A friend of mine got me a job as a software engineer at an AI copywriting company in the 18 months before the public release of ChatGPT. We started by having GPT-3 (3.0) write small snippets of ad...
Exemplary
A friend of mine got me a job as a software engineer at an AI copywriting company in the 18 months before the public release of ChatGPT. We started by having GPT-3 (3.0) write small snippets of ad text and product descriptions, but as time went on it was clear the next-big-thing would be long form AI writing. I had a lot of fun building the app. It was my first time building a website from the ground up for a successful business. But when my friend asked me if I'd ever use it to write something like a blog post I was adamant that I never would. Honestly, what's the point? If I'm writing something for people to read I'm writing it because I know what I want to say, and the act of writing refines my thoughts until I've written exactly what needs to be read.
Where does an AI fit in here? I could definitely use an AI editor or reviewer. I could get writer's block and need just the right phrase to get un-stuck. Or maybe I realized I've covered too many topics and could use suggestions for which ones to cut. But the AI doesn't have ideas of its own, it's just an accelerator for my keyboard. My friend never really understood what I meant, which is shocking as he is the one with a degree in creative writing and I'm the one with a CS degree. The only explanation I can think of is the old quote:
It Is Difficult to Get a Man to Understand Something When His Salary Depends Upon His Not Understanding It
My friend saw this startup as his path to lifetime financial security. He'd been given a generous equity package and if it did really well he'd have many millions of dollars. He was even at the point where he'd say things like "Hey if this works out I'll personally buy you a house."
But it wasn't meant to be. OpenAI used the startup for feedback on GPT-3's outputs (we had weekly reviews with an OpenAI rep). They used our inputs and outputs to train 3.5 and 4.0 and then released ChatGPT which more-or-less obsoleted the whole product.
You've succinctly captured my feelings about using GEN AI for ideation, it really can only come up with the most obvious things, so while it can seem convincing to get the ball rolling, truly...
If I'm writing something for people to read I'm writing it because I know what I want to say, and the act of writing refines my thoughts until I've written exactly what needs to be read.
You've succinctly captured my feelings about using GEN AI for ideation, it really can only come up with the most obvious things, so while it can seem convincing to get the ball rolling, truly novel insight remains human for now.
It's an energy activation thing, it's easy to type in what you want and have AI give you back obvious things, it takes hard work and thinking to get something novel.
But the risk I fight against is doing that hard thinking and coming up with the same obvious things the AI did, since there's no guarantee there will be any spark of inspiration.
So I guess all we can do is avoid being lazy, and be more brave.
Reminds me of Pixar's rules for writing: "Discount the First Idea That Comes to Mind. And the 2nd, and the 3rd and 4th and 5th. Get the Obvious Ones Out of the Way. Surprise Yourself." AI could...
Reminds me of Pixar's rules for writing: "Discount the First Idea That Comes to Mind. And the 2nd, and the 3rd and 4th and 5th. Get the Obvious Ones Out of the Way. Surprise Yourself."
AI could give you 100 ideas, to eliminate obvious ideas to help you get to the novel ideas.
I'm confused about the start up part of your friend's journey. Did openAI ask you to write this app? Was it paying for these input output or was it paying for your time on a monthly basis? How was...
I'm confused about the start up part of your friend's journey. Did openAI ask you to write this app? Was it paying for these input output or was it paying for your time on a monthly basis? How was it going to make your friend mega millions, vs they only ever wanted human data wranglers to train their LLM until its output is as good as your app's? Does your friend feel it was a fair exchange for his time?
The app we worked on was bringing in a few million in annual revenue on its own. GPT-3 was a closed beta. Helping OpenAI was a condition of beta access. And of course, if you’re friendly with...
The app we worked on was bringing in a few million in annual revenue on its own. GPT-3 was a closed beta. Helping OpenAI was a condition of beta access. And of course, if you’re friendly with OpenAI they’ll give you super duper early access to the newest models. If you’re trying to be a startup riding on their coattails they’re the ones playing kingmaker.
oh, wow. I guess things would have played out completely differently if (1) OpenAI kept their LLM to enterprise level access that companies can license for access (kinda like AWS?) or if (2) it...
oh, wow.
I guess things would have played out completely differently if (1) OpenAI kept their LLM to enterprise level access that companies can license for access (kinda like AWS?) or if (2) it took them long enough that your friend's startup can be sold for mega millions.
The way I understand it is like, if OpenAI had an orchard full of newly hybridized berries, and your friend's start up was a fleet of trucks bringing new berries to market at a rate like never before. And then they invented "teleport u-pick" open to everyone, including people who just wanted one berry a month. Kinda like that?
Yeah pretty much. It’s funny because our investors would always ask the CEO if we were safe against OpenAI cannibalizing the business. He was adamant that OpenAI would never want to handle the...
Yeah pretty much. It’s funny because our investors would always ask the CEO if we were safe against OpenAI cannibalizing the business. He was adamant that OpenAI would never want to handle the “last mile” themselves.
I think it's very natural for authors, musicians and actors to focus on the creative act that ChatGPT supplants when it's used, but I'd argue that they're only one half of the puzzle. One of the...
Exemplary
I think it's very natural for authors, musicians and actors to focus on the creative act that ChatGPT supplants when it's used, but I'd argue that they're only one half of the puzzle. One of the things I've realized in putting my own writing out there is that people read it in a very different context than I'm writing it, and my intent and goals in putting together a specific passage are only a portion of how they ultimately read and understand it. We see what we see and read what we read based not only on the work itself, but on the lifetime of experience and context that we supply when we read, watch or listen to that work.
In a comment above, the OP says that "once you find out [the work is AI] it becomes immediately degraded." And that may be true, for many people, but I find it odd to think that we would discard the meaning and ideas that we create, internally, upon seeing something, because we did not like the process behind them. Many artists have no particular intent or struggle in crafting their work, and it doesn't prevent people from deriving enjoyment from it.
The viewer is the final part of the creative process and still has heavy lifting to do; I don't subscribe to the notion that the lack of human involvement somehow handicaps reader enjoyment. The lack of facility in current-generation models is what does that. The same words, written by a human, would strike the reader no differently in a blind comparison.
We are fast-approaching a world where AI images and writing are going to effortlessly mimic the best of human creation. If you don't think it's coming, then you haven't been keeping up with the pace of acceleration - in a few short years we have gone from amateurish ramblings and abstract art to something very close to indistinguishable from human works. There will be something created that is poignant, or inspiring, or thought-provoking to you despite the utter lack of human involvement in its creation.
At that point you can either reject the feeling as artificial, or embrace it and recognize that the lack of an artist doesn't devalue what you feel - it just makes it wholly yours.
It does feel like the whole discussion of defining real/fake art is reductive and reactionary. And when most career artists are successful because their works are commodified, it sort of makes the...
It does feel like the whole discussion of defining real/fake art is reductive and reactionary. And when most career artists are successful because their works are commodified, it sort of makes the argument ring hollow.
I do agree with your idea of effort not being a requirement of art. My favorite image is the Pale Blue Dot. Compositionally: light streaks on a black background with a single white pixle right of center. Contextually:It's a photo of earth taken from Voyager 1, about six billion kilometers away.
We could debate for hours the semantics of this image. Is it even "real" art? Who should be credited with so many people involved in its creation? Should images like this be copyright protected? I don't think any of it matters. I look at the print of it every day and it holds a wealth of inspiration and consideration.
I think the more important discussion around AI art should be in its consumption. Like you said, art evokes an emotional response, so what happens if people could fine tune exactly what they want to see and feel.
Thinking about modern social media and internet usage, people are able to silo themselves to an antisocial degree and one of the few touch points are the art we share. Movies, music, games, TV. It's falling away with how algorithmic content has become and I don't know how to feel about people only consuming content designed and made just for them.
Combined with the explosion of profit driven social substitutes like AI partners, friends and therapy; fake people spamming content and very real social engineering risks; it feels like people are missing the Forrest for the trees.
Often when I read a great book or watch a great movie (or engage with any kind of art really), I find myself reading about the writer or director or whoever. I'll often seek out more of their work...
