TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

An unwilling illustrator found herself turned into an AI model

726 pointsby ghuntleyover 2 years ago

62 comments

fleddrover 2 years ago
I&#x27;m thinking a little bit of empathy doesn&#x27;t hurt. Reason from Hollie&#x27;s point of view. She didn&#x27;t ask for this and was working on cool stuff:<p><a href="https:&#x2F;&#x2F;holliemengert.com&#x2F;" rel="nofollow">https:&#x2F;&#x2F;holliemengert.com&#x2F;</a><p>Next, somebody grabs her work (copyrighted by the clients she works for), without permission. Then goes on to try and create an AI version of her style. When confronted, the guy&#x27;s like: &quot;meh, ah well&quot;.<p>Doesn&#x27;t matter if it&#x27;s legal or not, it&#x27;s careless and plain rude. Meanwhile, Hollie is quite cool-headed and reasonable about it. Not aggressive, not threatening to sue, just expressing civilized dislike, which is as reasonable as it gets.<p>Next, she gets to see her name on the orange site, reading things like &quot;style is bad and too generic&quot;, a wide series of cold-hearted legal arguments and &quot;get out of the way of progress&quot;.<p>How wonderful. Maybe consider that there&#x27;s a human being on the other end? Here she is:<p><a href="https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=XWiwZLJVwi4" rel="nofollow">https:&#x2F;&#x2F;www.youtube.com&#x2F;watch?v=XWiwZLJVwi4</a><p>A kind and creative soul, which apparently is now worth 2 hours of GPU time.<p>I too believe AI art is inevitable and cannot be stopped at this point. Doesn&#x27;t mean we have to be so ruthless about it.
评论 #33429230 未加载
评论 #33430156 未加载
评论 #33431021 未加载
评论 #33431850 未加载
评论 #33431823 未加载
评论 #33432121 未加载
评论 #33429222 未加载
评论 #33434366 未加载
评论 #33435170 未加载
评论 #33431417 未加载
评论 #33443624 未加载
评论 #33434840 未加载
评论 #33432974 未加载
评论 #33437109 未加载
评论 #33429231 未加载
评论 #33429931 未加载
评论 #33434574 未加载
评论 #33467097 未加载
评论 #33433586 未加载
评论 #33431288 未加载
评论 #33431663 未加载
cmovqover 2 years ago
I know this website is not a hivemind, but it&#x27;s interesting every time an article like this gets posted the majority opinion seems to be that training diffusion models on copyrighted work is totally fine. In contrast when talking about training code generation models there are multiple comments mentioning this is not ok if licenses weren&#x27;t respected.<p>For anyone who holds both of these opinions, why do you think it&#x27;s ok to train diffusion models on copyrighted work, but not co-pilot on GPL code?
评论 #33424649 未加载
评论 #33430638 未加载
评论 #33424693 未加载
评论 #33428709 未加载
评论 #33427440 未加载
评论 #33424795 未加载
评论 #33424683 未加载
评论 #33424751 未加载
评论 #33424844 未加载
评论 #33425376 未加载
评论 #33429083 未加载
评论 #33430918 未加载
评论 #33427318 未加载
评论 #33424639 未加载
评论 #33430781 未加载
评论 #33424616 未加载
评论 #33424618 未加载
评论 #33425484 未加载
评论 #33424554 未加载
评论 #33424650 未加载
评论 #33427945 未加载
评论 #33431033 未加载
评论 #33430896 未加载
评论 #33429537 未加载
评论 #33424672 未加载
评论 #33424698 未加载
评论 #33434744 未加载
评论 #33425940 未加载
评论 #33425740 未加载
评论 #33430711 未加载
评论 #33433574 未加载
评论 #33424708 未加载
评论 #33430308 未加载
评论 #33424735 未加载
评论 #33432053 未加载
评论 #33426375 未加载
评论 #33426967 未加载
vanadium1stover 2 years ago
I am a graphic artist. In the recent months I&#x27;ve read dozens of articles and threads like this. I still can&#x27;t see what the big deal is.<p>Graphic artists don&#x27;t have trade secrets or unique impossible techniques. If someone can see your picture, he can copy its style. It becomes publicly available as soon as you publish it. For the vast majority of graphic styles, if one author can do it, then hundreds of his colleagues can do it too, often just as well. If one author becomes popular and expensive - then his less popular colleagues can copy his style for cheaper. The market for this is enormous and this was the case for probably hundreds of years.<p>I personally am a non-brand artist like that. More often then not clients come to me with a reference not from my portfolio and ask me to produce something similar. I will do it, probably five times cheaper than the artist or studio who did the original. It may not be exactly as good, but it won&#x27;t be five times worse.<p>Some clients are happy to pay extra for the name brand, and will pay. Some want to spend less, and will settle for a non-brand copy.<p>The clients that are willing to pay for the name brand will still be there for the same reason they are now, and the existence of Stable Diffusion changes nothing to them. And the ones that just want the cheap copy would never contact the big name artist in the first place. The copy market will shift, but the big name artist doesn&#x27;t even have to be aware of it.
