There are two conflated meanings of laziness here: taking shortcuts to save time, and not using one’s time to work. Hard work as a converse of the first represents the waste of human potential (barring some of the side benefits you mentioned). Hard work as a converse of the second can be a virtue *even if* it is only a proxy for things we actually care about. Time not used creating art, improving one’s own or others’ condition, etc, represents a massive opportunity cost. The virtue of hard work is to recognize people who use the efficiency gains of the shortcuts to maximize their potential and create greater total output, rather than the same output for less work.
You're right that I conflated these two (and other) meanings of laziness, which I think is a reasonable thing to do — that's why these concepts are expressed with the same word in the first place! But your thinking about opportunity cost is on point, and if it had occurred to me I might have made it the main point in the conclusion.
I’m mulling it over and I think there’s definitely some more threads to be pulled in that direction.
For example, opportunity cost seems to me to be a fairly fundamental concept for a culture to grok, but I don’t think it’s a primary one. Seems like it requires a culture to fully internalize concepts like time, material scarcity, collective progress / betterment. It would be interesting to investigate under which circumstances this concept of hard work arises. In our current world at least, it seems like a very “fit” virtue, in the cultural Darwinism sense.
But is it a healthy one for the individual? Not as clear. Maybe that’s why we reward virtues in general - they in some sense involve personal sacrifice that is good for the group.
For me personally there's always been this unspoken ambition (see, I own it now) to be lazy AND successful. It feels like the pinnacle of existence to me, for some reason. Succeeding by putting in as little effort as possible feels like moving with grace through the terrain, finding the middle path instead of using brute force or expecting to be carried by someone else...
Interesting to use the word "grace" as an aesthetically charged descriptor of laziness! There really aren't that many positive descriptors of hard work, despite most people agreeing that it's something to strive for
I’ve been thinking a lot about the optimal balance between work and leisure. For ex, in the case of an ai automated world where we don’t have to work for our output, and perhaps our incomes are supplemented by UBI, what amount of work should I do?
When I was between jobs I wrote about 20 hours a week so let’s say that’s my optimal amount of work. You might be able to say I’m working hard at my craft even if you could also say I’m being lazy elsewhere (using a dishwasher, using ai art for my writing, etc).
My output might be bad, or an ai might write better than I do. But what does that matter to me who 1) needs something to do everyday and 2) might as well do something I can continually get better at and work on (Sisyphus does sound like pure torture to me). The value to me is in the hard work. It gives my life meaning.
At the same time, because I have devoted myself to writing, I have not devoted my time elsewhere: say to mastering painting. I could use an ai to paint just as a painter could use an ai to write. We’re both putting hard work into something and being lazy about something else. And I don’t think that takes anything away from either of us. The output is beside the point.
Thank you for such a thought provoking essay! I’ll be pondering this one for awhile....
There certainly is something to the "it's fun!" and "it gives my life meaning!" lines of thinking. Ultimately, something has to matter more than just the immediate outputs of our actions... And hard work, being hard, is a good candidate for this extra something. Maybe.
I think there's a subtle difference between dishwashers/cars and the AI art. AI art requires artists to make the art that the models are trained on. Dishwashers don't need specialized dish-washing-references to learn how to wash a new kind of dish. Cars don't need updates and upgrades to drive on a newly constructed road. For that reason, I think these AI's can be thought of to be doing a kind of "theft" of work.
When I buy a car, I'm paying right there for the work that went into making that car. My payment at the dealership is (assumedly) being distributed across the people who designed it, who worked in the factory that assembled it, who worked in logistics to transport it to my town, etc...
The models these AIs were trained on largely don't have permission from the original artists. There's an entire argument about how human artists get inspired and use ideas too, but they also need to put their own time and effort and knowledge into what they create. Time is possibly the most important as it is impossible to get any back once it is spent. There's probably some way to say that using an AI model trained on an artist's work without their permission is akin to stealing their literal life(time).
I can see an argument to be made for the work-value of the AI program *itself* (like the engineering of the dishwasher). I can also see an argument to be made for the work-value of creating the data sets (like the transportation of the dishwasher) ((and a difference between those that get artists permission and those that do not)). But the work-value from the original artists is largely ignored in most of the data sets (the people in the factories creating the physical dishwasher item).