Often when I read a great book or watch a great movie (or engage with any kind of art really), I find myself reading about the writer or director or whoever. I'll often seek out more of their work specifically rather than seeking out any old thing that might be in the same genre or make me feel a similar way — maybe my real interest is in the artist and their ideas. That relationship between the artist and the audience is pretty important, I think.
I'd argue that art is inherently pretty social and that — because there's no artist in a traditional sense — an AI generated story or song or whatever just doesn't really serve the same cultural purpose.
I think that's a valid and common way to approach media, but it's certainly not the only way to do so. I think the proper phrasing is not that art is inherently social, but that it is commonly...
I think that's a valid and common way to approach media, but it's certainly not the only way to do so. I think the proper phrasing is not that art is inherently social, but that it is commonly consumed in a social manner. To say that art is inherently anything is to attempt to define art, and I'm not sure that's a reasonable proposition given the broad range of perfectly valid definitions.
Art as a cultural phenomenon has changed pretty dramatically over the ages. It used to be that art was largely sacred, in one flavor or another, but relatively recent development has seen it veer secular. Definitions of what art is have oscillated back and forth for as long as anyone has been keeping track.
Go to a museum in Italy and you'll quickly realize that much of the defining art of the western canon was created on commission to flatter and elevate some fairly reprehensible people. That is, perhaps, an inherently-social purpose, but not the one I think you were talking about. Yet they remain examples of art, and of fairly important art, not because Michaelangelo or Raphael were trying to communicate anything in particular - but because of the importance of their subjects, perhaps from innovative techniques used in their creation, and most importantly how we collectively decided to treat the art afterwards.
In the same way, I don't believe that asking AI to create an artwork is very much different from popes and doges throwing money at artists to bolster their social standing, at least in terms of the artist's intent and fingerprint upon the work. I'm certain that in many cases the only part of the subject that truly impressed the artist was their pocketbook. In the same way, books have been written to pay rent, or for a wager, or in a drug-fuelled stupor, and they remain literature. Not every piece of art is trying to say something, nor is every artist. Not every person is looking for art that has something to say. We all produce and enjoy art differently.
That's a good point. I think my phrasing there wasn't great, and you're certainly right that commercial art muddies the waters. I think you hit on something that drives at what I was thinking. Art...
That's a good point. I think my phrasing there wasn't great, and you're certainly right that commercial art muddies the waters.
I think you hit on something that drives at what I was thinking. Art has changed drastically over time, but it's always been made by people. Art gets made and viewed within some kind of cultural context, and in order to exist within that, you have to be a person. A better way to phrase it might have been "art is inherently human, and humans are inherently social." Even referring to it as a "cultural phenomenon" (which I think is totally correct) is pretty explicitly tying art to personhood. To me, removing the artist from the equation just kind of puts this stuff in a different category, regardless of the artist's motivations or intent.
Commercial stuff definitely makes this messier. A boring stock photo on a corporate website absolutely technically counts as art, but I feel like slotting in an AI generated image in its place feels a lot more reasonable — it doesn't feel so different to me there. Maybe I'm kind of baking in some idea of "cultural value" here. That's admittedly arbitrary, but I'm not sure it feels so wrong...
In your earlier comment, you bring up the idea that AI programs will probably make poignant and inspiring things — I think you're right, but I don't think it makes those things art. I don't think we can describe art only in the context of the viewer's experience, we also have to consider the artist and the art's place in culture. I just don't think the latter is even relevant for AI art unless we start considering these programs to be people, which is very hard for me to take seriously right now.
I think there's an argument to be made that AI works may be beautiful and inspiring in the same way that a natural vista is - religious arguments aside, nobody "made" nature, but one can still...
I think there's an argument to be made that AI works may be beautiful and inspiring in the same way that a natural vista is - religious arguments aside, nobody "made" nature, but one can still enjoy beauty and vibes from it despite the lack of creative intent. The process of watching AI generation and looking for particularly beautiful outputs does somewhat remind me of sifting through seashells to find one that is especially unique, or finding a tree that grew in a fun way.
The problem arises when the AI outputs are, absent context, fairly hard to distinguish from human-produced content that is very definitely supposed to be art. It's like if someone engineered a tree that grew into random sculpted shapes - those sculptures would not be "art" in the same way as a human-created wooden sculpture despite being more or less identical, but they would be beautiful, and desired, and people would seek out novel or interesting examples of them.
I think the disconnect arises when people try to separate out art from beauty. For a lot of people, beauty and art are more or less equivalent; they want to fill their life or their home with things that are interesting, appealing and novel. For this purpose, AI art, seashells and funny trees probably work just as well as the traditional variety of art. It's only if you're interested in that extra element of human contribution that the origin of the art begins to matter. I think the availability of AI art just exposed a division that wasn't apparent before, between the people that wanted beauty and the people that wanted something more inflected with humanity.
I'm still not sure I wholly agree that AI art is in the same category as natural phenomena, though, because it's in a new category that didn't exist before. People do train trees to grow into various shapes, and it's decidedly on the "art" side of the fence because it was directed by a human. Yet what if someone genetically engineered that sculpture-tree? We "directed" it to grow that way, it's very unlikely that a tree in nature would grow to resemble a chipmunk or the Venus de Milo, so I think it'd be valid to claim credit for that output if one genetically engineered a tree that happened to do that reliably.
Yet something tells me it wouldn't fit that definition of "art" despite that process being monumentally harder than sculpting or training a tree to grow into that form. Anyone could plant the sculpture tree, even if they weren't the ones who did the heavy lifting of designing its genes. If creative effort and intent define art, then surely the work to create the tree counts - but since they're not the ones planting the tree, or pushing the button, that gap strips the end product of its assumed cultural value.
It feels wrong to me, because it doesn't neatly fit the paradigm of natural beauty, nor of human creativity. It's something in the middle, a natural creative force that we can nudge and direct. It makes me sad to see all of the hostility and anger about something so strange and unique.
When I was in college, I worked at a hotel front desk, and an old guy called on the phone to inquire about a room. We had one left, and I told him so. I asked if he wanted to put a card down to reserve it, and he said no - he'd drive right over. By the time he had, however, someone had used our online booking system to grab the room. I explained this to him, and he was instantly enraged. He blew up and started ranting about how he hated the internet, and how it had ruined everything. You could tell this wasn't the first scenario where his lack of facility with tech had cost him something, but instead of trying to adapt to the change he just hated it, and probably repeated that scene several more times throughout the remainder of his life.
I think about that guy a lot recently. Love it or hate it, the internet changed everything because it worked better in a variety of scenarios. It was faster, more effective and more capable than what we had before. People couldn't help but incorporate it into their lives, and those that didn't have faded to the sidelines of society, condemned to second-tier services in a world designed for everyone but them. I worry that in the philosophy and vitriol, a lot of people are setting themselves against this generation's transformative change.
Right. When you watch a Tarantino movie you’re thinking about the man that made it and how unique he is. Granted there is a lot of junk out there. AI generated “reality TV” or porn isn’t...
Right. When you watch a Tarantino movie you’re thinking about the man that made it and how unique he is.
Granted there is a lot of junk out there. AI generated “reality TV” or porn isn’t subverting much artistry.
I thought this video/letter beautifully expressed the feeling one gets from looking at "artwork" that comes out of generative AI models (and what a combo of Stephen Fry reading a letter from Nick...
I thought this video/letter beautifully expressed the feeling one gets from looking at "artwork" that comes out of generative AI models (and what a combo of Stephen Fry reading a letter from Nick Cave <3 ). Even if you didn't at first know that it was AI generated, once you find out it becomes immediately degraded. The human creative struggle is an essential part in giving meaning and value to an artwork that generative AI will never truly replace.
I find myself noticing AI generated images on articles or video thumbnails and immediately thinking lesser of the creative value of whatever content is behind it. Almost associating it directly with spam content. If they put so little effort in creating the image is the rest really worth my time?
I don't see it that way. I respect that you do, I just wish you expressed it in "I" form instead of "your" form. These things are subjective and other views are possible and legitimate. Anyway,...