评论 #33424516 未加载
评论 #33428725 未加载
评论 #33431770 未加载
评论 #33431124 未加载
评论 #33429197 未加载
评论 #33426729 未加载
gdubsover 2 years ago
Feels like there&#x27;s a difference between artists as drops in a vast ocean of training data, vs explicitly creating a model on one person&#x27;s work. And I think the conversation would benefit from not conflating the two.<p>I&#x27;m sort of a copyright &#x27;moderate&#x27; I suppose. I think people should get paid for their work, and trying to just rip-off a single person&#x27;s style (and I&#x27;m not at all saying this particular example was nefarious in intent) just feels gross. But I also think too much baggage and we stifle new ideas an innovations.<p>However, I also think that most of the conversation around large models like StableDiffusion lack an understanding of how these models actually work. There&#x27;s this misconception that they&#x27;re a kind of &#x27;collage machine&#x27;. The contribution of individual artists in these base models are like drops in a vast vast ocean. [edit: I repeat myself; recovering from Covid, forgive me.] They take this incredibly large set of digitized human creativity, and in turn we all get this amazing tool: a synthesizer for imagination.<p>Anyway, just my personal opinion. It&#x27;s become a very &#x27;us vs them&#x27;, lines-in-the-sand argument these days, and it&#x27;d be great if the conversation could be less heated and more philosophical.
评论 #33424220 未加载
评论 #33424188 未加载
评论 #33424279 未加载
评论 #33424695 未加载
评论 #33424520 未加载
评论 #33424363 未加载
评论 #33424191 未加载
评论 #33435074 未加载
评论 #33424173 未加载
Imnimoover 2 years ago
I feel like there are three issues at play here:<p>-Using her name to describe&#x2F;advertise the fine-tuned model.<p>-Using her illustrations to fine-tune the model.<p>-Using a larger body of potentially unlicensed images to train the base model.<p>For the first, if we had decided that the other steps were fair use or whatever, would it be better or worse if the fine-tuned model had been made available with no mention of the identity of the author of the training images? I&#x27;m not sure.<p>For the second, there is surely a limit after which this sort of thing becomes unambiguously unacceptable. Suppose you fine-tune so aggressively on a small dataset that eventually the model simply reproduces exactly the training images. Now you&#x27;re obviously violating copyright. But where exactly is line before that? If I have a base model that was trained on fully licensed images, and I make one single gradient descent step using a copyrighted image, making imperceptible changes to the model&#x27;s output, surely the resulting images not suddenly in violation. It seems to me that the standard should be that if a human were to draw the output by hand after looking at the training images, would we consider it a violation of copyright? As a thought experiment, imagine someone who lacks the ability to draw but can instead hand-write the weights of a neural network to produce the desired output - it shouldn&#x27;t matter which process they use.<p>For the third, what if I spent a long time prompt engineering on a model trained entirely on properly licensed data and was able to generate a prompt format that produced the outputs we see from this fine-tuned model? In other words, for any generative model, there is a space of reachable outputs, and it&#x27;s not so clear that these images did not already lie in that space before fine-tuning.
