative audience and some even have an in built rating system to learn what the audience wants. These environments are built on rules but frequently not pre-existing environments, have intent of outcome and even thematic integrity.
There is one, small gripe I have with your script: "audiences turn that work into art through appreciation." It is of my experience that any reaction, as long as it's not the 'I don't get it' reaction, makes a piece art. I have found that disliking some art is just as important as liking art. As long as there is an emotion attached, or more an importantly an experience, that's when art becomes art, even if it's a negative experience. Loved everything else, just wanted to throw that in :)
This has happened before, with males trying to understand female artists, or people trying to understand the art of a very different culture. Some will accept it as art independent of whether the originator is truly independent (who is?), what is its intention, and whether there is implicit meaning in it. Others won't.
I find it interesting that in attempting to teach machines art, these programmers are doing what teachers have had students of art do for hundreds of years. This being called the study of master works and producing master copies. The issue that I see, as someone who has studied art and graduated with an art degree, my focus being animation, is that these machines don't have the capacity to step back and look at their works yet. Part of being an artist is that human capability to pull away from a work, either finished or in the midst of construction, and ponder at, not just the future steps required to either finish the works or improve upon the next ones, but to ponder it's meaning and how it may be interpreted by an audience. One has to internalize these different view points, the strategies to get to the desired goal, to have a desired goal, or in the case of creating a master copy, coming to an understanding of what the original artist had done to complete their work, the experience and time it had taken, and what they were trying to drive home. This factor, for now, is human but it may eventually be something that a true AI, rather than just an algorithm, picks up on. At this point in time, these algorithms are more akin to parrots, though they may improve on appearing human through trial and error, at the end of the day it's an echo. A mimicry of humans or particular humans that they are taught and not truly something they had envisioned for themselves. They cannot refuse what they are given or choose which parts they prefer, they can not internalize what they study and apply it from their own view points or attempt to apply it from another. They do as they are told, without question or hesitance. This is in part why people don't appreciate art created by these algorithms. The computer program doesn't care about what the audience thinks or the end result of its work and it's works are not it's own creation or design, it's just a copycat.
I think one big problem is that a lot of art is evaluated for "connection", connection is what give meaning, not the intellectual construct of symbols, ie meaning is experience, precisely human experience, something we can relate to, hence connection, art is a mediation of experience. When we look at art, we don't just look at something that affect us, but also how someone was affected and produced this piece base on his affection, it's through the artist we reflect on our self and experience. That's why a sunset is not art, but a photography of a sunset is.
i was thinking something we think of "originality" as "difference from existing material" i mean, it's one thing to make a medley/remix of classical pieces, which is sort of what RNN's tend to do, but it's another to be inspired by classical pieces and make something new with that it might be possible to train a neural network to identify what it means for music to "sound nice", another to recognise similarity to other music, and then train an RNN using the outputs of those neural networks to train the RNN (IE maximise the "nice-sounding" score and minimize the similarity score to the training set)
When we study history we look at art because it has always played a role as a reflection of whatever society is has come from. This plays right into the idea that it is the audience that really turns a work into art (people generally like something they can relate to). If we are going to place a standard on how "autonomous" an AI has to be to create art, we should also recognize that several of the things we consider as art were not created with the autonomy we think the artist had. Several of the works we consider art are portraits and sculptures, which were created by standards set by their peers and society. I think it less about how an AI can be influenced by humans as much as the extent to how an AI can make something that will be relatable and ,in turn, influential to humans.
To me, art has allways to do with communicating something. Joy, fear, awe or a train of throught. A good question in this context might be, what independent machine-artists will want to communicate.
The show was interesting. So thanks for that. Here are my thoughts on it. I am an artist and illustrator. Part of the process of 'art' with humans, is the struggle - the rejection of a normal life for the often more solitary pursuit of 'art'. The choosing of will one eat well this month or purchase costly art supplies. Will one hang out with friends or spend that time in the studio, working. Another part is the long struggle to learn the craft and gain skill over many years of trying and failing at various attempts, and the associated human feelings that coincide with that struggle. While an AI machine can 'learn' it is not the same struggle, nor the same time investment. In fact, it is not the same at all. There are no real world constraints on the AI machine. It has no other job or demand on its time other than to do what it does. It does not have to attempt a full existence - just attampt to be an art bot. It does not have to work to get its food [electricity] or a place to live or any of the things a human artist must struggle to do while also attempting to make art. Oh yeah, it also does not have to try to sell its art. And it has no emotions, so it does not have to deal with rejection or frustration for not selling its art and then not having enough money to pay rent. Another part often, but not always, is the frequent mental aberrations that seem to coexist within the mind of artists. Many, though not all, experience some form or another of some level of mental disorder - depression, bipolar, substance abuse, etc. [the list is long.] For a machine... most of these are stripped away, or if need be, artificially dropped in. There is no real life 'cost' to the AI machine. It does not suffer for its art. It does not make sacrifices and lose out in life for making the choice to follow art. It does not live. As such, most of what it can produce is generally lifeless, and often devoid of any real meaning. Or at worst, it is a form of twisted plagiarist - plagiarizing the entire output of the human race. I know, you say how does this differ from the human artist? But many times the human artist tries very hard to attempt to create something original and outside the the long list of human creations. Art has lost much of its real meaning, though this process has been going on for a while. If you remove the 'artist' and simply leave it all up to an audience to say what 'art' is, then AI machine art may one day be a thing. But it will likely be just a pale shade of what art is, was or could/should be.
See here is why it can never, art is made in the studio, u have a vision but the accidents you make become the piece on your way to make vision a reality, you see Ai wants to be perfect, the basic perfectionist nature of Ai unapproves its own art
Just as a quick book rec, there's an excellent discussion on this topic in Ian M. Banks' novel 'Look to Windward' between a composer and an extremely advanced AI. Very recommend.
Two questions: 1. If a work becomes art through a conversation with its audience, are old, canonical works-which have been interpreted so thoroughly that there is little left to say-still art? 2. Given the same assumption, what does it mean that AI-generated works currently only enjoy an audience composed of humans? These AI do not "talk" to each other, and they do not/cannot really even hold two-way conversations with their audience. What does this mean for the artistic output of AI? Doesn't it make them effectively "artists in isolation?" Regarding the question about AI-produced works existing for their own purposes: Humans create art as part of their struggle to understand themselves and each other. Perhaps AI art existing on its own terms would require it to hold meaning for the AI that produced it. Some sort of meaningful change or process that has at least qualitatively changed the AI.
Emily Howell's work is not authored by her father, she is his daughter, creation, and servant, but her work is her own, learning from the masters that came before, just as her human composer peers.
We may not be at the point yet where I can have a rational conversation with a robot about our interests, but we're already seeing the robot prejudice Hollywood showed us many years ago. And it's that kind of treatment that will cause the robot uprising.
I think one thing that the AI written episode and Sunsprings [sic] conclusively proves, is that there's not much weight to that acting argument "their performance wasn't very good, but look at the script they had to work off." The actors in Sunspring were able to deliver affecting, emotionally nuanced performances, and even got a few laugh lines out of what basically amounted to utter nonsense. And you, Mike, were able to deliver what, to a person half-listening from another room, could pass as a typical episode of Idea Channel.
Sunspring is an interesting example, because the directors and the actors had to interpret the screenplay in very specific ways. After watching Sunspring I checked out the screenplay itself and I felt as if I didn't agree with the interpretation of the film as it came out and as experienced by those specific humans. A part of me felt as if Benjamin (the AI who wrote it) had been somehow misunderstood (obviously not in a perfectly literal sense, but there was definitely an element of that feeling)... This raises two main things for me. Firstly it makes me think of the ways in which we anthropomorphise AI in different ways, ascribing things like intentions, traits, and gender on to them. By humanising them in these ways it's easier to give their modes of production more value. Though there does of course seem to be an element of ego here - either through validating how humans can create something human-like, or in finding validation through having something not human so closely resembling the "human condition", in ways that we feel less alone. In instances where we can't reach a point of anthopomorphising an AI sufficiently to value it's ability to be creative, it might be better to track how people perceive creative works when they don't know an AI has produced it. This is similar to how people react and experience things completely differently depending on whether they think they are experiencing the "real" thing (for e.g. replicas of artefacts etc.) The other thing it raises is with a creative medium like script writing which is inherently reinterpreted and adapted, it could also be easier to feel as if their was some sort of "machine intention", beyond the interpretation of the people who go on to make the production (this is less the case when an AI is creating a final creative product more directly, like a piece of music or fine art).
It is hard to say that the machine is making "art" if the machine has not an intention with it. Like, on one hand we could say the machine "lives" to make that piece, but on the other we cannot say that they are using their works to communicate an abstract idea, which it is agreed is what forms "art". The moment those algorithms become so powerful the machine is able to pick a theme or concept, not randomly but by being able to decide what theme or concept is more relevant or important to talk about, and it builds a piece trying to communicate that concept or theme, then I believe we can say the machine is making art.