Even if you didn't at first know that it was AI generated, once you find out it becomes immediately degraded. The human creative struggle is an essential part in giving meaning and value to an artwork that generative AI will never truly replace.
I don't see it that way. I respect that you do, I just wish you expressed it in "I" form instead of "your" form. These things are subjective and other views are possible and legitimate.
Anyway, what I like about the whole AI thing is that it makes me think about questions like the one we are discussing here.
Not all of us are artistically gifted or want to spend the time to create the perfect image for whatever we're imagining. For example, I was recently talking with some friends about wand and broom...
Not all of us are artistically gifted or want to spend the time to create the perfect image for whatever we're imagining. For example, I was recently talking with some friends about wand and broom cowboy wizards. We found the idea hilarious, so I spent some time with DALL-E generating some images that matched what we were imagining.
There was still some refinement involved in fine-tuning the prompt to get images we were happy with (denim jackets, magic lassos, etc). The model even made the cowboy hats look pointy like a wizard cap! Sure, I could've hired some skilled freelancers to try and recreate what my friends and I were imagining, but that would have taken far more time, money, and effort than just playing with an image generator for half-an-hour.
Can’t say I care for the religious argument or putting process above the outcome. The way I think about this is “ghostwriters for everyone.” I think it’s okay-ish as long as you read it over and...
Can’t say I care for the religious argument or putting process above the outcome.
The way I think about this is “ghostwriters for everyone.” I think it’s okay-ish as long as you read it over and you’re happy that it represents you. It should be what you really want to say.
I haven’t tried it, but I doubt I’d use the results without further edits. More likely I’d use it for brainstorming; ChatGPT is good at making lists.
I don't know, I mean we shouldn't all aspire to have ghost writers, should we? I feel as though, having used Chat-GPT a few times as a writer, the effects that it has on me are almost identical to...
I don't know, I mean we shouldn't all aspire to have ghost writers, should we?
I feel as though, having used Chat-GPT a few times as a writer, the effects that it has on me are almost identical to if I were to commit an act of plagiarism. Sure, I can switch the words around from the source I'm taking from some so it sounds like me, and reflects what I want to say, but I still didn't say it. I'd argue that I barely even thought it. I just agreed with it. Put my name on it.
What's absent in that process is the work that goes into doing anything at all. Fine for like, an invitation or an email, sure. But for a creative pursuit taking away the mental work of first figuring out what you want to say, then putting it into your own words, it takes away the entire reason we write in my view. Makes us lazy, don't want go looking for tubers.
I see copying as essential to the creative process, rather than opposed to it. I’m reminded of different ways of making music. Musicians learn a song by playing it and that’s imitating what...
I see copying as essential to the creative process, rather than opposed to it. I’m reminded of different ways of making music. Musicians learn a song by playing it and that’s imitating what someone else did very closely. Sometimes we even memorize it. After you’ve internalized it, you can start playing variations and improvising. You can’t play it unless you’ve learned it.
But it’s legit for artists to build music out of samples too. It’s a different kind of music. There are also DJ’s.
You don’t have to imitate other musicians. You can imitate nature, or machines, taking things that were never meant to be music and finding a musical use for them. You can listen to anything.
Listening is how you judge emotional impact of what you’re making. Having good musical taste comes from listening.
A lot of the problems people are wary of come from skipping or taking shortcuts while editing, publishing things you don’t really stand behind because you didn’t care. If you don’t care enough to listen to the noise you’re publishing, why should anyone else care?
With writing it feels a little different because we generally assume written words ultimately come from people; there aren’t natural processes that generate writing. But machine-generated writing goes back to the printing press (no longer hand-written) and later, businesses started mailing computer-generated form letters. Then there were text-only computer games.
Now there are content farms and bots publishing huge amounts of generated text on the Internet. A lot of this is going to just be noise, just like a lot of the sounds we hear are noise. You do sometimes want to pay attention to noise. Computer-generated text is sometimes worth paying attention to as well, but we’re going to need a lot of filtering to find what we care about.
After we get over the novelty, saying “this is AI generated” isn’t going to be enough to care. There need to be other reasons why it’s worth reading.
I completely agree with your point about "copying" (I wouldn't call what you describe even copying), I'm a very large fan of the statement "everything is a remix." Look up every work that went...
I completely agree with your point about "copying" (I wouldn't call what you describe even copying), I'm a very large fan of the statement "everything is a remix." Look up every work that went into inspiring your favorite Shakespeare play and you can see that the dude was never some anomaly of originality and creativity, but just well read. They just remixed the texts they knew and used them to create their own unique works. It's how humans do, we build off each others knowledge to create new knowledge.
Kind of a jump from Shakespeare, but going with your music analogy Death Grips uses a million samples in their songs, yet when Exmilitary came out I was probably one of the most unique sounds I heard.
The argument could be made that Generative-AI is doing just this, remixing the knowledge that it's been fed to create it's own works and responses. However, I'd still argue that that's not really what's going on here. It reminds me of the question about if Koko the ape could understrand ASL. There was a bunch of unrelated tom-fuckery with that experiment, but the principle question was if Koko could actually use sign language, or it's just a case of monkey-see monkey-do. The consensus today is that it was probably the latter, she was just responding to positive reinforcement when making, or responding to, correct signs. We wanted to believe that she could sign because, well, that's something that we do and we like to think our way of communicating can be replicated in other species, that we don't need to understand them, but teach them to understand us. Really, she just wanted a banana.
Similarly with Chat-GPT, we've fed it information and given it enough positive reinforcement that it knows what do when someone asks it a question. It strings together words like this, it gets bananas. Shakespeare, or MC Ride, didn't do what they did for banana's, but because they made an intelligent thought to combine ideas they were familiar with into new ideas. imitation isn't intelligence, and when a user uses Chat-GPT they're only getting an imitation of what GPT's sources look like, and even then we'll never have access to those sources. It's answers will always lack the crucial step from imitation to remix, which is understanding it's sources to craft a new idea, because it doesn't understand anything except for how to get more banana's. When you use Chat-GPT to write something creative, you're basing what you write off of a second-hand copy, which is too far away from sampling others work to ever really get something original out of it in my eyes.
I think I've run out of ways to incorporate banana's into this semi-response/addendum/clarification-to-my-original-comment... thing. To connect it more directly back to my original comment about comparing using Chat-GPT to plagerism, you don't have the context needed to create something worthwhile when doing it this way. You're robbing yourself of the practice and knowledge it takes to remix ideas into something original, and instead leaning on something that could never do that because it's just not capable of that.
While it is true that human artists is inspired by and remixes other peoples work, we also have a whole life worth of experiences that influence what we create. A writer might have a near death...
While it is true that human artists is inspired by and remixes other peoples work, we also have a whole life worth of experiences that influence what we create. A writer might have a near death experience with almost getting hit by a bus, which inspires them to write something really compelling. So far, the AI models are not there yet, though some would probably argue that it is just a matter of feeding the models with enough data. There are some experiments of training AI models on sight, sound and smell. It becomes a rather philosophical question whether everything that defines a human personality can be reduced down to a string of data. It seems though we are still pretty far off from that level. Yes the technology moves fast, but it isn't guaranteed that the level of acceleration will continue. Sometimes we hit plateaus in technical advancement when the problems are starting to get exponential harder to solve for each small increment of improvement.
First, it's bananas, not banana's. Moving on, Shakespeare, humans in general, do develop skills and perform actions seeking "bananas." An infant sees adults (or older siblings) around themselves...
Shakespeare, or MC Ride, didn't do what they did for banana's, but because they made an intelligent thought to combine ideas they were familiar with into new ideas.
First, it's bananas, not banana's.
Moving on, Shakespeare, humans in general, do develop skills and perform actions seeking "bananas." An infant sees adults (or older siblings) around themselves walking upright, tries it, and is rewarded with positive attention. Learns to talk, and bananas. Gets good grades, bananas. Acts appropriately in ways the parents and friends desire, bananas.