评论 #33424271 未加载
评论 #33424941 未加载
评论 #33424742 未加载
评论 #33424290 未加载
评论 #33424853 未加载
sensanatyover 2 years ago
This is going to be a somewhat disjointed and nonsensical comment as I&#x27;m finding it difficult to find the proper words to put down regarding this whole issue.<p>Never in my life did I expect myself to become somewhat of a luddite when it comes to AI, especially considering how much I love the concept in general. Something about the way it seems to be working out in the real world has me questioning my opinions, though.<p>I think more than anything I&#x27;m fearful for the future and what advanced versions of these tools will mean for everyone. Will the world simply become overflooded with these A&quot;I&quot;s for every little thing imaginable? Will every billboard I see and every sound that comes out of a speaker be some weird, uncanny computer-generated thing engineered for the maximum possible user engagement? Will anything anyone ever touches be examined by a dozen different AIs, all looking for ways to cheaply and efficiently replicate it en-masse? Will we all just become a meat shell of fried dopamine receptors, with various AIs and algorithms dictating every facet of our lives?<p>A lot of people call the recent and probably future AI explosion some sort of progress, but from where I&#x27;m sitting it barely feels like progress at all, considering it&#x27;s already being used in the ways I&#x27;ve just described albeit at a less effective scale. Seeing my partner use TikTok to my dismay, and the efficacy of its algorithms and just how good it is at sucking countless hours out is just going to get worse once every single platform we interact with starts doing similar things.<p>It&#x27;s strange, because AI could truly go in any number of possible directions, but so far what I&#x27;m seeing is it&#x27;s going to lead us to ruin, and a lot of people seem to dismiss it or are even accepting of it (even in this very thread) simply in the name of so-called &quot;progress&quot;.
评论 #33431076 未加载
s1artibartfastover 2 years ago
The Supreme Court of the US is currently deciding if Andy Warhol&#x27;s orange prince[1] violates copyright.<p>It is essentially a cropped photo, defeatured, plus the color orange.<p>This could be done easily without the help of AI at all, just cropping and filtering.<p>The case is very interesting and grapples with the same questions.I highly recommended listening to the oral arguments.[2]<p>I doubt the Andy Warhol Foundation will prevail, but it raises all these same questions, without the AI: What constitutes a transformative use of prior work?<p>Can you imbue existing art with new ideas and make it your own?<p><a href="https:&#x2F;&#x2F;en.m.wikipedia.org&#x2F;wiki&#x2F;Orange_Prince_(1984)" rel="nofollow">https:&#x2F;&#x2F;en.m.wikipedia.org&#x2F;wiki&#x2F;Orange_Prince_(1984)</a><p><a href="https:&#x2F;&#x2F;www.oyez.org&#x2F;cases&#x2F;2022&#x2F;21-869" rel="nofollow">https:&#x2F;&#x2F;www.oyez.org&#x2F;cases&#x2F;2022&#x2F;21-869</a>
r_murpheyover 2 years ago
I have been a member of a local musicians union for more decades than I care to admit. If someone was offering to replace me with an AI after a career of dedication to the art, I would worry there is a real risk of incidents of devastating loss of income if there is no financial cushion. Even if I had not lost any gigs yet, I would call a couple of the local union board members to discuss it, and I believe this is a concrete situation which the local board would take up for discussion.<p>I&#x27;m sure there are good things to come out of AI in the arts, especially if it becomes a tool for the artist. But offering to put a financially struggling artist out of work with low effort, even temporarily, is a nightmare for the artist. Guilds and unions [0] [1] have talented artists on boards of directors and lawyers on staff who can help them codify some of the issues into their standard contract used by member artists. I have seen bands fired with no notice when they had used a standard union contract. The contract and access to the Union&#x27;s attorney is the only thing that protected them from loss work and not making rent that month.<p>[0] <a href="https:&#x2F;&#x2F;graphicartistsguild.org&#x2F;" rel="nofollow">https:&#x2F;&#x2F;graphicartistsguild.org&#x2F;</a> [1] <a href="https:&#x2F;&#x2F;www.usa829.org&#x2F;About-Our-Union&#x2F;Categories-Crafts#Scenic-Artist" rel="nofollow">https:&#x2F;&#x2F;www.usa829.org&#x2F;About-Our-Union&#x2F;Categories-Crafts#Sce...</a>
randyrandover 2 years ago
A great artist can take these same prompts, look at the same input images, and produce the same or even better results.<p>No one would call that stealing. Copying, yes. Stealing, No. A great artist can copy other artist&#x27;s style&#x27;s expertly. A large part of being an art student is learning how to do exactly this.<p>Stable Diffusion is just a truly great artist.
评论 #33430611 未加载
评论 #33429843 未加载
评论 #33430097 未加载
i_like_apisover 2 years ago
This Hollie Mengert’s style is (nice, but) not at all original. There are thousands of cartoons that look exactly like this. You could never even tell that anything like this is similar to “her” style.<p>But, even if she did have a distinctive style, there is nothing illegal or unethical about learning that style and producing your own similar artwork, whether you call it “in the style of”, or not.