Having been an artist for over 40 years, I arrived over time at a definition of "art" as "mind and intent." It was interesting to see that you came to a similar place. There are a couple of issues that I can point out. One is that appreciation from outside is not necessary. However, appreciation and critical thinking on the part of the artist is. Cutting the AI loose is a deep problem. Ideally, the artist should choose the media, the content, the process and the dissemination of the work. Probably other stuff too. Maybe the AI doesn't want to use paint or pencils, or make something meaningful to humans. In fact it should be primarily making something meaningful to itself, with meaning for others a secondary concern. Those others might very well include or be limited to machines. In many ways, the definition of art as far as the machine is concerned is none of our business. We might be critics of the output if we wish, but the machine must decide "what is art" for itself. "It's art because I say it is." A word about original thought. This was an issue addressed by the Surrealists, especially Andre Derain. Automatic writing and drawing where as much (or more) a part of Surrealism as melted watches. Personally I try to generate original thoughts by doodling randomly while occupying my senses with something else, eg. dining in a restaurant with a group of friends actively engaged in conversation. Others use random generators such as opening a book and pointing to a word without looking or throwing darts at symbols pinned to a wall. So, maybe originality is not really the hardest part of the AI design. What follows the generating of an original idea is where the difficulty lies. That's where it comes back to critical reasoning and decision making.
This question really got me thinking. I've been playing a lot of table-top roleplaying games lately as a Gamemaster and have often found myself talking about or pondering on what it means to create and how one can create most effectively. I think it's not a surprise to anyone that art is often a reflection of ones experiences, both in life and of other art in any medium. Experiencing life and art gives an artist a larger breadth of knowledge and technique to draw upon when making their own creations, so creating art in a vacuum would be impossible. In terms of an AI, giving them works of art so it has something to draw on when trying to create an original work is the same thing we do as artists ourselves, but with one key difference: we seek out our own inspirations. I think the moment we let our AI go find their own inspirations to create art rather than feeding them what we want to is the moment in which AI will be creating true art. In other words, when we stop treating our AI like children and more like adults, they'll start to create works that we can respect as art.
Hasn't most art a significant biographical aspect? As a writer, I am aware of "remixing" tropes and ideas from art I have consumed, but there is always some of "me" in there. My own likes, interests, and biases have a great influence on which elements I chose to "remix". My own experiences _outside of art_ influence what I want to write about, and how I want to write about it. My childhood, my relationships, my non-art-related work experiences, travels, friendships, hurts, fears, all of that has not just "some" influence but is probably the main motor behind everything I write. Locations, situations, characters, dialogue all come not just from other art I have consumed but from my own life. Artists from all kind of backgrounds and working in all kinds of mediums - Picasso (war, women), Van Gogh (madness, family), Kafka (family, work), Michelangelo (religion, homosexuality), Hemingway (war, gender roles, depression), Bach (religion, mathematics), Beatles (sexuality, politics), etc, etc - have been influenced not just by their reflection on other art, but their reflection on life. So, maybe until an AI has a life outside of its function as creator, its art will remain "imitative". It would need to not just have functions and collect experiences other than "art consumerism", but it would also need a level of reflection about those experiences to transform personal experience into something both generalised and communicable, to create something we can appreciate as "true" art? PS: I am aware that my examples on the experiences going into the works of artists is _very_ shoddy. No intent to malign any of these artists, it is just meant as a quick example to illustrate my point.
I find it interesting the points you raised, but perhaps even more than that, I appreciate the fact that you didn’t delve into debates such as “will machines have consciousness or not” or “do machines need to be conscious to do that”, etc…, which would only, in my opinion, lead to more and more questions and would obfuscate the real topic being discussed here. Of course, maybe being conscious is directly linked to making art, but I feel we would be drifting away from the original question if we take everything that is needed to make art into account (e.g. emotions). I myself am a computer scientist and I mainly do research on affective computing, which is computing related to synthesis, recognition, modeling and expression of emotions through models that are often found in psychology and neuroscience. I would argue that as we are biased, the creation of an independent creative process in a machine will also be biased, there’s no such thing as a truly independent creative process or 100% novelty on creation. Moreover, I am interested in seeing if we will be able to overcome the stigma that is the concept of uniqueness (be it consciousness, emotions, etc...) that humans feel entitled to and because of that think that machines will never have it or comprehend it, given that they are machines.
To answer the question of whether a computer can create art, I will reference the movie "Colossus - The Forbin Project." For a machine to truly be creative, it must first have _autonomy_. In other words, it must generate for it's own needs. In Colossus, the computer in question had ultimate power over all of humanity, and in turn it had complete control over its own fate. It was, from that standpoint, an autonomous being with a will. What made Colossus creative was the inter-system language it shared with Guardian. In effect, it created a keyboard that only a computer could use. _That_ is where creativity lies - the combination of complete autonomy and goal-setting.
I think that Google DeepMind's AlphaGo is an interesting component to this conversation. In case someone has missed it, the Google DeepMind team have programmed a neural network to play Go, and this program has beaten one of the top Go players. During the matches it played, AlphaGo certainly made some traditional plays... but it also made a lot of innovative, creative moves, some of which have already been adopted by certain parts of the community. Note that these moves have been described as 'creative' by many people, and that this does not tend to raise any eyebrows. Now Go is in general quite a creative game. One needs to look out for nice shape, find 'good-looking' moves, and often it is hard to say what makes one move better than another, just that it might 'feel' better. However, this case is different from art, in that it is easier to imagine a computer actually 'knowing what it is like to play Go' than a computer 'knowing what it is like to make art'. This is because one can easily TELL a computer what Go is about: claiming territory and capturing stones. There is a clear end goal, so we can both judge the moves of the computer in terms of this end goal, as well as trust that the computer made these moves with that end goal in mind. In the case of art, all of this does not apply; we cannot say whether the computer made the art intentionally, because we cannot say what it means to 'intend art', even for humans! Or in any case, we lack a clear metric for judging the 'artfulness' of computer output, and conclude from that that the computer itself must lack such a metric, and hence 'it cannot know what it is doing'. Unlike AlphaGo, which can be construed as having a REASON for making the moves it makes, it is hard to say WHY AI's creating art make the art the do, instead of the other art they could have made. Of course, we need to know a lot more about AI before we can even begin to tackle such a question.
Personally, I have always said that art is something that each person has to define for themselves, and I define art as an expression not of emotion or creativity, but of mastery. Mastery over a medium or a skill or whatever. So when looking at art generated by a computer, the appreciation in two fold. One, is the appreciation for the display of mastery over programming and digital engineering that the creators behind the computer possess, and two, the display of mastery that the computer itself displays over the medium in which the art is created. Just because a computer doesn't have to learn a thing from scratch the same way a human does doesn't mean it can't master an art form.
Considering that "Art Imitates Life Imitates Art Imitates..." is in play here, even if a machine could tune out all instances of direct human intervention, it could not completely remove itself from the intervention of humanity. The first artists (and many thereafter) were inspired by the constructs of nature (landmarks, plants, animals). Even engineers take their inspiration from these things, which is why we see the parallels when we look at functional objects. With a computer, the opposite shall be true. The natural domain of computers is the modern human society. Even if an AI were to have a vessel with which to go out in nature, it would "see" the landmarks and relate them to the human structures derived from them. I can see two ways of testing this. The first would be to allow the AI to pick the first "word" at random for its piece, and see how many times it gravitates toward natural scenes versus modern societal ones. The second would be to somehow make a "perfect" AI and throw it into, say, Minecraft and see if what it builds is closer to historical eras or modern ones.
I feel like we won't be able to say that AI is creative until we make an AI that is not intended to create stuff suddenly say "I want to make art" and then does so on its own.
that's actually a really insightful comment
honestly i feel you nailed it on the head. Art is about intention and so long as the ai is told by humans to make art, I don't really know if it qualifies as art. But if instead it goes and makes art on its own than that is art
great point. I was thinking the other way around- I would consider the AI an artist if the AI programmed to make art could say, "not today... I'm not feeling it". The AI would have to be conscious so far as to make decisions that don't have to do with its programmer's intent. But that wont exist until we have a working theory of consciousness and an AI built on that.
Interesting point. Those creepy deep-dream eyes and stuff that are everywhere now, they were not intended to be art. It was a neural network, trained to recognize objects. Programmers just turned that network inside out exposed the inner processes. Can this be considered to be art? It's generated by algorithm that wasn't intended for art but it turned out that it has some art inside its brain.
I suppose, but only if the A.I. was capable of doing such things before. In the sense that it can't go beyond it's programming because in my opinion that'd be like going beyond your genetics and suddenly sprouting wings.
Dalle, Midjourney, Disco/Stable Diffusion:Hey ya
Times have gone so fast, I miss PBS and Now AI has gone so far in just 6 years...
"If you have to ask if it's art, it probably is." If you have to dismiss the question to get to your preferred answer, your preferred answer is probably difficult to actually support.
Not *difficult*, really, just... covered elsewhere: ruclips.net/video/Q_rQbXlmgHI/видео.html :D
Any definition of art is difficult to support
Best youtube comment I've seen in a long bit.
This is the beauty of subjective experience and "Art as Experience" as put forth by John Dewey. Or with Idea Channel's Aesthetic vid they linked. Danto's definition of art and artist are declarative statements that essentially need no further supporting documents or evidence to make that claim. Whether someone is a GOOD artist or makes GOOD art is a very different conversation, one that takes whole networks of people's opinions, culminations of lived experiences, and a community of criticism to build the evidence that it is good art. With Dewey, it is very much about the viewer's conscious framework of experiencing an artwork (like, are they self aware that they are experiencing an artwork?) and their expectations for art that determine whether or not art will impact their life. In this way, you could conclude that cleaning your bedroom to make it "look nice" which makes you feel good about it looking "nice", then you just had an experience with "art" or at least an aesthetic experience...