Humanity uses a core, integrated praise and positive reinforcement loop. Instinctively. Someone does something you approve of, like, anything positive, you reward them. By offering them attention, praise, reinforcement. And when someone does things we consider to be less positive, be they unappealing, or ugly, disgusting, or even dangerous, we offer the reverse and show them negativity.
It happens on the individual level, it happens on the societal level. Modern society has even started to recognize this actively, and begun to experiment with using it purposefully (reference how media and online circles are starting to decide to not name certain kinds of criminals, under the theory those bad actors want the attention and immortality of notoriety, negative or otherwise.)
Bananas are inherent to how humans operate. People offer you approval and positivity, and you enjoy that. It makes you feel good. You seek to do it more. Tell a joke that gets a laugh, you share the joke with others. Hey, there's this joke I told, got a laugh, want to hear? Your boss tells you good job, and you think you like that better than being told off, so you seek to shape your work activities to get more "good job" and less "you fuck up".
Artists are no different. Not at all. Very, very, very few artists work solely in isolation, and only for themselves. It's almost unheard of. It's common for artists to be shy about their work, especially as beginners ... but that's just a function of the positive/negative reinforcement cycles. They fear rejection, and thus are reluctant to share.
But they want to. Because, also in them, mixed into (behind/beside, sometimes in front of, which we see as confidence) those fears is a desire to receive the approval. To be told, "wow, this is great, I love it, super job, very interesting."
It's one of the most common stumbling blocks of many new artists. Learning to take rejection. Beginning writers, for example, are often very bad at this. They reject the rejections, lash out, take it as personal insult, and refuse to accept it was their work that caused the less than positive response. The person they shared it with must be wrong, uncouth, uncultured, unintelligent, or otherwise unworthy or not ready to have been offered the chance to read the beginner's writing.
Some newbie writers get past this, and learn that rejection is just part of the process. Some even learn to separate the responses from their own views and beliefs about their writing.
My point is, if the objection to AGI is that it's been built purposefully on a positive/negative reinforcement loop ... that's no different than humans. When you write something "weird", people don't respond. At least, not positively. When you write something "good", they reward you with positivity. Bad, and they punish you with negativity.
Sometimes, a rare writer can find a very particular and welcome "kind of weird" and find success, but again that's just the positive/negative loop in action. They tried weird, found it somehow worked, and do it more. Meanwhile, a hundred others try weird, get lambasted, and quickly stop as they pivot to something else. Pivoting in search of, wait for it, positive response.
Humanity is all about the bananas. Since the day we crawled out of the ooze. Whichever proto-human form that did that did so because of a positive response. "Gee, this ooze isn't doing it for me. What if I crawled ... oh yes, much better. I like it out here out of the ooze. This is more positive."
And so the cycle has gone. Reaching for the next banana.
Seems like the movie “Click”. I think we’ll see a major division between those who can enjoy working hard on a mental task and those that have never known the pleasure.
Seems like the movie “Click”. I think we’ll see a major division between those who can enjoy working hard on a mental task and those that have never known the pressure pleasure.
For me, it's like a brainstorming buddy. When I get blocked on how to phrase an idea or come up with some points to discuss, I can get some general tips and advice. It's not like I lift the...
For me, it's like a brainstorming buddy. When I get blocked on how to phrase an idea or come up with some points to discuss, I can get some general tips and advice. It's not like I lift the responses wholesale insomuch as it helps me chat through some ideas I already have rolling around in my head or at least get the writing juices flowing.
#offtopic In this context, I don't think Nick is referencing the biblical creation story as a literal example. I think he's referencing it in his letter as a way to use a very widely-known story...
Can’t say I care for the religious argument or putting process above the outcome.
#offtopic
In this context, I don't think Nick is referencing the biblical creation story as a literal example. I think he's referencing it in his letter as a way to use a very widely-known story about creation to make his point. Whether or not the reader/listener actually believes in the biblical creation story is actually immaterial to the point he's making.
But if you’re not religious, the natural world wasn’t created, it just is. Its meaning doesn’t come from anyone’s effort. So the analogy falls apart. We can appreciate nature without thinking,...
But if you’re not religious, the natural world wasn’t created, it just is. Its meaning doesn’t come from anyone’s effort. So the analogy falls apart. We can appreciate nature without thinking, “someone put a lot of work into that.”
We could appreciate generated text in a similar way, looking for bits we find interesting, like finding a nice shell on the beach, or an interesting rock. Maybe you could build a wall or some kind of artwork out of rocks?
The analogy to nature doesn’t really work either, though. AI chatbots are interesting because they’re based on huge amounts of text that actually was written by people. There is randomness, too, but purely random text generation isn’t interesting.
Maybe a better analogy would be building new structures out of stone found in ruins.
I understand your point, I think I didn't explain mine quite right. I think the creation story was referenced as a work of fiction. You could replace God with Odin or Viracocha and make the same...
I understand your point, I think I didn't explain mine quite right. I think the creation story was referenced as a work of fiction. You could replace God with Odin or Viracocha and make the same point as long as your target audience understood the story you are referencing.
Further off topic - I think my opinion here is being swayed by the fact that Stephen Fry read it. Skimming Wikipedia I better understand why Nick Cave would make that reference, as his work contains a lot of religious references.
I have this perception that the British (though Nick is Australian) has/makes references to biblical stories without the same cultural overtones that it would have if an American did. In the US if someone makes a biblical reference I will likely immediately label them as conservative unless I already know them. I don't have the same reflex with people from across the pond.
Yes, understood, and I agree. Biblical references used to be common in the US too, and reading older books you will see more of them. It’s common cultural background.
Yes, understood, and I agree. Biblical references used to be common in the US too, and reading older books you will see more of them. It’s common cultural background.
I'm a writer. I think it's kind of disappointing so many people need an AI to be able to string words together. But, at the same time, I am a firm believer in how fucking hard it is to write well....
I'm a writer. I think it's kind of disappointing so many people need an AI to be able to string words together. But, at the same time, I am a firm believer in how fucking hard it is to write well. Because it is.
Of all the art in the world, I've found writing is the one casual people will tend to assume is the easiest. Then they sit down to do it one weekend and find out "holy shit, this is way fucking harder than I thought."
Because, what usually happens, they assume over the weekend they'll put their "great idea" to paper, be done by Monday, and by Friday have become a sensation as everyone praises them. The praise spreads from their immediate friends and family, talent agents knock on their door, contracts are signed, and by the end of the month they're a global phenomenon.
But by Monday, what's actually happened 99.9% of the time is they've hardly managed any pages, and what is there sucks. They see their friends and family struggling to get through it, searching desperately for something kind and inoffensive to tell the wanna-be writer. Which is difficult because, again, writing is hard!
Other kinds of art doesn't tend to get assumed to be easy. Few people assume drawing or painting, for example, is "easy and quick." They might not agree it's Mount Everest, but they don't tend to think "I'll just learn drawing this weekend" like they do with writing. They tend to recognize that graphical arts, for example, will take a time investment. Which is true for any skill, any skill. Including creative ones.
I'm not anti-AGI. I do think that idea isn't enough, which is another common assumption by casuals. All AGI gives right now is the ability to shape an idea from aether to some kind of form on the page. How good that form might be is a separate discussion (and however good or bad it is now, it will inevitably improve over time).
We're moving toward a future where casuals will have access to a holistic AGI system. One where they can feed it their ideas, and it'll give them back stories, tv shows, paintings, movies, whatever. A place where you can say "what if the Terminator was the aliens from Predator instead of robots, and they're traveling in time to find the most dangerous humans and honorably hunt them; go." And get back an actual product you can read/watch/listen to/whatever, instead of just having disjointed "ideas" in your head.
Right now, it takes an actual artist to do that. Often multiple artists (for something like a show or movie). To take an idea and put it in physical form. Which takes work, which takes time and effort (lots of it) to learn, and then do.
AGI is building toward a place where, at the very least, the "workshop version" of ideas can be put in physical form. And, again at the very least, those versions can certainly be given to actual artists who've learned their craft and be polished further.