评论 #33424300 未加载
评论 #33426989 未加载
评论 #33432324 未加载
djoldmanover 2 years ago
It&#x27;s interesting to think about all the controversy surrounding machine-generated images in contrast with a scenario where the same images were generated&#x2F;drawn&#x2F;created directly by a human.<p>In this specific scenario, what would the artist think if 1000 people started drawing in her style and released those images?
评论 #33423770 未加载
评论 #33424177 未加载
评论 #33431858 未加载
评论 #33424043 未加载
评论 #33425489 未加载
chmod775over 2 years ago
It fascinates me how this AI generated art looks good on first glance, but turns into absolute nightmare fuel the longer you look at it.<p>Contorted faces, impossible limbs, mushed hands, surreal cityscape.<p>They look like something a madman pretending to be sane would draw, glimpses of his twisted mind showing through cracks in the facade.<p><a href="https:&#x2F;&#x2F;i.imgur.com&#x2F;EXQlxtF.png" rel="nofollow">https:&#x2F;&#x2F;i.imgur.com&#x2F;EXQlxtF.png</a><p><a href="https:&#x2F;&#x2F;i.imgur.com&#x2F;F1IzpR0.png" rel="nofollow">https:&#x2F;&#x2F;i.imgur.com&#x2F;F1IzpR0.png</a><p><a href="https:&#x2F;&#x2F;i.imgur.com&#x2F;2KvkNnu.png" rel="nofollow">https:&#x2F;&#x2F;i.imgur.com&#x2F;2KvkNnu.png</a>
评论 #33436343 未加载
评论 #33433567 未加载
mwlpover 2 years ago
In sophomore year of college, I came across a Japanese illustrator who sold sticker packs for messenger apps. I liked their art style a lot, to the point where I themed my entire computer setup (wallpaper&#x2F;terminal&#x2F;editor) around their work. Some friends and I worked on a project for an information retrieval class and their art continued to be a centerpiece for the website&#x27;s theme, and we even jokingly snuck an image into our final paper.<p>As much as I loved their style, it seemed they rarely put out new content. A few weeks ago I trained a SD embedding on their work-- it was the coolest thing ever. I thought back to the class project, how I &quot;stole&quot; one of their artworks to use as the favicon. &quot;Nobody outside this class will see this&quot;, I thought. But a pixel art anime girl wearing headphones was perfect branding for the app. &quot;If I ever decide to publish this project, maybe I&#x27;ll commission the artist for an official logo&quot;, I thought at the time. Now I wonder if I&#x27;d just use the SD embedding...
评论 #33424725 未加载
评论 #33425034 未加载
spicyusernameover 2 years ago
It really does feel like the headline for 1970 - 2070 is going to be &quot;Technology desperately tries, but fails, to make the world a better place&quot;.<p>What technology has _really_ improved what it is like to be a human being, around other human beings, and to lead a fulfilling life?<p>Every &quot;disrupted&quot; industry feels like it just has been replaced by a less-human, more unequal, and more dystopian version of itself over the past 50 years.<p>Those who can survive tech jobs seem to have turned out to be the least equipped people to properly navigate us towards utopia.
评论 #33426299 未加载
评论 #33425510 未加载
评论 #33429432 未加载
评论 #33438535 未加载
评论 #33423903 未加载
评论 #33424171 未加载
评论 #33431556 未加载
评论 #33430135 未加载
VBprogrammerover 2 years ago
I don&#x27;t really have a stance on the moral or ethical points but some of the results in the illustrations included here are amazing. If you mixed them up and asked me to identify those which were originals and those which where AI generated I would fail miserably. That in amazing in my book.
评论 #33432822 未加载
yuzuquatover 2 years ago
There&#x27;s already a lot of discussion on the legal&#x2F;moral arguments here so I&#x27;d like to comment on something more concrete.<p>As I understand it, an illustration for a magazine like the New York Times might net anywhere from $100 to $1000 and require 8 hours of work. An illustrator working for someone like the new york times or magic the gathering would likely consider this the pinnacle of a stable job. Many, including my comic books teacher, spent years moonlighting a service job before making it and publishing (Kikuo Johnson). With the advent of generative AI art, it seems immoral from a fiduciary responsibility point of a view that an art director <i>doesn&#x27;t</i> train an AI model on their illustrator&#x27;s art before laying them off.<p>I have no doubt that generative AI will continue to push forward irrespective of the legal arguments being made. I&#x27;m fearful for the frictional unemployment that comes. Having come from art school (and luckily working in tech), my illustration peers are creative but such creativity doesn&#x27;t necessarily translate into creative use of tooling, business-savyness or marketing. All I can say is that I empathize with a lot of the fear and hope for the best.