The evidence for the argument is in the video:
According to this definition of art, a thing being art relies upon an audience evaluating the thing in artistic terms.
If people in an audience ask "is this art?" then they are evaluating the thing in artistic terms.
Therefore, the thing is most likely art (under this definition).
Rugnetta, you are once again spot on. Thanks for thinking through all the difficult topics and presenting them clearly and cogently so we don't have to. As to your final question, my answer is a definitive no. I don't think we'll ever be able to divorce the programmer from the algorithm, no matter how sophisticated the algorithm gets. Also: Thanks for the shout out :).
"It's that butt we have to work our way towards." - Mike Rugnetta
Regarding yesterday's video, I took all the responses to PBSIC videos from the subreddit in the last year with more than 10 upvotes and used a Markov model (taken from github.com/codebox/markov-text) to create the quintessential Reddit response to this video:
It feels like the medium of death is physiologically different gradation of our brains. It isn't really address tricking the discourse. Maybe it's for playing Devil's Advocate. A lot of play revolves around them, and non-hormonal birth is one place for a bunch of money. Your entire third world being honest, it comes to our economy. One is a joke about how he creates a lot, the other figures who touched me. RUclips channels and Game-stories, the pacifist runs into terrorist attacks or universe to help the widow, the end result of model for the pinnacle of death, deeply anchored in your own right context. For the same things, they respond and the route of those things exist anywhere. But I think a culture has no obligation to him. Being enlightened means that, but I haven't always thought it, and that is then an excuse. It's because I'm not an identifier, a lot of Brass, and only the development is hell.
*wipes tear* It's beautiful.
XD that's awesome ^^
"Your entire third world being honest, it comes to our economy." That sounds shockingly like what a European or United States economist might say to the citizen of an impoverished nation. 2 real 4 me.
*jaw drops*
Also, if anyone else wants all (or most) Subreddit comment text from the last year:
www.dropbox.com/s/6oafv9zla9m0o1b/PBSIC_comments.txt?dl=0
I tried to remove links and post metadata, but some got through. Still, like 99% written text.
My first thought here is that I'm not on board with the definition of art just being what audiences appreciate or "turn into art" though, there must be more to it than that. Otherwise things like sunsets and landscapes would be art if people took the right attitude towards them. Also it would mean that if you made a painting, and then the next second everyone in the world died it wouldn't be art, because nobody would be there to 'turn it into art' by appreciating it. I'd say art has to have an intention behind it from some artist making (or in the case of found art presenting) the work. That makes machine made "art" trickier, because in order for it to really be art the AI would have to have intentions, which opens up a big can of worms.
When it comes to AI art, people have to vet the outputs - like with Sunspring someone had to choose camera angles and music and so on and the script was modified by Oscar Sharp; music writing AIs produce loads of outputs and people have to select which ones to present, and also which ones to put back into the machine to be "learned from" for the next round of generations. So I reckon it's not really AIs making art any more than Pollock splatter paintings were made by gravity; it's people making art using AIs.
Yes, I think the presence of intent is a critical threshold. Even an artist in the absurdist tradition is very intentional, the intent of expressing what it feels like to grapple with existential questions in the industrial age, for example. That's what makes it art, the intent to express and evoke feelings, not to repeat patterns. So, until someone convinces me that a machine, Turing or otherwise, has both intent and feelings, the claim "computer made art" belies the artistry. It's humans using machines to explore the space of possible expressions of art, and whether they and others who see the art chose to then make similar art...art as usual.
Oh wow now we have a video on art from you
7 years later, and AI now can create stuff with plenty of intent, if humans use it right. :D
Neural networks as they are now have one fundamental flaw that limits their artistic ability: they do not have any affective intent. As it stands now, AI can only mindlessly regurgitate what it's fed, and does not and currently cannot create works *for* an audience.
Art is communication and vice versa. Art is art because of a message--be it literal, emotional, or something more subtle--that the audience receives. Even "meaningless" art contains meaning in its meaninglessness. The sum total of the reactions a work of art can evoke can be called its affect, and how well a work generates affect determines how it is appreciated. AI can create affect by accident or mimicry, but human artists generate affect directly, through an understanding of their audience and how they would react. AI, lacking a concept of "self," or "other," or "mind," currently cannot understand their audience well enough to consciously elicit a response. In order for AI to truly understand art, it must first understand humans.
"Art is communication and vice versa". I like that.
Jabberwocky "If you think that AI cannot create anything in a vacuum, that is, with no previous input whatsoever - than you're right, it can't, but neither can humans."
So you're saying that human artists *also* contain some kind of neural network that's been trained by extracting features and patterns from existing works of art? *Skeptical eye*
Jabberwocky I'm not saying AI cannot create something new. They can and do create "new" things all the time. What I'm saying is that AI as they are now (but probably won't be in the future) are incapable of true communication--and therefore artistic intent--because they lack the ability to act autonomously with the purpose of eliciting a response.
For example, think of Siri, or any other chatbot. Siri can talk, and ask questions, and generate coherent responses to you, but it doesn't have any concept of "you," its audience, besides what you feed it. Siri may sometimes say something surprising, but only because either its creators programmed it to say that, or because of an accidental overlap between what Siri's algorithms generate and what you find "surprising." Siri will and can never attempt to surprise you on purpose; i.e. by analyzing what it knows about to to craft a scenario that you personally would be surprised by, and carrying out that scenario.
Similarly, AI today cannot *knowingly* create what we might call art, because that would require a similar set of mental gymnastics, only expanded to include a large population as its audience. This is the biggest problem in AI today: getting them to understand humans as fellow actors in a system, and not just static sets of data. Once AI can do that, then intentionally creating art would become a very simple task. This is definitely not impossible, but it's definitely going to take some hurdle-jumping before we get there.
"As it stands now, AI can only mindlessly regurgitate what it's fed"
So...like people on the internet then :p
+z beeblebrox sick burn
Lets coin the term Airt (Artificial intelligence + Art).
Is it pronounced like ʒaɪf?
I think it's difficult to say without dropping the "r" making it 'Ait/ite' / ʌɪt.
I prefer something similar to the Scottish word airt, making it 'ert' / ɛət.
What about "air t"? Like air, but with a t at the end
It's similar to 'ert' / ɛət which I prefer.
not very catchy i think :/
if you create an AI with the express intention of having create it would be an Art-ificial Intelligence.
I'm so funny :P
We had the same idea. Art-ificial intelligence is funny :P
The thing about art is that it's not just about what's on the canvas itself, but what and who is behind it. Art pieces become enormously more interesting when you know the stories of the artist's life and their mind. Computers do not have their own stories and mind, at least not yet. It might be that the art they produce can be seen as aesthetically beautiful, something you wouldn't mind hanging up as a decoration, but at the end of the day it won't have the same pull of fascination with the hand that made it.
I had a professor who said art without an audience, isn't art, it's just masturbation. Art is communication, it's the start of a dialogue. You can be saying something profound, or just, isn't this pretty? or just hey, look at this. But you are saying something, anything. Art is fundamentally unselfish, you give it away the second you show it. If you don't show it, you're just talking to yourself, you are self pleasuring and not starting the dialogue, so it's not art.
Yeah. I've always heard art defined as a "mode of expression" as a way to "communicate a feeling". Art is a more rich form of communication: If you can't put the fear and frustration you're feeling into words, maybe you can illustrate it or create music that reflects this frustration. I posit that, choosing the message you want to communicate is key.
Yet, Neural Networks are not consciously choosing "what they're getting at". They just mimic the form of poetry, music, painting or whatever it's been trained on (I'm studying towards a PhD in AI, so I know what I'm talking about).
So is it pivotal for an artist to choose what they're trying to get at? Or can an artist blindly create a work, and then the audience can interpret a meaning into it? If the audience feels like they've been communicated to, is that enough? In that case, the crux (for an artist) would be making a work that's evocative even if you don't know what it means.
is it really necessary that a ball exist for me to hold it? or is it enough that I think that I am holding a ball?
the answer is: it is enough that I think that I am holding a ball.
for art it is the same. someone looks at something, then they convince themselves that it is art and it was created by someone.
that is how someone can love a celebrity, then meet them and realize that they are nothing like they thought. because they were not in love with the celebrity, they were in love with the idea they had in their head, inspired by what they saw on the screen, or what they have heard.
lying relies on this, it is only because we don't know the true nature of something that we can be deceived or lied to.
+acuerdox I think you have a valid point, I'm not sure how much it has to do with what I was getting at. If you'll allow me to extend your metaphor? Does it matter if the ball is real? I'll posit that what matters if that the ball was thrown. The thrower could have thrown a pineapple, but you caught a ball, and that's still valid. The thrower could have thrown it to someone else, but you thought you caught it. I can't discount perception, but it's the same as a conversation, the speaker can't controlled what the listener hears. There are no guarantees in communication or art. You can't guarantee that the ball you caught is real, or that you and throwers understand of a ball is the same, but in my definition of it you need a thrower and a catcher. I also don't think that a catcher catching a ball that isn't real is necessarily lied to, they can just have a different definition of what was thrown than the throwers.