Genies don't go back in the bottle. AI isn't going away. It's too interesting of an idea. The people who've had it, and who are working on it, have already gotten the positive reinforcements. Too many people who've been exposed to it have already had positive reactions to it.
Sure I think writing is worth learning, but most don't, because most don't bother to learn it. And having an AGI that can take prompts like "write me a thank you letter" or "write me a story about ghosts going to school with fish" is very interesting to a lot of these folks who haven't bothered.
I've started dabbling with Stable Diffusion. Why? Because I write, but I don't do graphics. So it's fun for me to give it prompts (and I'm finding the prompt engine isn't even remotely as good at words as I am, which is frustrating, but I expect it's going to improve as time goes by) and get back pictures. I'm not good at pictures. I'm good at describing pictures, but not actual pictures.
But aside from that, my main purpose was the thought that instead of having to deal with a graphical artist (something I really do hate doing, because we don't speak the same language and they're often very difficult and unpleasant about it) for a book cover, I could maybe get imagery from SD. I'm poking prompts and fiddling with learning how to get images back with the thought of being able to take these into my covers work flow.
Prior to this I've managed to teach myself some of the rudiments of graphical design and layout, and produced some decent covers, but the stumbling block has been finding images to use. They're not free, you see. Even stock images are expensive for what I need sometimes. Custom images from a graphics artist can run hundreds of dollars.
If I can get images out of SD, that's something I'm extremely interested in. Extremely. It's not remotely as simple as telling Stable Diffusion "give me a book cover with X and Y on it." But I can get images. And I can take those images into GIMP and do what I've already been learning how to do; massage and alter and fiddle with them to make them work with my cover design. Adding elements, combining them, layering on graphics and so on.
Which saves me from having to dig through online for something I can make work. I can just let SD run, flip through what it's kicking out, and when inspiration strikes take that one (or those ones) and see what I can come up with.
I don't care about not owning the cover, which I won't if I source AI images. If I manage to put out a cover that resonates so much people take it, fine. I care about the story, not the cover. That cover is just there to get them to look at the story. And if that cover won't cost me hundreds of dollars, I consider that a win. I think sooner or later I'll get it working, but right now it's still a process I'm learning and fiddling with.
I'm sure an actual graphical artist can do it better, and certainly faster, than I can. Kind of how I can work with a text AGI to get a better story, and polish it better and faster, than they could. But the genie definitely isn't going back in the bottle either way.
https://www.theredhandfiles.com/chatgpt-making-things-faster-and-easier/ Stephen Fry is an incredible orator / narrator though, and one of my favorite people in the world. So I would highly...
Stephen Fry is an incredible orator / narrator though, and one of my favorite people in the world. So I would highly recommend still watching the video, but perhaps just reading while you listen.
A friend of mine got me a job as a software engineer at an AI copywriting company in the 18 months before the public release of ChatGPT. We started by having GPT-3 (3.0) write small snippets of ad text and product descriptions, but as time went on it was clear the next-big-thing would be long form AI writing. I had a lot of fun building the app. It was my first time building a website from the ground up for a successful business. But when my friend asked me if I'd ever use it to write something like a blog post I was adamant that I never would. Honestly, what's the point? If I'm writing something for people to read I'm writing it because I know what I want to say, and the act of writing refines my thoughts until I've written exactly what needs to be read.
Where does an AI fit in here? I could definitely use an AI editor or reviewer. I could get writer's block and need just the right phrase to get un-stuck. Or maybe I realized I've covered too many topics and could use suggestions for which ones to cut. But the AI doesn't have ideas of its own, it's just an accelerator for my keyboard. My friend never really understood what I meant, which is shocking as he is the one with a degree in creative writing and I'm the one with a CS degree. The only explanation I can think of is the old quote:
My friend saw this startup as his path to lifetime financial security. He'd been given a generous equity package and if it did really well he'd have many millions of dollars. He was even at the point where he'd say things like "Hey if this works out I'll personally buy you a house."
But it wasn't meant to be. OpenAI used the startup for feedback on GPT-3's outputs (we had weekly reviews with an OpenAI rep). They used our inputs and outputs to train 3.5 and 4.0 and then released ChatGPT which more-or-less obsoleted the whole product.
You've succinctly captured my feelings about using GEN AI for ideation, it really can only come up with the most obvious things, so while it can seem convincing to get the ball rolling, truly novel insight remains human for now.
It's an energy activation thing, it's easy to type in what you want and have AI give you back obvious things, it takes hard work and thinking to get something novel.
But the risk I fight against is doing that hard thinking and coming up with the same obvious things the AI did, since there's no guarantee there will be any spark of inspiration.
So I guess all we can do is avoid being lazy, and be more brave.
Reminds me of Pixar's rules for writing: "Discount the First Idea That Comes to Mind. And the 2nd, and the 3rd and 4th and 5th. Get the Obvious Ones Out of the Way. Surprise Yourself."
AI could give you 100 ideas, to eliminate obvious ideas to help you get to the novel ideas.
Even if it did have novel insight, what am I going to do? Stop thinking for myself? I’d be better off dead.
I'm confused about the start up part of your friend's journey. Did openAI ask you to write this app? Was it paying for these input output or was it paying for your time on a monthly basis? How was it going to make your friend mega millions, vs they only ever wanted human data wranglers to train their LLM until its output is as good as your app's? Does your friend feel it was a fair exchange for his time?
The app we worked on was bringing in a few million in annual revenue on its own. GPT-3 was a closed beta. Helping OpenAI was a condition of beta access. And of course, if you’re friendly with OpenAI they’ll give you super duper early access to the newest models. If you’re trying to be a startup riding on their coattails they’re the ones playing kingmaker.
oh, wow.
I guess things would have played out completely differently if (1) OpenAI kept their LLM to enterprise level access that companies can license for access (kinda like AWS?) or if (2) it took them long enough that your friend's startup can be sold for mega millions.
The way I understand it is like, if OpenAI had an orchard full of newly hybridized berries, and your friend's start up was a fleet of trucks bringing new berries to market at a rate like never before. And then they invented "teleport u-pick" open to everyone, including people who just wanted one berry a month. Kinda like that?
Yeah pretty much. It’s funny because our investors would always ask the CEO if we were safe against OpenAI cannibalizing the business. He was adamant that OpenAI would never want to handle the “last mile” themselves.
I think it's very natural for authors, musicians and actors to focus on the creative act that ChatGPT supplants when it's used, but I'd argue that they're only one half of the puzzle. One of the things I've realized in putting my own writing out there is that people read it in a very different context than I'm writing it, and my intent and goals in putting together a specific passage are only a portion of how they ultimately read and understand it. We see what we see and read what we read based not only on the work itself, but on the lifetime of experience and context that we supply when we read, watch or listen to that work.
In a comment above, the OP says that "once you find out [the work is AI] it becomes immediately degraded." And that may be true, for many people, but I find it odd to think that we would discard the meaning and ideas that we create, internally, upon seeing something, because we did not like the process behind them. Many artists have no particular intent or struggle in crafting their work, and it doesn't prevent people from deriving enjoyment from it.
The viewer is the final part of the creative process and still has heavy lifting to do; I don't subscribe to the notion that the lack of human involvement somehow handicaps reader enjoyment. The lack of facility in current-generation models is what does that. The same words, written by a human, would strike the reader no differently in a blind comparison.
We are fast-approaching a world where AI images and writing are going to effortlessly mimic the best of human creation. If you don't think it's coming, then you haven't been keeping up with the pace of acceleration - in a few short years we have gone from amateurish ramblings and abstract art to something very close to indistinguishable from human works. There will be something created that is poignant, or inspiring, or thought-provoking to you despite the utter lack of human involvement in its creation.
At that point you can either reject the feeling as artificial, or embrace it and recognize that the lack of an artist doesn't devalue what you feel - it just makes it wholly yours.
It does feel like the whole discussion of defining real/fake art is reductive and reactionary. And when most career artists are successful because their works are commodified, it sort of makes the argument ring hollow.
I do agree with your idea of effort not being a requirement of art. My favorite image is the Pale Blue Dot. Compositionally: light streaks on a black background with a single white pixle right of center. Contextually:It's a photo of earth taken from Voyager 1, about six billion kilometers away.