评论 #33428888 未加载
评论 #33428838 未加载
评论 #33427833 未加载
btillyover 2 years ago
It is worth asking, what legal rights of the author&#x27;s may have been violated here?<p>Glancing at <a href="https:&#x2F;&#x2F;cyber.harvard.edu&#x2F;property&#x2F;library&#x2F;moralprimer.html" rel="nofollow">https:&#x2F;&#x2F;cyber.harvard.edu&#x2F;property&#x2F;library&#x2F;moralprimer.html</a> the following ending sentence seems particularly applicable:<p><i>If a person uses the identity of an author, or the works of the author, for her own benefit without the author&#x27;s permission, then she may have violated the author&#x27;s right of publicity or may be guilty of misappropriation of the author&#x27;s work.</i><p>And the previous line is nearly applicable:<p><i>If authorship of a work is attributed to an author against her will, or misattributed, the author may have a state action for defamation against the person responsible for the attribution.</i><p>Of course &quot;moral rights&quot; are particularly weak in the USA. I&#x27;m sure that there would be a much better case in the EU.<p>Of course this gets directly to the question of what happens when laws conflict with technology. People in technology generally think that technology should win. People who benefit from the laws think that the laws should win. Both popular opinion and real world results generally wind up somewhere between.
评论 #33424292 未加载
评论 #33424006 未加载
评论 #33423965 未加载
评论 #33424044 未加载
zasdffaaover 2 years ago
You can&#x27;t copyright a style. And this is automation, which put lots of people out of work a while back eg <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Luddite" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Luddite</a> amongst other protests (people like Hollie, too). But Disney right now is a soulless factory as much as any other and they&#x27;ll be onto this tech like a shot, and Hollie... I don&#x27;t know. I wish her well.
irrationalover 2 years ago
Why would I want to spend years developing my own art style if I know that it can be copied and used by one of these AI engines in a few hours? Are we going to end up with less human created artwork?
评论 #33429539 未加载
评论 #33432045 未加载
mring33621over 2 years ago
this whole argument is bullshit<p>artists are no more a protected class than programmers<p>learn to use the automation to further your own goals&#x2F;career<p>or move out of the way<p>At some point, some non-artist, non-director, non-screenwriter will make a brilliant movie using these ML-assisted technologies. After few iterations of this success, most people will shut up and learn to harness the benefits.
Waterluvianover 2 years ago
My dad once told me about when he was a kid, some Christmas specials would only be played a few times on CBC right before Christmas. If you missed them, you had to wait a year. He said that became a tradition.<p>When he went to share his tradition with us, he was a bit bothered by the idea of &quot;get it on VHS!&quot; He actually protested the idea. &quot;I didn&#x27;t want you watching it at any time. It should be a tradition.&quot; And indeed it became a tradition for my brothers and I.<p>My kids are 3 and 5 and &quot;Nightmare Before Christmas&quot; was a huge hit last Halloween. My kids wanted to watch it 3 or 4 times that Halloween week. It&#x27;s available for streaming so that was easy. But then they watched it in November. And January. And a bunch in the spring and summer. ...And they never asked for it this year.<p>There&#x27;s a quality bestowed through scarcity. When you have it all the time, on-demand, there is no scarcity. When you can generate art instantly, in any style, of anything, I think it stops being exciting. For example, I don&#x27;t think the concept of &quot;ahahaha look, it&#x27;s the Avengers but as Muppets!&quot; or &quot;It&#x27;s me, but as a Simpson&#x27;s character!&quot; will have any amusement value by next year.<p>Maybe I should be asking Gene Roddenberry but I&#x27;ll ask all of you: do you think there&#x27;s something lost by eliminating scarcity?