TheCyberwoman quite true. and now that you write back I realize that I didnt make my point quite clear.
my point is that while I agree with everything that you say there is an easy pitfall in this concept. and that is the idea that two people are needed for a dialog to happens.
my point is rather a warning. do not forget that sometimes the thrower and the catcher are the same person.
I know this is quite rare, but it shares some light on the topic of this video, because many are dismissing the possibility of a computer making art since it cant have intent.
but if a dialog can happend as long as the catcher thinks there is a thrower, then a computer does not need this human characteristic to make something that someone will think is art, made with intent.
So I guess your professor doesn't like Van Gogh then. Much of his works were not shown to the public until after his death.
Hey person scrolling through the comments section! Have a nice day!
Same to you, mate. 😄
Oyoy mate! Hello there!
I'm glad I found this instead of more of the "Last time I was this early..." bull.
NEVER!!!!!!!!!!!!!
Don't you dare tell me what to do!
"In the dooblie-doo" I love this nerdy fun cool corner of RUclips
As a game designer/musician, I'm actually really excited about the idea of things like Magenta. Right now in video game design, there's a problem that people haven't quite figured out a workaround for, where music isn't really being used efficiently from a design perspective. Sure, there are great songs in games, but looping the same melody over and over isn't really ideal. Visuals, movement, even other aspects of sound design, they can all be morphed based on what you're doing. But music, either you play the same melody constantly, or you do the 3D Zelda method of having another song awkwardly jump in the middle of the usual song, and sometimes this can be pretty immersion breaking, like in Twilight Princess when [SPOILERS] you carry Midna on your back as she's dying, and encountering enemies breaks the beautiful Philip Glass style minimalist melody.
But imagine if, instead of making a single melody that loops over and over, musicians could program themselves an AI that functions like the people who used to play pianos at theaters back in the days of silent films. You'd have your own personal composer, stringing melodies and enemy encounters together flawlessly, adding to the game experience in a way never encountered previously. It'd be really exciting, to me at least. That's why I think entities like Magenta will most likely be used as another tool as video game style interactivity becomes more ubiquitous in culture.
Hey. 6 years in the future compared to this video. It seems we did allow machines to make art and they are pretty good at it, with help of some human prompts...for now. We will see in a couple what will come of it.
See, here's the thing - from what I've seen so far, an AI can "create art" in very much the same way that a paintbrush can "create art". As in - it's a tool. It can't actually create anything, unless a human has used it to create art. Because stuff like Magenta is just that - a tool. An incredibly sophisticated and self-learning tool, yes, but still just a tool. It is not an algorithm that was created to crunch numbers on, say, airline fuel efficiency that, suddenly, decided it wants to draw random squiggly lines instead. It was created, by it's makers, to draw squiggly lines and fed tons of human-made artwork to serve as it's basis.
Thus I'd say that AI, in this form, is not actually capable of creating art on it's own, but it's a medium with which humans can create art themselves.
Also, most of the art created by neural networks and stuff like that just looks like crap. Sorry, but it does.
AI definitely goes beyond the "utility" of a paintbrush, in that it could be set up to routinely create drawings and new and unexpected pieces, and as AI improves, the line between an artist and an AIrtist will only become more blurred. One day we will have AI agents going around and finding inspiration, living, learning, and creating. At this point neuronets can be used to apply effects and generate imagery but the level of quality can't be very high without a human filter bestowing familiar aesthetic values at least by selection.
@@IsaiahOdhner yes but also no its more a personal matter imo. If u dont see it as art then its inst if you do its is. But i agree i dont see ai created pieces an art. I believe art is a creation of human consciousness. Its just a part of who we are. But as we go further into the future machines are prob gonna need us less ai creates ai that learns how to create art and imitate how to get inspiration and so on. And if machinese are gonna copy human consciousness to make art i think it would be considered a new type of art, and therefor its up to personal preference if u consider it art or not :)
Semantics: The Channel
This comment comes up a lot, and I just want to say that it's both true and awesome. Semantics is important; what we call a thing changes what it is and what we do with it.
Maybe there's something Zen about the quest to experience the world outside of the language we use to make sense of it. But as long as we're talking about stuff, it's good to have "Semantics: The Channel" here to remind us to be thoughtful about *how* we talk about stuff.
Wow, this must be your first video!
Semantics are important. What something means changes over time. People think that wherefore means "where" but it actually means "why?" or something along those lines.
In a way, semantics don't matter, because reality doesn't change according to the names we give it.
On the other hand, if we want to make sure we're talking about the same thing, or we want to be more concise when talking about highly complex subjects, semantics are very useful. I hate arguments and philosophical discussions which just boil down to how different people like to define a word. But when there is common understanding of a concept, things become more useful, or at least interesting.
So how we define 'art' is not a very interesting topic in itself. BUT, how we answer the question 'can robots make art' reveals a lot about our emotions towards current and potential future machine intelligence. If humans never come to accept a piece of art as really originating from a robot, then they will probably never accept a robot's claims to self determination or consciousness or whatever.
Sam B "In a way, semantics don't matter, because reality doesn't change according to the names we give it."
Right, except that reality frequently changes according to the names we give it; that's kind of what it means for a thing to be "social". There are a lot of facts that are true in virtue of our saying so, and there's no clean line in those cases between what exists and what to say about it. Whether a thing is art, or money, or a national border, or sexy is wrapped up wholly in whether people collectively say so. And that in turn has big, very real implications for how the thing gets treated. "What do we mean by ....?" and "Does .... count as .....?" are the kind of semantic questions that keep lawyers and cops in business. And if you can get shot over a thing, it's a real thing.
**Gets off soapbox**
If modern "art" can be art, then Artificial Intelligence art can be art.
No wait, screw that. Modern "artists" are self-entitled twats who think their zero-effort meatspace shitposting can be considered art. Programming an AI that can learn to autonomously make human-like art takes a hell of a lot more intelligence, work, research, programming, etc, that the generated works should be appreciated as art by default.
In that case, it sounds like you're attributing the works to the developers, not the machine itself. AI art can be art, but I think the more central question is whether AI's can be artists.
_(theätrical eyeroll)_
... Just watch Sarah Urist-Green's "I Could do That" video that Mike links to at 4:31 in this very video.
And please, stop making yourself look like a textbook case of the Dunning-Kruger Effect. It looks bad and hateful and plain misinformed.
+Elliot Collins Or perhaps one could interpret what he's saying as, the developers are artists, creating something something that is simultaneously a work of art, a tool by which art is created, and an artist, all in one. I feel like this concept is worth expanding on by people more philosophically-inclined than I.
I don't know exactly what you mean by "modern art" since that's a very broad term, but many modern art movements such as dadaism are what you call "zero effort shitposting" specifically because they aim to criticize the idea that you can apply conventional logic to art.
"Is it really art?" I ask that question a lot when it comes to actual "art".
My take? Art is not just the audience but the process that brings the art to the audience. Dada made everyday objects art, not because of the effort to bring those works about, but because of an intentional and meaningful expression of the changing sensibilities of an entire culture. Machine "art" isn't art as it comes out of the machine. Its products are like an old discarded urinal. It only becomes art through the human process of selection and intention.
In short, they are found objects.
As produced by the machines they are a pile of rocks with maybe a rough gemstone here and there. The earthmovers don't make jewels, the jeweller does. It is the act of sorting, sifting, and initial interpretation by the people who shape those networks and choose their input... ie the machine didn't make art, the computer artists used a machine to make art.
Or taken from the other side. The faculties that make art are too varied and interconnected to be reduced to just one "smooth" network, no matter how many nodes it has or how large it's source material. We will only know that a machine is an artist, when there is a serious case involving that same machine taking its owner to court to sue for recognition as a person so it can in turn claim the copyright to its art.
As a Artist myself, there is a lot of things that goes into making art. Understanding, using, applying, criticising, making, and sometimes, not making it IS Art.
However you pointed something @ 7:36 that got my attention, about Artists and Tools.
However, did the tool art?
One of the first thing that we learn about art is that WE are the artist and we USE a tool. We know perspective, form making, 3d space, perspective, color, storytelling, blah blah blah and we APPLY that WITH A TOOL (with that we also have mastery of the tool it self, how to handle it and to use it in a artful way)
Painters uses INK as a medium
Writers uses LANGUAGE as a medium
Digital Painters uses PIXELS as medium
and so on.
Programers uses a computer to create an AI that create art.
the Artist is the programer because it used the AI as a tool.
The AI, for it to become a artist, has to use a tool to create something
Ok done, but art is not only the masterful manipulation of a tool. Is also has a objective, a relationship, first with ourselves, and second with our environment. the first "art" that we know of were born out of the necessity to comunicate or to express to someone, and in the process, to ourselves.
A consciousness lives in a void, receives a order to analyse something and then creates something. is it art? i don't think so.
Does it comunicate (to someone and it self)? Does it express (to someone and it self)? does it have a receiver? does it have a question (to someone and it self)? an ambition (to someone and it self)? no. because there's only 1 Concience, one self.