We could debate for hours the semantics of this image. Is it even "real" art? Who should be credited with so many people involved in its creation? Should images like this be copyright protected? I don't think any of it matters. I look at the print of it every day and it holds a wealth of inspiration and consideration.
I think the more important discussion around AI art should be in its consumption. Like you said, art evokes an emotional response, so what happens if people could fine tune exactly what they want to see and feel.
Thinking about modern social media and internet usage, people are able to silo themselves to an antisocial degree and one of the few touch points are the art we share. Movies, music, games, TV. It's falling away with how algorithmic content has become and I don't know how to feel about people only consuming content designed and made just for them.
Combined with the explosion of profit driven social substitutes like AI partners, friends and therapy; fake people spamming content and very real social engineering risks; it feels like people are missing the Forrest for the trees.
Often when I read a great book or watch a great movie (or engage with any kind of art really), I find myself reading about the writer or director or whoever. I'll often seek out more of their work specifically rather than seeking out any old thing that might be in the same genre or make me feel a similar way — maybe my real interest is in the artist and their ideas. That relationship between the artist and the audience is pretty important, I think.
I'd argue that art is inherently pretty social and that — because there's no artist in a traditional sense — an AI generated story or song or whatever just doesn't really serve the same cultural purpose.
I think that's a valid and common way to approach media, but it's certainly not the only way to do so. I think the proper phrasing is not that art is inherently social, but that it is commonly consumed in a social manner. To say that art is inherently anything is to attempt to define art, and I'm not sure that's a reasonable proposition given the broad range of perfectly valid definitions.
Art as a cultural phenomenon has changed pretty dramatically over the ages. It used to be that art was largely sacred, in one flavor or another, but relatively recent development has seen it veer secular. Definitions of what art is have oscillated back and forth for as long as anyone has been keeping track.
Go to a museum in Italy and you'll quickly realize that much of the defining art of the western canon was created on commission to flatter and elevate some fairly reprehensible people. That is, perhaps, an inherently-social purpose, but not the one I think you were talking about. Yet they remain examples of art, and of fairly important art, not because Michaelangelo or Raphael were trying to communicate anything in particular - but because of the importance of their subjects, perhaps from innovative techniques used in their creation, and most importantly how we collectively decided to treat the art afterwards.
In the same way, I don't believe that asking AI to create an artwork is very much different from popes and doges throwing money at artists to bolster their social standing, at least in terms of the artist's intent and fingerprint upon the work. I'm certain that in many cases the only part of the subject that truly impressed the artist was their pocketbook. In the same way, books have been written to pay rent, or for a wager, or in a drug-fuelled stupor, and they remain literature. Not every piece of art is trying to say something, nor is every artist. Not every person is looking for art that has something to say. We all produce and enjoy art differently.
That's a good point. I think my phrasing there wasn't great, and you're certainly right that commercial art muddies the waters.
I think you hit on something that drives at what I was thinking. Art has changed drastically over time, but it's always been made by people. Art gets made and viewed within some kind of cultural context, and in order to exist within that, you have to be a person. A better way to phrase it might have been "art is inherently human, and humans are inherently social." Even referring to it as a "cultural phenomenon" (which I think is totally correct) is pretty explicitly tying art to personhood. To me, removing the artist from the equation just kind of puts this stuff in a different category, regardless of the artist's motivations or intent.
Commercial stuff definitely makes this messier. A boring stock photo on a corporate website absolutely technically counts as art, but I feel like slotting in an AI generated image in its place feels a lot more reasonable — it doesn't feel so different to me there. Maybe I'm kind of baking in some idea of "cultural value" here. That's admittedly arbitrary, but I'm not sure it feels so wrong...
In your earlier comment, you bring up the idea that AI programs will probably make poignant and inspiring things — I think you're right, but I don't think it makes those things art. I don't think we can describe art only in the context of the viewer's experience, we also have to consider the artist and the art's place in culture. I just don't think the latter is even relevant for AI art unless we start considering these programs to be people, which is very hard for me to take seriously right now.
I think there's an argument to be made that AI works may be beautiful and inspiring in the same way that a natural vista is - religious arguments aside, nobody "made" nature, but one can still enjoy beauty and vibes from it despite the lack of creative intent. The process of watching AI generation and looking for particularly beautiful outputs does somewhat remind me of sifting through seashells to find one that is especially unique, or finding a tree that grew in a fun way.
The problem arises when the AI outputs are, absent context, fairly hard to distinguish from human-produced content that is very definitely supposed to be art. It's like if someone engineered a tree that grew into random sculpted shapes - those sculptures would not be "art" in the same way as a human-created wooden sculpture despite being more or less identical, but they would be beautiful, and desired, and people would seek out novel or interesting examples of them.
I think the disconnect arises when people try to separate out art from beauty. For a lot of people, beauty and art are more or less equivalent; they want to fill their life or their home with things that are interesting, appealing and novel. For this purpose, AI art, seashells and funny trees probably work just as well as the traditional variety of art. It's only if you're interested in that extra element of human contribution that the origin of the art begins to matter. I think the availability of AI art just exposed a division that wasn't apparent before, between the people that wanted beauty and the people that wanted something more inflected with humanity.
I'm still not sure I wholly agree that AI art is in the same category as natural phenomena, though, because it's in a new category that didn't exist before. People do train trees to grow into various shapes, and it's decidedly on the "art" side of the fence because it was directed by a human. Yet what if someone genetically engineered that sculpture-tree? We "directed" it to grow that way, it's very unlikely that a tree in nature would grow to resemble a chipmunk or the Venus de Milo, so I think it'd be valid to claim credit for that output if one genetically engineered a tree that happened to do that reliably.
Yet something tells me it wouldn't fit that definition of "art" despite that process being monumentally harder than sculpting or training a tree to grow into that form. Anyone could plant the sculpture tree, even if they weren't the ones who did the heavy lifting of designing its genes. If creative effort and intent define art, then surely the work to create the tree counts - but since they're not the ones planting the tree, or pushing the button, that gap strips the end product of its assumed cultural value.
It feels wrong to me, because it doesn't neatly fit the paradigm of natural beauty, nor of human creativity. It's something in the middle, a natural creative force that we can nudge and direct. It makes me sad to see all of the hostility and anger about something so strange and unique.
When I was in college, I worked at a hotel front desk, and an old guy called on the phone to inquire about a room. We had one left, and I told him so. I asked if he wanted to put a card down to reserve it, and he said no - he'd drive right over. By the time he had, however, someone had used our online booking system to grab the room. I explained this to him, and he was instantly enraged. He blew up and started ranting about how he hated the internet, and how it had ruined everything. You could tell this wasn't the first scenario where his lack of facility with tech had cost him something, but instead of trying to adapt to the change he just hated it, and probably repeated that scene several more times throughout the remainder of his life.
I think about that guy a lot recently. Love it or hate it, the internet changed everything because it worked better in a variety of scenarios. It was faster, more effective and more capable than what we had before. People couldn't help but incorporate it into their lives, and those that didn't have faded to the sidelines of society, condemned to second-tier services in a world designed for everyone but them. I worry that in the philosophy and vitriol, a lot of people are setting themselves against this generation's transformative change.
Right. When you watch a Tarantino movie you’re thinking about the man that made it and how unique he is.
Granted there is a lot of junk out there. AI generated “reality TV” or porn isn’t subverting much artistry.
I thought this video/letter beautifully expressed the feeling one gets from looking at "artwork" that comes out of generative AI models (and what a combo of Stephen Fry reading a letter from Nick Cave <3 ). Even if you didn't at first know that it was AI generated, once you find out it becomes immediately degraded. The human creative struggle is an essential part in giving meaning and value to an artwork that generative AI will never truly replace.
I find myself noticing AI generated images on articles or video thumbnails and immediately thinking lesser of the creative value of whatever content is behind it. Almost associating it directly with spam content. If they put so little effort in creating the image is the rest really worth my time?