Yes, the "is AI art theft?" angle is an important one in the current debate, and one that I purposefully did not discuss here. It probably is a bigger concern for actual artists than whether AI model users are lazy, but I find it interesting how the lazy insult is commonly thrown instead of making the more subtle theft point.
That said, I find myself quite unconvinced that AI art is theft. The problem with this line of reasoning is that AI models don't really do anything with the art besides downloading and viewing it (or, well, the AI analog of viewing). In particular, AI models don't: store, sample, copy, modify, remix, adapt, or share the art in the training set. One way to realize this is that AI models are files that are much smaller than the set of art files used for training. At the time of image generation, the model has no access whatsoever to the original art.
If an artist has published art online, then regardless of whether they gave permission for others to store it, sample it, remix it, etc., it can be safely assumed that they gave permission to download and view it. (When I say download, I mean just the downloading of content that happens every time you visit a webpage.) Possibly, our definitions of permissions will now adjust now that "downloading and viewing" has extra consequences, but at the moment it seems very hard to make a case that AI art is theft while you and I viewing a copyrighted image isn't.
I don't think the downloading and viewing part is the problem. I think it's the way that the models (and by extension, users) interpret and use that art to create new images.
For example, I doubt anyone would find an issue with me downloading a copy of every post on this blog to keep on my hard drive. However, if I *republished* the content, that's where things get muddy. If I reposted it openly and with attribution, then at worst I'm taking away from traffic to your blog. However, if republished it without attribution or *charged* for access, most people would consider that theft.
Now, AI doesn't usually copy things exactly. However, it can overfit to it's data in a way that a human artist would be able to knowingly avoid. (For an example, the Afghan woman photo.) With the size of the training sets, it would be practically impossible to tell if an individual generated image was overfitting. Following the earlier example: maybe I don't repost an entire article verbatim, but I have the same sentences with a few choice words changed through the use of an automated thesaurus. I think most people would still consider that to be plagiarism as most of the work was done by you.
Theft and plagiarism in art isn't a black-or-white issue. There's different degrees of inspiration, similarities in technique, etc... Still, I err on the side of extreme caution and say that unless the model was made with permission of all artists involved, it isn't right to *publish* anything made with it - freely or otherwise.
Yeah it's going to be a contentious issue for sure. IP laws aren't prepared for it.
Still, your argument strikes me as being not about AI, but about plagiarism in general. We could replace "AI" with "pen" in your text and have a coherent argument: pens don't plagiarize by default, but they can be used by unscrupulous users to plagiarize. That doesn't mean we should ban pens.
We don't need to go into the myriad ways this analogy breaks down, but hopefully it makes the point clear: intentional plagiarism, whether done with AI or a pen or a text processor, is bad. The invention of AI art doesn't change that. I've said before, and will say again, that using AI to purposefully copy a living artist's work or style against their wishes is a great way to be an asshole.
Possibly, AI art makes *unintentional* plagiarism much easier, and that's a problem. I don't think we have strong norms on what to do when an artist accidentally makes something that's very similar to copyrighted work. I can see a world in which this becomes much more common, and that's the crux of what IP specialists will have to deal with.
This is a good essay but I don't totally buy the argument that we value things primarily because they are rare. By that logic, a drawing I make, if I don't make a lot of drawings, is highly valuable. But in reality nobody cares what I draw. There are clearly other factors at play in determining the value of something.
I also don't believe that hard work is particularly rare. A lot of poor people work pretty hard just to put food on the table. In fact, I think someone working 55 hour weeks working multiple jobs in the food service industry probably works harder than your average CEO. But we value their labor totally differently. Why is that?
I wonder if one of the reasons we value hard work has to do with talent versus skill. Someone might have a great talent to paint, for example, but if they don't improve their skill, they will never become as great as they could be, therefore depriving society of some of their potential. Isn't that part of why we encourage hard work? To develop talents into skills?
Your first paragraph answers the second: you're absolutely right that we value things based on a huge combination of factors, of which hard work and rarity are only two. But all other things being equal, hard work as a personal quality is rarer than laziness. A better word might be 'scarce'—houses aren't particularly rare, but they're scarce, there aren't infinitely many of them.
Your last paragraph points to the question of opportunity cost, which I agree is a good point that didn't make it into the post.
Fine post. 2 things: Nassim Nicholas Taleb seems to consider lazy: fine, a sign of smarts, slack and more ( he praised some at his team as "lazy" - at least in the Malcolm Gladwell story about him - I can not recall it from the 2 books of him I read partly).