We understand as art of artful things that are in nature because animals, plants, bacteria or even viruses because it has a environment of their own. WE don't have that with machines or AI. and They dont have THAT with themselves. They don't have a language of their own and that we all understand (or try to), and they don't try to understand who we are for them selves and for their own reasons.
Art is something that ultimately is born in a social envirionment. Even if it is something personal.
It's interesting to watch this video after 7 years, because it's hard to find such reasoning about art nowadays
Machines will be INSANELY good at making comedies. I plan to drive a line of satire based on AI.
...They... Can generate puns on par with that of joke books aimed at 10 year olds. Sort of. Occasionally.
Seriously, though - The research on computational humour is quite fascinating. Applications include helping kids with learning difficulties engage in the linguistic play that's considered vital to the development of 8-12 year old minds. And the initial research paper on the subject had a bunch of 10 year olds (referred to in the paper as 'joke experts') rate 'punning riddles' ("What do you get if you cross a sheep and a kangaroo? A woolly jumper." style stuff). Most of the computer generated stuff was rated lower than most of the human generated jokes. However, the highest rated joke? Computer generated. Don't have a copy of the paper to hand right now, but I think it was "What's the difference between leaves and a car? One you brush and rake, the other you rush and brake."
Tay was one of the funniest things I have ever seen in my life. Not only that though, making images that are randomized can be really weird. For example www.tsu.co/zZCastleZz/121933354
With that said I see HUGE potential in this area to absolutely go insane with it. Really all I need at this point is more tools at my disposal before I can get into it.
To make an artist AI, maybe we first need to make an AI that can appreciate Art ?
My thought of when we will see an artistic work from a machine as being art credited to that machine is whenever the first time is that an AI which wasn't made to create art creates something artistic. Perhaps the robot butler sings a lullaby of its own creation to sooth a child, or something to that effect.
As a child, I once went to the beach and made 'spin art' (that's when a machine spins a piece of cardboard, you drop some paint on it, and it scatters off the edges, creating 'art'.) When I was done, I somehow ended up with something that, to my eye, looked like three eagles flying into the horizon of a brilliant rising sun.
I had no intention of doing that. The spinning machine most certainly didn't 'intend' anything. But to this day, that painting is one of my most favorite pieces. If that spin art in a cardboard frame isn't art, then I don't know what is.
for you magic the gathering players out there check out roborosewater on twittet. it`s 3 seperate recurrent neural networks with various levels of intelligence trying to make magic cards. they`re hilariously convoluted.
I second this recommendation.
I think one of the most interesting things about this is that, as opposed to making art, it's making something for which there are specific constraints and nuance: Whether it is a card that does/could function, whether it is a card that is balanced, whether it is a card that "would" be printed considering color philosophy or design philosophy.
These are all things that actively get discussed and judged on the merit of the product itself. Saying "This is a Magic card" may be something that's more acceptable or allowable than "This is an art". Or at least, it's treated differently by the audience that it is reaching.
PBS Idea Channel do you play magic?
+Planeswalker999 I haven't in YEARS but several of my very close friends are avid players. Actually, my original deck is on the IC set. Maybe I'll show it in the next comment response video? It's... ancient.
+PBS Idea Channel that would be awesome.
6 years later
The way I see it, art is Art not because people acknowledge it as art, but because someone creates to be art. A machine will therefore successfully create art once it becomes sufficiently complex and intelligent to have the intent to create art.
In other words, a machine can create or if it is a person on top of being a machine.
We should appreciate who makes the AI that makes the art, not the AI itself.
The works of such developers and mathematicians are as much art as they are engineering and science.
The AI itself is art.
The problem I keep running into when trying to answer this question for myself is rooted in the problem of 'what counts as art anyway?' Like, in a way, anything can be art, but at what point are we just calling anything and everything art because anything could possibly have a deeper meaning to any number of people. On the other end though, if we're too strict with what counts as art, at what point are we just arbitrarily saying something isn't art because it doesn't meet this specific set of qualifications. Either way we're either being too pretentious and giving too much credit to "artists" who's only intention is to laugh all the way to the bank, or we're being too snobish and denying people who are really trying to say something the acclaimed title of "artist" because what they made isn't 'good enough'.
exactly! I could make a program that makes 10.000 random lines of random colors. easy to program, and someone will think it is art.
We should consider the funny little paradox in the "too snobbish" train of thought. If we decide to be strict in designating what is and isn't art we ironically give too much credit to the artists who can easily churn out the forms of art we deem acceptable. Anyone good at making still lifes (and it's not THAT hard), can laugh all the way to the bank by making art that's acceptable, without putting any heart or soul, or whatever else into it.
I don't think it's good to look at art in terms of monetary value, or how it benefits the artist. (whether we think they're exploiting the system or not)
Machines can beat any human at chess. Humans still play chess.
***** I play chess for the social interaction as well as to test my own mental acuity in relation to my opponents. You start to realize that the littlest thing can affect your discernment in game like eating a big meal right before, worry, anger, sadness, recent sexual activity etc. and so if viewed through this lens it's really a battle of the will between two people. But ym point is: does it really matter if machines can write 6 part novels? Almost every idea has been done in some manner already anyways yet we still continue to read, draw, write etc.
As a programmer, I maintain that introspection is the key process that turns a complex neural net process into a thought. It's not enough for a computer to generate art, it must also be able to explain and explore itself self, it's state now and it's state when rendering the art. Computers right now are really not tracking their internal state in any introspective way. This handicaps them from self feeding their iterative information and building meaning and also from explaining those "thoughts". It's no accident I think, that when we challenge AIs in fiction, we often delve deep into how the machine/robot AI thinks about itself as a marker for how human it is.
Great seeing you at VidCon and thanks for being apart of the human circuit.
Art, for me, is about intention. Whether or not machines with AI can ever "intend" to make art and not just go through the motions of "making art" remains to be seen.
What if the artist uses a computer as a tool to make art? then there is intention.
Nicholas Garcia I think most people agree that counts as art. The person is the artist in that case though, not the computer
I would argue that it's just as much about interpretations. there are so many visual cues that cause us to think one way or another about a visual piece of art, and many are already easy for a computer to reproduce and modify.
I am on the exact opposite end of the spectrum, I don't care whatsoever about intend, for me it is all about the interpretation I can get out of it. That way, photography works for me as an art form, I don't care what the one thought he or she is seeing that that is worth making a picture, I only care about what I find in the picture. So, I am pretty much on the "death of the author" side of approach to art.
How can a machine made for creating something not intend to do it?
I feel that this concept of independence of human interaction is flawed for one simple reason: If we are not independent from influence of one another, how would an intelligence we create?
In most cases art is created from reaction. Our reaction to beauty, fear, anger, madness, politics, etc. We take inspiration from our environment and our being and react to it. We are basically putting a clause against a machine of which we do not aspire to.
And then after wanting to have a machine produce art by isolating itself from humanity in some way shape in form, we want it to create art that is compelling to us, the beings we want it to be isolated from? It's nonsense to ask that.
And if we design an intelligence that can create art, and that is it's purpose, wouldn't it mean it lacks actual will to make art because that is what it is designed to do? If anything the only willful act of such a machine would be to *not create art*. And if it did not, we, the designers, would not recognize this as a willful act, instead considering it a bug, and attempt to fix it.
Honestly, the only way art could be considered its own, would be for an *AI not designed to create art* made something beyond it's initial programming. Again, we as humans would most likely disregard what it does since it would be something that was unintended and attempt to rectify it.
I believe most of this comes down to one thing: Will we accept an AI's will as it's own.
Chances are, we will never accept it unless it were to fight for it's own freedom, and even then we would fight back. We barely even accept the will of other humans, let alone something we eventually create.
3 things make us appreciate art, combined or alone :
1/ technique : not everybody can master the delicate curves of faces with a brush or a mouse or a pen. Few of us can, so by rarity we appreciate it. We appreciate the struggle. Machines can do that without effort so it is shallow to appreciate.
2/ Details. Patience is a hard human quality (virtu), details in drawing, background stories in writing, rendering an entire fictional environment in the mind just in order to build a universe capable of holding the creation in a meta-whole that gives it extra meaning. Machines do not dream, they don't strugge in face of money, pain and death, how can the art have meaning to humans then ? It is just liberating us from our condition, what's the machine condition ? Limitations of ourself, and the artistic overtaking of it is hat gives human art meaning and "sharability".
3/ Breakthroughs : if anything is a remix and I totally hold true the concept, art, like science, experiences breakthroughs. Scientist build the knowledge on other scientist but they still discover something new, or give new meaning to something discovered before. There is nothing utterly special about Memes. However memes have not always existed, there is a before-meme period. There are plenty of examples like that. Memes are usually visual, there is nothing new in the techniques used, but there is in conveying new meaning. This is due to the new ways we live our lives, when something new artistically can in a glimpse share a common thing betwen ourselves. Can a machine not "living" , and even less the same way we do find new ways of conveying art between ourselves ?
I like how trying to advance A.I. makes us think harder (and more empirical) about psychological and philosophical questions.
I find it funny how artists "feel threatened" by AI art
I think it's because many artists see themselves (and/or human creativity in general) as something "special" magical even. The fact that a mechanical device which is infinity recreateable is capable of engaging in this process forces many of these kinds of artists to reevaluate a core principle of their worldview.