评论 #33425410 未加载
评论 #33425405 未加载
评论 #33426818 未加载
评论 #33425463 未加载
评论 #33425264 未加载
评论 #33425936 未加载
评论 #33425331 未加载
评论 #33426154 未加载
评论 #33426507 未加载
评论 #33425690 未加载
评论 #33426426 未加载
评论 #33425040 未加载
sussmannbakaover 2 years ago
Read the thread. Hell, read THIS thread. It’s one thing to steal somebody’s lifework and point towards ongoing legal cases that will make it ok because the practice will benefit big companies. It’s another thing to be an asshole about it. If it was “wow love her style”, sure, still can’t pay her bills with that but it’s actually “mid style, fuck you also go away”. If ML people continue to treat the sole source of their models like this, they shouldn’t be surprised to be treated in a similar fashion.
评论 #33431836 未加载
tayloriusover 2 years ago
Just to imagine how some good <i>might</i> come out of this for Hollie. Lots of famous artists&#x27; original works are very valuable, even though they have a lot of imitators. No matter how many hours of GPU time you have, you can&#x27;t produce an original work by Hollie Mengert. Maybe the AI content firehose could act to enhance her reputation?<p>Maybe artists will start training their own AI replacements. At least then they&#x27;ll control it in some sense, and perhaps they could rent it out to people.
评论 #33432930 未加载
SilverBirchover 2 years ago
I think one of the things that really isn&#x27;t being picked up about this, is that this isn&#x27;t just about copying. Hollie Mengert isn&#x27;t only objecting to her work being stolen. She&#x27;s objecting to having her name associated with it - and she&#x27;s objects because she doesn&#x27;t think it&#x27;s good. And she&#x27;s absolutely right. People really aren&#x27;t engaging much with that as a problem for this tool. I&#x27;d put a simple test to you - sit and look for 1 minute at each of the images she created and the machine created. There&#x27;s a very simple way of telling who did what. The longer you look at Hollie&#x27;s images the more you notice. The longer you look at the generated images the less it makes sense. The perfect example of this is the image of the woman in the middle of the party - there is so much more going on it that single image than all of the generated images together - the focus and composition, and the interaction of the different characters are features basically missing from most of the generated images.<p>I do wonder if a large blind spot of the engineers who are working on this is that they don&#x27;t actually understand art and therefore don&#x27;t understand what they&#x27;re missing.
Waterluvianover 2 years ago
Does anyone else feel painfully unsure of their opinion on all of this? I honestly don&#x27;t recall the last major thing I&#x27;ve felt this completely uncertain about. All my opinions generally lean in one direction at least a little bit.<p>On one hand, I think it might be ridiculous for an artist to get to &quot;own&quot; a &quot;style&quot; of art. In the first example on this page, none of the art looks plagiarized. It looks like what every artist has done: been inspired by or borrowed ideas from other sources.<p>But on the other hand, if left unchecked, this will further harm our creative industries. We&#x27;re going to be starving out our artists because robots can generate art _far_ more easily than they can. If this continues, it disincentivizes anyone from trying the already very uphill battle of making a living by creating art. One might say, &quot;capitalism, baby! we don&#x27;t need those artists, because we have AI and look at what it can do in seconds!&quot; But I think that even if AI can &quot;discover&quot; new art styles and trends, there&#x27;s something lost by humans not doing it.<p>I don&#x27;t think AI will be able to replace human creativity for discovering new paradigms as fast as it will replace human application of existing paradigms. And by doing the latter really well with AI, we&#x27;re killing our ability to do the former. We&#x27;ll end up with a sterile art trajectory.<p>I guess my uncertainty is: something about this _feels wrong_ and yet I cannot point to any one moral&#x2F;ethical thing that feels wrong about it.
评论 #33425251 未加载
评论 #33424992 未加载
评论 #33430212 未加载
评论 #33425511 未加载
评论 #33425124 未加载
评论 #33425811 未加载
yieldcrvover 2 years ago
Reading all of that, the biggest issue was just the art style naming.<p>One of the key features of Stable Diffusion is adding &quot;in the style of &lt;artist name&gt;&quot; to the prompt. This just has a contemporary&#x2F;living artist and actually lets an individual train Stable Diffusion for anyone to add that style to their own Stable Diffusion instance, instead of waiting for Stability.ai to release another dataset.<p>He has since renamed the style, but he should just say &quot;inspired by &lt;artist name&gt;&quot;<p>It is so similar to what someone inspired by a particular artist would do that I can&#x27;t make a separate standard.<p>Basically the pushback comes from the level of discipline that was once required in the past (2 months ago) compared to now. That level of discipline is no longer required.