Another: How hard is this hard work? Scott Alexander point out the obvious: For those with talent it is often EASY to do their thing. He was praised for his essays that came to him like breathing. But math? "to this day I believe I deserve a fricking statue for getting a C- in Calculus I. It should be in the center of the schoolyard, and have a plaque saying “Scott Alexander, who by making a herculean effort managed to pass Calculus I, even though they kept throwing random things after the little curly S sign and pretending it made sense.”" https://slatestarcodex.com/2015/01/31/the-parable-of-the-talents/
- I can not draw for my life. I know people who can and they all can without effort - or with the same kinda effort I put into a session of CK3, CIV4 or GTAV. You found a way to get the AI to do the pics you wanted? Job well done. Hope it wasn't too hard. ;)
Hmm it occurs to me, after reading your comment, that part of the value of hard work may come from a desire to level the field. Sure, we care about end results, but we also don't want to only praise/reward the most talented and skilled, lots of other people need praise and reward too — kids, beginners, people born in less than ideal situations, etc.
It is complicated - our concepts of hard work has contradictions, as you argued well. Maybe, when we praise great work (art/biz/school), we should praise just the result and drop the talk of "how hard it must have been". - When teachers/ parents/ bosses know the work was hard indeed, and due to this there was an improvement, then they should acknowledge the effort, ofc.. Statues for C-students won"t happen, though. Music: Dire Straits "Money for nothing" - On a final note, the ancient Greek and Romans did consider hard toil a curse. I consider all born in the first world very lucky to be mostly free of it.
When all that surrounds us, at last, is completely fake; when human ingenuity and study and long-cultivated abilities have become nothing more than a joke, then we can be free! We'll be free to be as mediocre as we wish, and yet be surrounded by the our trappings, all fake, of art and literature and science. Our brilliant machines will even be able to generate a great hall of people who will stand up and applaud us for our brilliance.
There are two conflated meanings of laziness here: taking shortcuts to save time, and not using one’s time to work. Hard work as a converse of the first represents the waste of human potential (barring some of the side benefits you mentioned). Hard work as a converse of the second can be a virtue *even if* it is only a proxy for things we actually care about. Time not used creating art, improving one’s own or others’ condition, etc, represents a massive opportunity cost. The virtue of hard work is to recognize people who use the efficiency gains of the shortcuts to maximize their potential and create greater total output, rather than the same output for less work.
You're right that I conflated these two (and other) meanings of laziness, which I think is a reasonable thing to do — that's why these concepts are expressed with the same word in the first place! But your thinking about opportunity cost is on point, and if it had occurred to me I might have made it the main point in the conclusion.
I’m mulling it over and I think there’s definitely some more threads to be pulled in that direction.
For example, opportunity cost seems to me to be a fairly fundamental concept for a culture to grok, but I don’t think it’s a primary one. Seems like it requires a culture to fully internalize concepts like time, material scarcity, collective progress / betterment. It would be interesting to investigate under which circumstances this concept of hard work arises. In our current world at least, it seems like a very “fit” virtue, in the cultural Darwinism sense.
But is it a healthy one for the individual? Not as clear. Maybe that’s why we reward virtues in general - they in some sense involve personal sacrifice that is good for the group.
Yeah, opportunity cost is a relatively complex concept, and many (most?) adults today aren't aware of it, I think.
I really like your analysis.
For me personally there's always been this unspoken ambition (see, I own it now) to be lazy AND successful. It feels like the pinnacle of existence to me, for some reason. Succeeding by putting in as little effort as possible feels like moving with grace through the terrain, finding the middle path instead of using brute force or expecting to be carried by someone else...
Interesting to use the word "grace" as an aesthetically charged descriptor of laziness! There really aren't that many positive descriptors of hard work, despite most people agreeing that it's something to strive for
True. The best image I have of hard work is hitting at a rock again and again knowing it will crack one day
I’ve been thinking a lot about the optimal balance between work and leisure. For ex, in the case of an ai automated world where we don’t have to work for our output, and perhaps our incomes are supplemented by UBI, what amount of work should I do?
When I was between jobs I wrote about 20 hours a week so let’s say that’s my optimal amount of work. You might be able to say I’m working hard at my craft even if you could also say I’m being lazy elsewhere (using a dishwasher, using ai art for my writing, etc).