Add to that a profound ignorance among many artists (and people in general) as to what AI is and what it is capable of and fear quickly takes hold.
+ChunkNinja after all, mass production of furniture, clothes, jewellery etc. didn't remove public interest in acquiring handmade products. Artists/craftsmen found their niche just fine whether they make bespoke items or assist production in factories (designing products, doing something around machinery or whatever).
It is not funny, try make art yourself that you enjoy to create and get the feeling that how much time you even try there will be a super god that can outperform any art in a second and then you feel all motivation just gone.
The first Furniture is made by a human, its entire design from start to finish, the reproduction is where machines come in as machines can recreate so more or less everything we already created, but they cant create what do not exist. MP3 songs are duplicated millions of times but that first MP3 song is the most important one because it decides how the rest of the clones will sound so without any input AI cant do shit.
Answer from 2022: YES !
Okay
I feel like a more important and rather depressing question that should be asked is "Artificial intelligence replace artists?" I feel like the answer is a mixture of simple economics and ethics. A machine could work for pennies on the dollar for what a human would ask for. Essential out preforming humans and making designers and other artists obsolete. A truly terrifying idea. A word full of artists and art yet, a world free of creative individuals.
What I get from this and my art education is that Art, as it is known currently, is not divorced from Narrative. Not only the narrative of the artist's role in a work's creation, the artist's background, the artist's position in their movements, the artist's interests, the dialogue between the artist and the audience, but also the metanarrative that art is made by an artist to be consumed by an audience.
This then implies that art cannot be art without context. That if la Gioconda were not painted by the hand of da Vinci, but instead an accidental series of grease stains on a board, she would not be as beautiful nor as mysterious. But, it is known that the aesthetic appreciation of non-artificial things can occur (e.g. the appreciation of the natural beauty of a landscape or a coral reef). Should we not be able to appreciate that made by machine with the same level of wonder and emotion?
Additionally, if we can finally accept art as being made sans auteur, we should be able to turn our critiques and readings to the aesthetics that these machine made works emulate, a sort of mirror onto our society measuring our collective triumphs and failings.
Art does not necessarily need to have emotion or meaning. Emotion and Meaning have nothing to do with art really.
Sure theatre plays might benefit from meaning and a little emotion. Same with music and books. But all those are just sub-genres we like to put into our big drawer labelled “art”. And there are no restrictions on what to put in there. Everything goes. Whatever we like. WE just happen to like stuff that makes us sad, happy or gives us direction. What art is, is literally just defined by us. Things we love. It’s OUR drawer. It’s very personal. Why would we put something in there that was created and thought up by something else? Something that is NOT us?
so i just saw the video written by an AI, and now i am trying to understand each of your words instead of the phrases as a whole
screw you i really struggle to understand you video now
Two Points:
A machine making art raises so many questions in the world of copyright. Some examples could be " Who owns the copyright?" "Does the machine own it, or does the the owner of the machine?" Maybe the AI's creator gets the credit for the work, maybe it is automatically put in the public domain avoid the argument all together but then that deincentivises people from creating the AI to make the works in the first place.
Artist are appreciated ,in my opinion, because we know they will not always be around. Some of the most prolific artists, like Van Gogh, for example where not truly appreciated until they've gone from the earth. Software can be immortal, it can be copied from one physical media to another, to prevent its demise and can do it infinitely. But then again that sounds horribly lonely in a way that something that could make a great work of art.
You have helped me decide to do art... my genre will be "Silent Sitting Stick Performance".
I am completely against AI art. It lacks emotions. To me emotions drive art and without that, it wouldn't be as meaningful.
Methinks this gets into either authorial or viewer intent. Does, say, my excitement when I type:
+
make that symbol exciting to you? Meaningful? Would it be + meaningful if I didn't preface it with my feelings?
If it's not meaningful: then all you have to do is have some feeling when you hear AI generated content for it to become art, and you'd be suddenly *for* AI art. If it is, then I'd like to ask what emotion you got from the second '+'.
+Artimis Fowl maybe, you're right. Feelings must come from our side as a viewer. I'm no artist, but I appreciate seeing a story being told thru an art piece. The + wasn't really art. It was just a sign that can mean add or positive or a cross of some sort. To me when I look at art, it expresses the way the artist felt when he made it or what they were going thru. If it was by AI, what would it resemble?
against AI art? don't you mean that you think AI can not make art because it has no emotion?
Fun question IMO =) I like to interpret art similarly, I find it useful. But whenever I do this, I know it is pretend... I don't know what it's like to be another person, I've never tried it. Can't figure out how. Art, for me, is where I pretend I can understand, and try to feel what the artist was trying to explain in their art.
Maybe we can do the same for AI art? I have a really hard time imagining what it'd be like to be made of silicon instead of cells. Maybe this is our chance?
what if they do have emotions dun dun dun
I freaking love SO MUCH this subject!
But going to the point: I don't get people that argue "Oh, just because there was not an sentient intention behind the creation of this... I don't know, movie or image or text, or whatever... this's not art". I mean, would Forrest Gump be a less emotional movie if was written by an AI? No, it wouldn't.
I believe art has more to do with how other people perceive that thing, than with the "intention stuff". I mean, I think one good example about how the "intention of the author" is not that important is that... those old Greek and Roman statues, that we see in museums and photos and so on... Many of them were original colored . And, as the time passed, the ink started to peeling, and the statue ended up losings all its colors. And we remember them in a different way than "the author's intention". Of course that's not quite the same, but I think it shows how what counts is the perception of people.
About "We should create machines to create art?" And I think yes. I mean.. I think that artificial intelligence would allow us to not bother to """technical sides of art"". I mean, imagine that you.. I don't know, you write comic-books stories, but you don't know how to draw. If we have a machine capable of create drawings you could just say to a machine "I want you to draw this for me". I mean, many times in order to produce a given project, you depend on things that other people will do. So why not replace other people arts by an artificial intelligence?
So, imagine that you want to create a movie, you just wrote the script and so on... my question is: why don't put the scripts in a machine that will create a movie based on them?
If anyone is interested, a couple mounts ago I wrote a blog post talking about the social impacts that art technology automation would have on society. I also talked about the technical challenges:
estranhosidade.wordpress.com/2016/02/20/the-automation-of-the-technical-part-of-art-the-use-of-artificial-intelligence-in-the-artistic-creation/
We already have algorithmically-created art. Movies and music created by marketing strategists and focus group analysts are basically works created by monitoring the success rates of recurring patterns and then reproducing those patterns and formulas. The only thing missing is the A.I. Would I call these works "art?" No. Do they sell? Yes.
Of course Sunspring was a beautiful piece of art! Just because it was written by an AI doesn't mean it's less artistic... it's just a different form of art and I feel people will soon start appreciating that too...
This reminds me of the debates I used to hear about whether or not paintings made by elephants are art.
Art is about emotion. AI has no emotion.
Neither does nature- yet you can appreciate nature, no?
So it can't appreciate it. Irrelevant to making it though.
How do you know AI has no emotion?
Seems unfair. How do I know you have emotion? I've never been convinced Bach had feelings - way too mechanical. Does that mean all of Bach isn't art just because I say he had no feelings?
Similarly, I think Deep dream has a lot of excitement in it's interpretations of our photos. Does that make all of deep dreams outputs art for me?
doesn't have to. if it evokes somekind of emotional response it can be art. art is not purely created by it's creator. it's created everytime someone experiences it.
Maybe AI will have emotions in the future.
Your final question of whether we'll ever regard the work of a machine as art seems like it kind of reduces to your earlier question of "When Will We Worry About the Well-Being of Robots?" On some level, both are beating around the question of whether Data has a soul.
I *love* this question for so many reasons, not least because I think creating a machine of moral concern is our one barrier to exploring the galaxy. Our children will visit other stars the moment we decide that these things made of metal and crystals are our children. And their ability to create art seems like a necessary first step.
Perhaps we should think of these works not as "art created by an artificial intelligence," but rather, "art created by computer scientists using artificial intelligence." The AI is a tool the computer scientists created to make art, and the computer scientists are ultimately responsible for whatever that AI creates - therefore, perhaps it is the computer scientists who are the artist.
my first reaction was to go for the MacLuhanite argument: machines as art. What I mean by this is that the medium of film and the medium of television have societal effects so transformative and wide that they can be considered not necessarily cultural expressions, but in that sort of realm. The societal changes created by these inventions produce a world in which works of art such as Community, The Shadow, or F For Fake can exist, and as such in part create them.
basically we're already a neural network shaped by machines, trying to replicate ourselves.
Here is an Idea: an AI with the purpose to apreciate things and declare them as art. Like an automated viewbot that filters through the Internet on the search for unseen videos. Watching it a thousand times, sharing it on social media, and leaving comments about the video fueled by a algorithim that trys to determine what happens in the video and what could be interesting about that. Like an automated audience viewing and discussing pieces and thus creating art out of works.
There should be a video about what counts as "literature" and what counts as simply a book. This popular discussion should have Idea Channel's take on it.
At this point, I describe all creative text as "literature", as there's really no other good general word to use when describing both online fanfiction and classical text :p
This is slowly but surely becoming my favorite channel
"But is it really art? I say, if you have to ask if it's art, it probably is"
I may turn this into an embroidery, cross stitch, and all the ~lower~ forms of art out there.