Yizahiover 2 years ago
I foresee super quick closure of almost all graphic artists in a closed and heavily moderated pay per view&#x2F;use communities. Or they will simply starve. This so called A&quot;I&quot; producing a ton a derivative art will mess up a lot of industries.
评论 #33427684 未加载
评论 #33426211 未加载
detritusover 2 years ago
It managed that from just 32 example illustrations?<p>Irrespective of everything else here, I find that alone hugely impressive.<p>ed - for &#x27;hugely impressive&#x27; also insert &#x27;slightly terrifying&#x27;.
评论 #33433128 未加载
jarrell_markover 2 years ago
Artist using Disney trademarks without their permission has style taken without their permission
评论 #33429407 未加载
protomythover 2 years ago
I get the feeling the ruling on Andy Warhol Foundation for the Visual Arts, Inc. v. Goldsmith will point how the Supreme Court will see this. I&#x27;m starting to think that the entity who owns the data used for training should own the output.
lurquerover 2 years ago
One way to consider copyright: when the concept gained traction 16th and 17th centuries, the doctrine existed to protect the author’s ability to get someone to finance the printing of his work. That is reproduction was costly and required an investment by a printer. The printer faced more economic harm than the author if he spent $20,000 typesetting and printing a thousand copies of something only to find a competitor beat him to the market. To be clear, the copyright attaches to the author, but it is only valuable because it enables the author to induce a printer to publish his work.<p>Now, as the costs of publishing have dropped to zero, the concept is beginning to make less sense.
asciimovover 2 years ago
I think it&#x27;s interesting that for the past 50+ years we have been having an ongoing philosophical debate about the ethics of genetic manipulation. The warning stories are abound in sci-fi, and though we have the tools to do so, we use extreme restraint in their research.<p>Yet we haven&#x27;t had such discussions in computer science about the use of AI. While other fields wrestle with the ethics of doing things, we have had no such discussions. (Due to the nature of our education we are blind to the demands of ethics.) We have made the tools so easy to use, and so accessible, that even if we should discuss the implications of said tools the cat is already out of the bag.<p>Lately, I&#x27;ve been thinking about that poem &quot;First they came...&quot; by Martin Niemöller and wondering if now is the time to heed it&#x27;s warning.<p><pre><code> First they came for the socialists, and I did not speak out— Because I was not a socialist. Then they came for the trade unionists, and I did not speak out— Because I was not a trade unionist. Then they came for the Jews, and I did not speak out— Because I was not a Jew. Then they came for me—and there was no one left to speak for me.</code></pre>
评论 #33429680 未加载
评论 #33424032 未加载
评论 #33424557 未加载
whateveracctover 2 years ago
I think people are being to extreme in their thinking around AI art.<p>Yes, it&#x27;s legal to learn from and imitate art styles.<p>Sure, AI doing the same thing is the same thing.<p>But if I actually copy someone&#x27;s IP (e.g. characters), I am infringing. You can do that with AI. And if anything, it&#x27;s easier to accidentally infringe due to how the AI works. So I think AI will end up coming with more required diligence.<p>This doesn&#x27;t even touch on how art will still be driven by taste and skill. No matter how much tech types wish art wasn&#x27;t actually a skill (is it out of envy? idk. But there&#x27;s a strong HN voice acting like art skill isn&#x27;t actual skill).
lucasfcostaover 2 years ago
I feel for her, but there’s nothing inherently wrong with this and there is no going back.<p>These are original derivatives. In the same way a human is inspired by other arts and may mimic their style, machines can too, they’re just way more efficient. What are they going to do? It feels unfair to punish the machine simply for being more efficient than humans.<p>Instead, artists need to find a novel way to generate revenue by leveraging their uniqueness. Perhaps using Stable Diffusion themselves and turning their creations into NFTs to prove their authenticity.
hosejaover 2 years ago
The journo just can&#x27;t help itself and leads with hostile language right from the start. Hopefully it will soon be made obsolete.
ccbccccbbcccbbover 2 years ago
It&#x27;s simple: if you don&#x27;t respect other&#x27;s intellectual property, expect that no one respects yours, and <i>if you give away for free what others sell for a living just because you have another source of income, expect the latter to be taken away from you too.</i><p>Truly it&#x27;s a sad and ugly state of affairs these days.
swordsmithover 2 years ago
So with these generative models running rampant, what&#x27;s to even motivate aspiring artists to develop and hone their craft, if their years of work can be copied so easily?<p>Maybe it doesn&#x27;t practically matter, because some art style generative model can be developed and feed into the diffusion model so it can generate art of new styles.