My output might be bad, or an ai might write better than I do. But what does that matter to me who 1) needs something to do everyday and 2) might as well do something I can continually get better at and work on (Sisyphus does sound like pure torture to me). The value to me is in the hard work. It gives my life meaning.
At the same time, because I have devoted myself to writing, I have not devoted my time elsewhere: say to mastering painting. I could use an ai to paint just as a painter could use an ai to write. We’re both putting hard work into something and being lazy about something else. And I don’t think that takes anything away from either of us. The output is beside the point.
Thank you for such a thought provoking essay! I’ll be pondering this one for awhile....
There certainly is something to the "it's fun!" and "it gives my life meaning!" lines of thinking. Ultimately, something has to matter more than just the immediate outputs of our actions... And hard work, being hard, is a good candidate for this extra something. Maybe.
I suppose there is always the option to not work hard on anything and still feel virtuous about it??? But that’s hard for me to imagine.
Masterpiece.
Have had the same question numerous times but I have been too lazy to put the work to answer it so thoroughly 😂
Thank you!
I think there's a subtle difference between dishwashers/cars and the AI art. AI art requires artists to make the art that the models are trained on. Dishwashers don't need specialized dish-washing-references to learn how to wash a new kind of dish. Cars don't need updates and upgrades to drive on a newly constructed road. For that reason, I think these AI's can be thought of to be doing a kind of "theft" of work.
When I buy a car, I'm paying right there for the work that went into making that car. My payment at the dealership is (assumedly) being distributed across the people who designed it, who worked in the factory that assembled it, who worked in logistics to transport it to my town, etc...
The models these AIs were trained on largely don't have permission from the original artists. There's an entire argument about how human artists get inspired and use ideas too, but they also need to put their own time and effort and knowledge into what they create. Time is possibly the most important as it is impossible to get any back once it is spent. There's probably some way to say that using an AI model trained on an artist's work without their permission is akin to stealing their literal life(time).
I can see an argument to be made for the work-value of the AI program *itself* (like the engineering of the dishwasher). I can also see an argument to be made for the work-value of creating the data sets (like the transportation of the dishwasher) ((and a difference between those that get artists permission and those that do not)). But the work-value from the original artists is largely ignored in most of the data sets (the people in the factories creating the physical dishwasher item).
Yes, the "is AI art theft?" angle is an important one in the current debate, and one that I purposefully did not discuss here. It probably is a bigger concern for actual artists than whether AI model users are lazy, but I find it interesting how the lazy insult is commonly thrown instead of making the more subtle theft point.
That said, I find myself quite unconvinced that AI art is theft. The problem with this line of reasoning is that AI models don't really do anything with the art besides downloading and viewing it (or, well, the AI analog of viewing). In particular, AI models don't: store, sample, copy, modify, remix, adapt, or share the art in the training set. One way to realize this is that AI models are files that are much smaller than the set of art files used for training. At the time of image generation, the model has no access whatsoever to the original art.
If an artist has published art online, then regardless of whether they gave permission for others to store it, sample it, remix it, etc., it can be safely assumed that they gave permission to download and view it. (When I say download, I mean just the downloading of content that happens every time you visit a webpage.) Possibly, our definitions of permissions will now adjust now that "downloading and viewing" has extra consequences, but at the moment it seems very hard to make a case that AI art is theft while you and I viewing a copyrighted image isn't.
I don't think the downloading and viewing part is the problem. I think it's the way that the models (and by extension, users) interpret and use that art to create new images.
For example, I doubt anyone would find an issue with me downloading a copy of every post on this blog to keep on my hard drive. However, if I *republished* the content, that's where things get muddy. If I reposted it openly and with attribution, then at worst I'm taking away from traffic to your blog. However, if republished it without attribution or *charged* for access, most people would consider that theft.
Now, AI doesn't usually copy things exactly. However, it can overfit to it's data in a way that a human artist would be able to knowingly avoid. (For an example, the Afghan woman photo.) With the size of the training sets, it would be practically impossible to tell if an individual generated image was overfitting. Following the earlier example: maybe I don't repost an entire article verbatim, but I have the same sentences with a few choice words changed through the use of an automated thesaurus. I think most people would still consider that to be plagiarism as most of the work was done by you.
Theft and plagiarism in art isn't a black-or-white issue. There's different degrees of inspiration, similarities in technique, etc... Still, I err on the side of extreme caution and say that unless the model was made with permission of all artists involved, it isn't right to *publish* anything made with it - freely or otherwise.