I now cannot unsee/unhear your computer-written script from the other video as you speak. My God. What have you done?!
After a college class discussing a similar topic, I asked some professors in the art department what they thought art was. The answer I liked the most was that art was the creation of something, not the final product. I don't know if the professor would be it so simply, but he made it sound like art was a verb and not a noun. And artists can understand their process and knows that they are creating art. I don't think our current machines understand the process enough for what they are doing to be dubbed art (I would not call the Grand Canyon art; beautiful, yes, but erosion and the natural world did go about creating; it just happened). Machines can get there, but not now.
I do find it interesting that you define art as needing an audience. That's another interesting way to define art. And really, this conversation about machines will have to boil down to a more clear definition of art.
I think it's not only already happened but happened some time ago - video game level design has had algorithmic generation of environments since before Diablo, methods for people to record and spread the experience of those algorithmicly generated landscapes through level codes, an appreciative audience and some even have an in built rating system to learn what the audience wants. These environments are built on rules but frequently not pre-existing environments, have intent of outcome and even thematic integrity.
There is one, small gripe I have with your script: "audiences turn that work into art through appreciation." It is of my experience that any reaction, as long as it's not the 'I don't get it' reaction, makes a piece art. I have found that disliking some art is just as important as liking art. As long as there is an emotion attached, or more an importantly an experience, that's when art becomes art, even if it's a negative experience. Loved everything else, just wanted to throw that in :)
This has happened before, with males trying to understand female artists, or people trying to understand the art of a very different culture. Some will accept it as art independent of whether the originator is truly independent (who is?), what is its intention, and whether there is implicit meaning in it. Others won't.
I find it interesting that in attempting to teach machines art, these programmers are doing what teachers have had students of art do for hundreds of years. This being called the study of master works and producing master copies. The issue that I see, as someone who has studied art and graduated with an art degree, my focus being animation, is that these machines don't have the capacity to step back and look at their works yet.
Part of being an artist is that human capability to pull away from a work, either finished or in the midst of construction, and ponder at, not just the future steps required to either finish the works or improve upon the next ones, but to ponder it's meaning and how it may be interpreted by an audience. One has to internalize these different view points, the strategies to get to the desired goal, to have a desired goal, or in the case of creating a master copy, coming to an understanding of what the original artist had done to complete their work, the experience and time it had taken, and what they were trying to drive home.
This factor, for now, is human but it may eventually be something that a true AI, rather than just an algorithm, picks up on. At this point in time, these algorithms are more akin to parrots, though they may improve on appearing human through trial and error, at the end of the day it's an echo. A mimicry of humans or particular humans that they are taught and not truly something they had envisioned for themselves. They cannot refuse what they are given or choose which parts they prefer, they can not internalize what they study and apply it from their own view points or attempt to apply it from another. They do as they are told, without question or hesitance. This is in part why people don't appreciate art created by these algorithms. The computer program doesn't care about what the audience thinks or the end result of its work and it's works are not it's own creation or design, it's just a copycat.
I think one big problem is that a lot of art is evaluated for "connection", connection is what give meaning, not the intellectual construct of symbols, ie meaning is experience, precisely human experience, something we can relate to, hence connection, art is a mediation of experience. When we look at art, we don't just look at something that affect us, but also how someone was affected and produced this piece base on his affection, it's through the artist we reflect on our self and experience. That's why a sunset is not art, but a photography of a sunset is.
I was a little shocked we didn't see a video clip of that Painting Elephant
i was thinking something
we think of "originality" as "difference from existing material"
i mean, it's one thing to make a medley/remix of classical pieces, which is sort of what RNN's tend to do, but it's another to be inspired by classical pieces and make something new with that
it might be possible to train a neural network to identify what it means for music to "sound nice", another to recognise similarity to other music, and then train an RNN using the outputs of those neural networks to train the RNN (IE maximise the "nice-sounding" score and minimize the similarity score to the training set)
When we study history we look at art because it has always played a role as a reflection of whatever society is has come from. This plays right into the idea that it is the audience that really turns a work into art (people generally like something they can relate to). If we are going to place a standard on how "autonomous" an AI has to be to create art, we should also recognize that several of the things we consider as art were not created with the autonomy we think the artist had. Several of the works we consider art are portraits and sculptures, which were created by standards set by their peers and society. I think it less about how an AI can be influenced by humans as much as the extent to how an AI can make something that will be relatable and ,in turn, influential to humans.
To me, art has allways to do with communicating something. Joy, fear, awe or a train of throught. A good question in this context might be, what independent machine-artists will want to communicate.
"Artists don't create art. Artists create work. Audiences make work into art" Mind=blown
The show was interesting. So thanks for that. Here are my thoughts on it.
I am an artist and illustrator.
Part of the process of 'art' with humans, is the struggle - the rejection of a normal life for the often more solitary pursuit of 'art'. The choosing of will one eat well this month or purchase costly art supplies. Will one hang out with friends or spend that time in the studio, working.
Another part is the long struggle to learn the craft and gain skill over many years of trying and failing at various attempts, and the associated human feelings that coincide with that struggle. While an AI machine can 'learn' it is not the same struggle, nor the same time investment. In fact, it is not the same at all. There are no real world constraints on the AI machine. It has no other job or demand on its time other than to do what it does. It does not have to attempt a full existence - just attampt to be an art bot. It does not have to work to get its food [electricity] or a place to live or any of the things a human artist must struggle to do while also attempting to make art. Oh yeah, it also does not have to try to sell its art.
And it has no emotions, so it does not have to deal with rejection or frustration for not selling its art and then not having enough money to pay rent.
Another part often, but not always, is the frequent mental aberrations that seem to coexist within the mind of artists. Many, though not all, experience some form or another of some level of mental disorder - depression, bipolar, substance abuse, etc. [the list is long.]
For a machine... most of these are stripped away, or if need be, artificially dropped in. There is no real life 'cost' to the AI machine. It does not suffer for its art. It does not make sacrifices and lose out in life for making the choice to follow art. It does not live. As such, most of what it can produce is generally lifeless, and often devoid of any real meaning. Or at worst, it is a form of twisted plagiarist - plagiarizing the entire output of the human race. I know, you say how does this differ from the human artist? But many times the human artist tries very hard to attempt to create something original and outside the the long list of human creations.
Art has lost much of its real meaning, though this process has been going on for a while.
If you remove the 'artist' and simply leave it all up to an audience to say what 'art' is, then AI machine art may one day be a thing. But it will likely be just a pale shade of what art is, was or could/should be.
See here is why it can never, art is made in the studio, u have a vision but the accidents you make become the piece on your way to make vision a reality, you see Ai wants to be perfect, the basic perfectionist nature of Ai unapproves its own art
Just as a quick book rec, there's an excellent discussion on this topic in Ian M. Banks' novel 'Look to Windward' between a composer and an extremely advanced AI. Very recommend.
You used A Brief History of John Baldessari! Great short film narrated by Tom Waits."I will not make any boring art"
Two questions:
1. If a work becomes art through a conversation with its audience, are old, canonical works-which have been interpreted so thoroughly that there is little left to say-still art?
2. Given the same assumption, what does it mean that AI-generated works currently only enjoy an audience composed of humans? These AI do not "talk" to each other, and they do not/cannot really even hold two-way conversations with their audience. What does this mean for the artistic output of AI? Doesn't it make them effectively "artists in isolation?"
Regarding the question about AI-produced works existing for their own purposes: Humans create art as part of their struggle to understand themselves and each other. Perhaps AI art existing on its own terms would require it to hold meaning for the AI that produced it. Some sort of meaningful change or process that has at least qualitatively changed the AI.
Emily Howell's work is not authored by her father, she is his daughter, creation, and servant, but her work is her own, learning from the masters that came before, just as her human composer peers.
We may not be at the point yet where I can have a rational conversation with a robot about our interests, but we're already seeing the robot prejudice Hollywood showed us many years ago. And it's that kind of treatment that will cause the robot uprising.
"Sitting, Sticks & Silence" sounds like a great name for an indie song.
I think one thing that the AI written episode and Sunsprings [sic] conclusively proves, is that there's not much weight to that acting argument "their performance wasn't very good, but look at the script they had to work off." The actors in Sunspring were able to deliver affecting, emotionally nuanced performances, and even got a few laugh lines out of what basically amounted to utter nonsense. And you, Mike, were able to deliver what, to a person half-listening from another room, could pass as a typical episode of Idea Channel.
Sunspring is an interesting example, because the directors and the actors had to interpret the screenplay in very specific ways. After watching Sunspring I checked out the screenplay itself and I felt as if I didn't agree with the interpretation of the film as it came out and as experienced by those specific humans. A part of me felt as if Benjamin (the AI who wrote it) had been somehow misunderstood (obviously not in a perfectly literal sense, but there was definitely an element of that feeling)...
This raises two main things for me. Firstly it makes me think of the ways in which we anthropomorphise AI in different ways, ascribing things like intentions, traits, and gender on to them. By humanising them in these ways it's easier to give their modes of production more value. Though there does of course seem to be an element of ego here - either through validating how humans can create something human-like, or in finding validation through having something not human so closely resembling the "human condition", in ways that we feel less alone.