I don't see it that way. I respect that you do, I just wish you expressed it in "I" form instead of "your" form. These things are subjective and other views are possible and legitimate.
Anyway, what I like about the whole AI thing is that it makes me think about questions like the one we are discussing here.
Not all of us are artistically gifted or want to spend the time to create the perfect image for whatever we're imagining. For example, I was recently talking with some friends about wand and broom cowboy wizards. We found the idea hilarious, so I spent some time with DALL-E generating some images that matched what we were imagining.
There was still some refinement involved in fine-tuning the prompt to get images we were happy with (denim jackets, magic lassos, etc). The model even made the cowboy hats look pointy like a wizard cap! Sure, I could've hired some skilled freelancers to try and recreate what my friends and I were imagining, but that would have taken far more time, money, and effort than just playing with an image generator for half-an-hour.
Can’t say I care for the religious argument or putting process above the outcome.
The way I think about this is “ghostwriters for everyone.” I think it’s okay-ish as long as you read it over and you’re happy that it represents you. It should be what you really want to say.
I haven’t tried it, but I doubt I’d use the results without further edits. More likely I’d use it for brainstorming; ChatGPT is good at making lists.
I don't know, I mean we shouldn't all aspire to have ghost writers, should we?
I feel as though, having used Chat-GPT a few times as a writer, the effects that it has on me are almost identical to if I were to commit an act of plagiarism. Sure, I can switch the words around from the source I'm taking from some so it sounds like me, and reflects what I want to say, but I still didn't say it. I'd argue that I barely even thought it. I just agreed with it. Put my name on it.
What's absent in that process is the work that goes into doing anything at all. Fine for like, an invitation or an email, sure. But for a creative pursuit taking away the mental work of first figuring out what you want to say, then putting it into your own words, it takes away the entire reason we write in my view. Makes us lazy, don't want go looking for tubers.
I see copying as essential to the creative process, rather than opposed to it. I’m reminded of different ways of making music. Musicians learn a song by playing it and that’s imitating what someone else did very closely. Sometimes we even memorize it. After you’ve internalized it, you can start playing variations and improvising. You can’t play it unless you’ve learned it.
But it’s legit for artists to build music out of samples too. It’s a different kind of music. There are also DJ’s.
You don’t have to imitate other musicians. You can imitate nature, or machines, taking things that were never meant to be music and finding a musical use for them. You can listen to anything.
Listening is how you judge emotional impact of what you’re making. Having good musical taste comes from listening.
A lot of the problems people are wary of come from skipping or taking shortcuts while editing, publishing things you don’t really stand behind because you didn’t care. If you don’t care enough to listen to the noise you’re publishing, why should anyone else care?
With writing it feels a little different because we generally assume written words ultimately come from people; there aren’t natural processes that generate writing. But machine-generated writing goes back to the printing press (no longer hand-written) and later, businesses started mailing computer-generated form letters. Then there were text-only computer games.
Now there are content farms and bots publishing huge amounts of generated text on the Internet. A lot of this is going to just be noise, just like a lot of the sounds we hear are noise. You do sometimes want to pay attention to noise. Computer-generated text is sometimes worth paying attention to as well, but we’re going to need a lot of filtering to find what we care about.
After we get over the novelty, saying “this is AI generated” isn’t going to be enough to care. There need to be other reasons why it’s worth reading.
I completely agree with your point about "copying" (I wouldn't call what you describe even copying), I'm a very large fan of the statement "everything is a remix." Look up every work that went into inspiring your favorite Shakespeare play and you can see that the dude was never some anomaly of originality and creativity, but just well read. They just remixed the texts they knew and used them to create their own unique works. It's how humans do, we build off each others knowledge to create new knowledge.
Kind of a jump from Shakespeare, but going with your music analogy Death Grips uses a million samples in their songs, yet when Exmilitary came out I was probably one of the most unique sounds I heard.
The argument could be made that Generative-AI is doing just this, remixing the knowledge that it's been fed to create it's own works and responses. However, I'd still argue that that's not really what's going on here. It reminds me of the question about if Koko the ape could understrand ASL. There was a bunch of unrelated tom-fuckery with that experiment, but the principle question was if Koko could actually use sign language, or it's just a case of monkey-see monkey-do. The consensus today is that it was probably the latter, she was just responding to positive reinforcement when making, or responding to, correct signs. We wanted to believe that she could sign because, well, that's something that we do and we like to think our way of communicating can be replicated in other species, that we don't need to understand them, but teach them to understand us. Really, she just wanted a banana.
Similarly with Chat-GPT, we've fed it information and given it enough positive reinforcement that it knows what do when someone asks it a question. It strings together words like this, it gets bananas. Shakespeare, or MC Ride, didn't do what they did for banana's, but because they made an intelligent thought to combine ideas they were familiar with into new ideas. imitation isn't intelligence, and when a user uses Chat-GPT they're only getting an imitation of what GPT's sources look like, and even then we'll never have access to those sources. It's answers will always lack the crucial step from imitation to remix, which is understanding it's sources to craft a new idea, because it doesn't understand anything except for how to get more banana's. When you use Chat-GPT to write something creative, you're basing what you write off of a second-hand copy, which is too far away from sampling others work to ever really get something original out of it in my eyes.
I think I've run out of ways to incorporate banana's into this semi-response/addendum/clarification-to-my-original-comment... thing. To connect it more directly back to my original comment about comparing using Chat-GPT to plagerism, you don't have the context needed to create something worthwhile when doing it this way. You're robbing yourself of the practice and knowledge it takes to remix ideas into something original, and instead leaning on something that could never do that because it's just not capable of that.
While it is true that human artists is inspired by and remixes other peoples work, we also have a whole life worth of experiences that influence what we create. A writer might have a near death experience with almost getting hit by a bus, which inspires them to write something really compelling. So far, the AI models are not there yet, though some would probably argue that it is just a matter of feeding the models with enough data. There are some experiments of training AI models on sight, sound and smell. It becomes a rather philosophical question whether everything that defines a human personality can be reduced down to a string of data. It seems though we are still pretty far off from that level. Yes the technology moves fast, but it isn't guaranteed that the level of acceleration will continue. Sometimes we hit plateaus in technical advancement when the problems are starting to get exponential harder to solve for each small increment of improvement.
First, it's bananas, not banana's.
Moving on, Shakespeare, humans in general, do develop skills and perform actions seeking "bananas." An infant sees adults (or older siblings) around themselves walking upright, tries it, and is rewarded with positive attention. Learns to talk, and bananas. Gets good grades, bananas. Acts appropriately in ways the parents and friends desire, bananas.
Humanity uses a core, integrated praise and positive reinforcement loop. Instinctively. Someone does something you approve of, like, anything positive, you reward them. By offering them attention, praise, reinforcement. And when someone does things we consider to be less positive, be they unappealing, or ugly, disgusting, or even dangerous, we offer the reverse and show them negativity.
It happens on the individual level, it happens on the societal level. Modern society has even started to recognize this actively, and begun to experiment with using it purposefully (reference how media and online circles are starting to decide to not name certain kinds of criminals, under the theory those bad actors want the attention and immortality of notoriety, negative or otherwise.)
Bananas are inherent to how humans operate. People offer you approval and positivity, and you enjoy that. It makes you feel good. You seek to do it more. Tell a joke that gets a laugh, you share the joke with others. Hey, there's this joke I told, got a laugh, want to hear? Your boss tells you good job, and you think you like that better than being told off, so you seek to shape your work activities to get more "good job" and less "you fuck up".
Artists are no different. Not at all. Very, very, very few artists work solely in isolation, and only for themselves. It's almost unheard of. It's common for artists to be shy about their work, especially as beginners ... but that's just a function of the positive/negative reinforcement cycles. They fear rejection, and thus are reluctant to share.
But they want to. Because, also in them, mixed into (behind/beside, sometimes in front of, which we see as confidence) those fears is a desire to receive the approval. To be told, "wow, this is great, I love it, super job, very interesting."