Yeah it's going to be a contentious issue for sure. IP laws aren't prepared for it.
Still, your argument strikes me as being not about AI, but about plagiarism in general. We could replace "AI" with "pen" in your text and have a coherent argument: pens don't plagiarize by default, but they can be used by unscrupulous users to plagiarize. That doesn't mean we should ban pens.
We don't need to go into the myriad ways this analogy breaks down, but hopefully it makes the point clear: intentional plagiarism, whether done with AI or a pen or a text processor, is bad. The invention of AI art doesn't change that. I've said before, and will say again, that using AI to purposefully copy a living artist's work or style against their wishes is a great way to be an asshole.
Possibly, AI art makes *unintentional* plagiarism much easier, and that's a problem. I don't think we have strong norms on what to do when an artist accidentally makes something that's very similar to copyrighted work. I can see a world in which this becomes much more common, and that's the crux of what IP specialists will have to deal with.
This is a good essay but I don't totally buy the argument that we value things primarily because they are rare. By that logic, a drawing I make, if I don't make a lot of drawings, is highly valuable. But in reality nobody cares what I draw. There are clearly other factors at play in determining the value of something.
I also don't believe that hard work is particularly rare. A lot of poor people work pretty hard just to put food on the table. In fact, I think someone working 55 hour weeks working multiple jobs in the food service industry probably works harder than your average CEO. But we value their labor totally differently. Why is that?
I wonder if one of the reasons we value hard work has to do with talent versus skill. Someone might have a great talent to paint, for example, but if they don't improve their skill, they will never become as great as they could be, therefore depriving society of some of their potential. Isn't that part of why we encourage hard work? To develop talents into skills?
Your first paragraph answers the second: you're absolutely right that we value things based on a huge combination of factors, of which hard work and rarity are only two. But all other things being equal, hard work as a personal quality is rarer than laziness. A better word might be 'scarce'—houses aren't particularly rare, but they're scarce, there aren't infinitely many of them.
Your last paragraph points to the question of opportunity cost, which I agree is a good point that didn't make it into the post.
Fine post. 2 things: Nassim Nicholas Taleb seems to consider lazy: fine, a sign of smarts, slack and more ( he praised some at his team as "lazy" - at least in the Malcolm Gladwell story about him - I can not recall it from the 2 books of him I read partly).
Another: How hard is this hard work? Scott Alexander point out the obvious: For those with talent it is often EASY to do their thing. He was praised for his essays that came to him like breathing. But math? "to this day I believe I deserve a fricking statue for getting a C- in Calculus I. It should be in the center of the schoolyard, and have a plaque saying “Scott Alexander, who by making a herculean effort managed to pass Calculus I, even though they kept throwing random things after the little curly S sign and pretending it made sense.”" https://slatestarcodex.com/2015/01/31/the-parable-of-the-talents/
- I can not draw for my life. I know people who can and they all can without effort - or with the same kinda effort I put into a session of CK3, CIV4 or GTAV. You found a way to get the AI to do the pics you wanted? Job well done. Hope it wasn't too hard. ;)
Hmm it occurs to me, after reading your comment, that part of the value of hard work may come from a desire to level the field. Sure, we care about end results, but we also don't want to only praise/reward the most talented and skilled, lots of other people need praise and reward too — kids, beginners, people born in less than ideal situations, etc.
It is complicated - our concepts of hard work has contradictions, as you argued well. Maybe, when we praise great work (art/biz/school), we should praise just the result and drop the talk of "how hard it must have been". - When teachers/ parents/ bosses know the work was hard indeed, and due to this there was an improvement, then they should acknowledge the effort, ofc.. Statues for C-students won"t happen, though. Music: Dire Straits "Money for nothing" - On a final note, the ancient Greek and Romans did consider hard toil a curse. I consider all born in the first world very lucky to be mostly free of it.
Sisyphus reminds me of this motivational talk by a Navy SEAL named Jocko: https://youtu.be/7qNGWjDzxpQ?feature=shared
When all that surrounds us, at last, is completely fake; when human ingenuity and study and long-cultivated abilities have become nothing more than a joke, then we can be free! We'll be free to be as mediocre as we wish, and yet be surrounded by the our trappings, all fake, of art and literature and science. Our brilliant machines will even be able to generate a great hall of people who will stand up and applaud us for our brilliance.