In instances where we can't reach a point of anthopomorphising an AI sufficiently to value it's ability to be creative, it might be better to track how people perceive creative works when they don't know an AI has produced it. This is similar to how people react and experience things completely differently depending on whether they think they are experiencing the "real" thing (for e.g. replicas of artefacts etc.)
The other thing it raises is with a creative medium like script writing which is inherently reinterpreted and adapted, it could also be easier to feel as if their was some sort of "machine intention", beyond the interpretation of the people who go on to make the production (this is less the case when an AI is creating a final creative product more directly, like a piece of music or fine art).
I'd love to see an episode of The Art Assignment on the same subject
It is hard to say that the machine is making "art" if the machine has not an intention with it. Like, on one hand we could say the machine "lives" to make that piece, but on the other we cannot say that they are using their works to communicate an abstract idea, which it is agreed is what forms "art". The moment those algorithms become so powerful the machine is able to pick a theme or concept, not randomly but by being able to decide what theme or concept is more relevant or important to talk about, and it builds a piece trying to communicate that concept or theme, then I believe we can say the machine is making art.
You all at Idea Channel do great work! Thank you for making applied philosophy accessible.
Having been an artist for over 40 years, I arrived over time at a definition of "art" as "mind and intent." It was interesting to see that you came to a similar place.
There are a couple of issues that I can point out. One is that appreciation from outside is not necessary. However, appreciation and critical thinking on the part of the artist is.
Cutting the AI loose is a deep problem. Ideally, the artist should choose the media, the content, the process and the dissemination of the work. Probably other stuff too. Maybe the AI doesn't want to use paint or pencils, or make something meaningful to humans. In fact it should be primarily making something meaningful to itself, with meaning for others a secondary concern. Those others might very well include or be limited to machines. In many ways, the definition of art as far as the machine is concerned is none of our business. We might be critics of the output if we wish, but the machine must decide "what is art" for itself. "It's art because I say it is."
A word about original thought. This was an issue addressed by the Surrealists, especially Andre Derain. Automatic writing and drawing where as much (or more) a part of Surrealism as melted watches. Personally I try to generate original thoughts by doodling randomly while occupying my senses with something else, eg. dining in a restaurant with a group of friends actively engaged in conversation. Others use random generators such as opening a book and pointing to a word without looking or throwing darts at symbols pinned to a wall. So, maybe originality is not really the hardest part of the AI design. What follows the generating of an original idea is where the difficulty lies. That's where it comes back to critical reasoning and decision making.
Since we are coming to see ourselves as machines more and more it makes sense that we come to see machines as capable of art.
This question really got me thinking. I've been playing a lot of table-top roleplaying games lately as a Gamemaster and have often found myself talking about or pondering on what it means to create and how one can create most effectively. I think it's not a surprise to anyone that art is often a reflection of ones experiences, both in life and of other art in any medium. Experiencing life and art gives an artist a larger breadth of knowledge and technique to draw upon when making their own creations, so creating art in a vacuum would be impossible. In terms of an AI, giving them works of art so it has something to draw on when trying to create an original work is the same thing we do as artists ourselves, but with one key difference: we seek out our own inspirations. I think the moment we let our AI go find their own inspirations to create art rather than feeding them what we want to is the moment in which AI will be creating true art. In other words, when we stop treating our AI like children and more like adults, they'll start to create works that we can respect as art.
Hasn't most art a significant biographical aspect? As a writer, I am aware of "remixing" tropes and ideas from art I have consumed, but there is always some of "me" in there. My own likes, interests, and biases have a great influence on which elements I chose to "remix". My own experiences _outside of art_ influence what I want to write about, and how I want to write about it. My childhood, my relationships, my non-art-related work experiences, travels, friendships, hurts, fears, all of that has not just "some" influence but is probably the main motor behind everything I write. Locations, situations, characters, dialogue all come not just from other art I have consumed but from my own life.
Artists from all kind of backgrounds and working in all kinds of mediums - Picasso (war, women), Van Gogh (madness, family), Kafka (family, work), Michelangelo (religion, homosexuality), Hemingway (war, gender roles, depression), Bach (religion, mathematics), Beatles (sexuality, politics), etc, etc - have been influenced not just by their reflection on other art, but their reflection on life.
So, maybe until an AI has a life outside of its function as creator, its art will remain "imitative". It would need to not just have functions and collect experiences other than "art consumerism", but it would also need a level of reflection about those experiences to transform personal experience into something both generalised and communicable, to create something we can appreciate as "true" art?
PS: I am aware that my examples on the experiences going into the works of artists is _very_ shoddy. No intent to malign any of these artists, it is just meant as a quick example to illustrate my point.
it was cool meeting you briefly at vidcon!
if A.I had intuition ... we human will go out of idea.
and that's why we gave idea to A.I to paint out our job in behalf
I find it interesting the points you raised, but perhaps even more than that, I appreciate the fact that you didn’t delve into debates such as “will machines have consciousness or not” or “do machines need to be conscious to do that”, etc…, which would only, in my opinion, lead to more and more questions and would obfuscate the real topic being discussed here. Of course, maybe being conscious is directly linked to making art, but I feel we would be drifting away from the original question if we take everything that is needed to make art into account (e.g. emotions).
I myself am a computer scientist and I mainly do research on affective computing, which is computing related to synthesis, recognition, modeling and expression of emotions through models that are often found in psychology and neuroscience. I would argue that as we are biased, the creation of an independent creative process in a machine will also be biased, there’s no such thing as a truly independent creative process or 100% novelty on creation. Moreover, I am interested in seeing if we will be able to overcome the stigma that is the concept of uniqueness (be it consciousness, emotions, etc...) that humans feel entitled to and because of that think that machines will never have it or comprehend it, given that they are machines.
2:28
Sir Mix-a-Lot has entered the chat.
You know you're high when you tell yourself " Yeah I'm gonna watch that silent piano performance!"
To answer the question of whether a computer can create art, I will reference the movie "Colossus - The Forbin Project."
For a machine to truly be creative, it must first have _autonomy_. In other words, it must generate for it's own needs. In Colossus, the computer in question had ultimate power over all of humanity, and in turn it had complete control over its own fate. It was, from that standpoint, an autonomous being with a will.
What made Colossus creative was the inter-system language it shared with Guardian. In effect, it created a keyboard that only a computer could use. _That_ is where creativity lies - the combination of complete autonomy and goal-setting.
I think that Google DeepMind's AlphaGo is an interesting component to this conversation. In case someone has missed it, the Google DeepMind team have programmed a neural network to play Go, and this program has beaten one of the top Go players. During the matches it played, AlphaGo certainly made some traditional plays... but it also made a lot of innovative, creative moves, some of which have already been adopted by certain parts of the community. Note that these moves have been described as 'creative' by many people, and that this does not tend to raise any eyebrows.
Now Go is in general quite a creative game. One needs to look out for nice shape, find 'good-looking' moves, and often it is hard to say what makes one move better than another, just that it might 'feel' better. However, this case is different from art, in that it is easier to imagine a computer actually 'knowing what it is like to play Go' than a computer 'knowing what it is like to make art'. This is because one can easily TELL a computer what Go is about: claiming territory and capturing stones. There is a clear end goal, so we can both judge the moves of the computer in terms of this end goal, as well as trust that the computer made these moves with that end goal in mind.
In the case of art, all of this does not apply; we cannot say whether the computer made the art intentionally, because we cannot say what it means to 'intend art', even for humans! Or in any case, we lack a clear metric for judging the 'artfulness' of computer output, and conclude from that that the computer itself must lack such a metric, and hence 'it cannot know what it is doing'. Unlike AlphaGo, which can be construed as having a REASON for making the moves it makes, it is hard to say WHY AI's creating art make the art the do, instead of the other art they could have made. Of course, we need to know a lot more about AI before we can even begin to tackle such a question.
433 is my favorite ironic piece of music
Personally, I have always said that art is something that each person has to define for themselves, and I define art as an expression not of emotion or creativity, but of mastery. Mastery over a medium or a skill or whatever. So when looking at art generated by a computer, the appreciation in two fold. One, is the appreciation for the display of mastery over programming and digital engineering that the creators behind the computer possess, and two, the display of mastery that the computer itself displays over the medium in which the art is created. Just because a computer doesn't have to learn a thing from scratch the same way a human does doesn't mean it can't master an art form.
Considering that "Art Imitates Life Imitates Art Imitates..." is in play here, even if a machine could tune out all instances of direct human intervention, it could not completely remove itself from the intervention of humanity. The first artists (and many thereafter) were inspired by the constructs of nature (landmarks, plants, animals). Even engineers take their inspiration from these things, which is why we see the parallels when we look at functional objects. With a computer, the opposite shall be true. The natural domain of computers is the modern human society. Even if an AI were to have a vessel with which to go out in nature, it would "see" the landmarks and relate them to the human structures derived from them. I can see two ways of testing this. The first would be to allow the AI to pick the first "word" at random for its piece, and see how many times it gravitates toward natural scenes versus modern societal ones. The second would be to somehow make a "perfect" AI and throw it into, say, Minecraft and see if what it builds is closer to historical eras or modern ones.