It's one of the most common stumbling blocks of many new artists. Learning to take rejection. Beginning writers, for example, are often very bad at this. They reject the rejections, lash out, take it as personal insult, and refuse to accept it was their work that caused the less than positive response. The person they shared it with must be wrong, uncouth, uncultured, unintelligent, or otherwise unworthy or not ready to have been offered the chance to read the beginner's writing.
Some newbie writers get past this, and learn that rejection is just part of the process. Some even learn to separate the responses from their own views and beliefs about their writing.
My point is, if the objection to AGI is that it's been built purposefully on a positive/negative reinforcement loop ... that's no different than humans. When you write something "weird", people don't respond. At least, not positively. When you write something "good", they reward you with positivity. Bad, and they punish you with negativity.
Sometimes, a rare writer can find a very particular and welcome "kind of weird" and find success, but again that's just the positive/negative loop in action. They tried weird, found it somehow worked, and do it more. Meanwhile, a hundred others try weird, get lambasted, and quickly stop as they pivot to something else. Pivoting in search of, wait for it, positive response.
Humanity is all about the bananas. Since the day we crawled out of the ooze. Whichever proto-human form that did that did so because of a positive response. "Gee, this ooze isn't doing it for me. What if I crawled ... oh yes, much better. I like it out here out of the ooze. This is more positive."
And so the cycle has gone. Reaching for the next banana.
Seems like the movie “Click”. I think we’ll see a major division between those who can enjoy working hard on a mental task and those that have never known the
pressurepleasure.For me, it's like a brainstorming buddy. When I get blocked on how to phrase an idea or come up with some points to discuss, I can get some general tips and advice. It's not like I lift the responses wholesale insomuch as it helps me chat through some ideas I already have rolling around in my head or at least get the writing juices flowing.
#offtopic
In this context, I don't think Nick is referencing the biblical creation story as a literal example. I think he's referencing it in his letter as a way to use a very widely-known story about creation to make his point. Whether or not the reader/listener actually believes in the biblical creation story is actually immaterial to the point he's making.
But if you’re not religious, the natural world wasn’t created, it just is. Its meaning doesn’t come from anyone’s effort. So the analogy falls apart. We can appreciate nature without thinking, “someone put a lot of work into that.”
We could appreciate generated text in a similar way, looking for bits we find interesting, like finding a nice shell on the beach, or an interesting rock. Maybe you could build a wall or some kind of artwork out of rocks?
The analogy to nature doesn’t really work either, though. AI chatbots are interesting because they’re based on huge amounts of text that actually was written by people. There is randomness, too, but purely random text generation isn’t interesting.
Maybe a better analogy would be building new structures out of stone found in ruins.
I understand your point, I think I didn't explain mine quite right. I think the creation story was referenced as a work of fiction. You could replace God with Odin or Viracocha and make the same point as long as your target audience understood the story you are referencing.
Further off topic - I think my opinion here is being swayed by the fact that Stephen Fry read it. Skimming Wikipedia I better understand why Nick Cave would make that reference, as his work contains a lot of religious references.
I have this perception that the British (though Nick is Australian) has/makes references to biblical stories without the same cultural overtones that it would have if an American did. In the US if someone makes a biblical reference I will likely immediately label them as conservative unless I already know them. I don't have the same reflex with people from across the pond.
Yes, understood, and I agree. Biblical references used to be common in the US too, and reading older books you will see more of them. It’s common cultural background.
I'm a writer. I think it's kind of disappointing so many people need an AI to be able to string words together. But, at the same time, I am a firm believer in how fucking hard it is to write well. Because it is.
Of all the art in the world, I've found writing is the one casual people will tend to assume is the easiest. Then they sit down to do it one weekend and find out "holy shit, this is way fucking harder than I thought."
Because, what usually happens, they assume over the weekend they'll put their "great idea" to paper, be done by Monday, and by Friday have become a sensation as everyone praises them. The praise spreads from their immediate friends and family, talent agents knock on their door, contracts are signed, and by the end of the month they're a global phenomenon.
But by Monday, what's actually happened 99.9% of the time is they've hardly managed any pages, and what is there sucks. They see their friends and family struggling to get through it, searching desperately for something kind and inoffensive to tell the wanna-be writer. Which is difficult because, again, writing is hard!
Other kinds of art doesn't tend to get assumed to be easy. Few people assume drawing or painting, for example, is "easy and quick." They might not agree it's Mount Everest, but they don't tend to think "I'll just learn drawing this weekend" like they do with writing. They tend to recognize that graphical arts, for example, will take a time investment. Which is true for any skill, any skill. Including creative ones.
I'm not anti-AGI. I do think that idea isn't enough, which is another common assumption by casuals. All AGI gives right now is the ability to shape an idea from aether to some kind of form on the page. How good that form might be is a separate discussion (and however good or bad it is now, it will inevitably improve over time).
We're moving toward a future where casuals will have access to a holistic AGI system. One where they can feed it their ideas, and it'll give them back stories, tv shows, paintings, movies, whatever. A place where you can say "what if the Terminator was the aliens from Predator instead of robots, and they're traveling in time to find the most dangerous humans and honorably hunt them; go." And get back an actual product you can read/watch/listen to/whatever, instead of just having disjointed "ideas" in your head.
Right now, it takes an actual artist to do that. Often multiple artists (for something like a show or movie). To take an idea and put it in physical form. Which takes work, which takes time and effort (lots of it) to learn, and then do.
AGI is building toward a place where, at the very least, the "workshop version" of ideas can be put in physical form. And, again at the very least, those versions can certainly be given to actual artists who've learned their craft and be polished further.
Genies don't go back in the bottle. AI isn't going away. It's too interesting of an idea. The people who've had it, and who are working on it, have already gotten the positive reinforcements. Too many people who've been exposed to it have already had positive reactions to it.
Sure I think writing is worth learning, but most don't, because most don't bother to learn it. And having an AGI that can take prompts like "write me a thank you letter" or "write me a story about ghosts going to school with fish" is very interesting to a lot of these folks who haven't bothered.
I've started dabbling with Stable Diffusion. Why? Because I write, but I don't do graphics. So it's fun for me to give it prompts (and I'm finding the prompt engine isn't even remotely as good at words as I am, which is frustrating, but I expect it's going to improve as time goes by) and get back pictures. I'm not good at pictures. I'm good at describing pictures, but not actual pictures.
But aside from that, my main purpose was the thought that instead of having to deal with a graphical artist (something I really do hate doing, because we don't speak the same language and they're often very difficult and unpleasant about it) for a book cover, I could maybe get imagery from SD. I'm poking prompts and fiddling with learning how to get images back with the thought of being able to take these into my covers work flow.
Prior to this I've managed to teach myself some of the rudiments of graphical design and layout, and produced some decent covers, but the stumbling block has been finding images to use. They're not free, you see. Even stock images are expensive for what I need sometimes. Custom images from a graphics artist can run hundreds of dollars.
If I can get images out of SD, that's something I'm extremely interested in. Extremely. It's not remotely as simple as telling Stable Diffusion "give me a book cover with X and Y on it." But I can get images. And I can take those images into GIMP and do what I've already been learning how to do; massage and alter and fiddle with them to make them work with my cover design. Adding elements, combining them, layering on graphics and so on.
Which saves me from having to dig through online for something I can make work. I can just let SD run, flip through what it's kicking out, and when inspiration strikes take that one (or those ones) and see what I can come up with.
I don't care about not owning the cover, which I won't if I source AI images. If I manage to put out a cover that resonates so much people take it, fine. I care about the story, not the cover. That cover is just there to get them to look at the story. And if that cover won't cost me hundreds of dollars, I consider that a win. I think sooner or later I'll get it working, but right now it's still a process I'm learning and fiddling with.
I'm sure an actual graphical artist can do it better, and certainly faster, than I can. Kind of how I can work with a text AGI to get a better story, and polish it better and faster, than they could. But the genie definitely isn't going back in the bottle either way.
Could someone link to the letter?
https://www.theredhandfiles.com/chatgpt-making-things-faster-and-easier/
Stephen Fry is an incredible orator / narrator though, and one of my favorite people in the world. So I would highly recommend still watching the video, but perhaps just reading while you listen.