评论 #33430243 未加载
jdthediscipleover 2 years ago
In this case the AI guy made public which images he used.<p>However, if someone wants, they can simply never publicize what the training set was, copy other artists and always plausibly deny they used any unlicensed input data, were they to get into legal trouble.<p>I don&#x27;t really see a way around this tbh.
jiggawattsover 2 years ago
Can styles be copyrighted? So for example, if a human artist makes a novel work in the <i>style</i> of another artist, is there any legal grounds for claiming copyright?<p>Assume that the specific characters, names, etc aren’t being used. Think cartoon in the style of Disney, but not Mickey Mouse.
snapcasterover 2 years ago
I feel for her, this is a new frontier we don&#x27;t have the legal framework or the cultural norms to handle this appropriately yet. As a developer, i can see doing something like this because &quot;its cool&quot; and not considering the impact to other people
savant_penguinover 2 years ago
Maybe I&#x27;m oversimplifying but the question is: &quot;does an artist _own_ a style?&quot;<p>To me the answer is no.
评论 #33424373 未加载
honkycatover 2 years ago
Won&#x27;t recordings of music put live performers out of business?<p>...I say this facetiously, but I think the comparison is fair. at this point: it is so cheap and easy and useful to do it feels impossible to stop.<p>$2 and an internet connection. When you are at that point, the game is lost.
tanbog6over 2 years ago
Kind of off topic, but it would be a fascinating experiment if you could take all the art that influenced a real artists and train an ai on it an compare to the artist actual output.<p>Do humans have original ideas or are we synthesizers of what we take in?
DeathArrowover 2 years ago
I think the only disputable thing is using a person&#x27;s name in the model name.<p>If model&#x27;s name would have been Funny Cartoon Creator instead of X Y, then there wouldn&#x27;t be any reason for outrage and vitriolic articles.
makzover 2 years ago
Call me Luddite but this has gone too far, please stop it already.
评论 #33428192 未加载
rendallover 2 years ago
Training data should absolutely be under control of the creator &#x2F; owner of the IP, and AI trained with stolen IP should itself be considered stolen. The lack of clarity here is only because of the newness of the technology, but this will be as clear to our future selves as &quot;warez&quot; are now known to be stolen work.
6stringmercover 2 years ago
How completely unsurprising - a person who grew up in a country most famous for chop yo dolla gives two fucks about the context of achieving his personal desires at the expense of those actually willing to do the work. It’s his own goddamn narrative. I hope Disney bankrupts him as a consequence to his ignorant and selfish actions.
komali2over 2 years ago
&gt; Aside from the IP issues, it’s absolutely going to be used by bad actors: models fine-tuned on images of exes, co-workers, and, of course, popular targets of online harassment campaigns. Combining those with any of the emerging NSFW models trained on large corpuses of porn is a disturbing inevitability.<p>I&#x27;m wondering when the destruction of the traditional internet and trust networks, ala Neal Stephenson&#x27;s &quot;Fall, or, Dodge in Hell&quot; is brought about. In the book (spoilers I suppose), an engineer develops a set of tools to cheaply and easily generate an internet hoax, the example the engineer uses to demonstrate the technology being a fake nuking of a town in if I remember correctly Ohio. Through sockpuppeting and a few choice hiring of actors (and just a teensy bit of image and video manipulation), he&#x27;s able to convince the world for about a day that the USA has lost a remote town to a nuclear strike, and in fact it remains a conspiracy for decades after whether the town actually was nuked or not.<p>The point that I&#x27;m thinking about now though is the engineer develops these tools a bit further, then, after (with consent) targeting them wholesale on a single person, releases them to the world, which basically completely destroys the modern internet as misinformation and information become perfectly interchangeable.<p>A lot of reviews leave out how important this plot point was for the development of the various technologies that feed the main plot of the novel, that being an mind-upload scifi, namely technology around anonymous distributed ledger identification based on traits only AI can accurately tie to a person (i imagined it as like, millions of subtle characteristics such as how someone used their mouse, the pauses between typing certain characters, etc).<p>Long story short it become impossible to trust anything viewed directly on the internet without extensive editor (in the publication sense of the word) support.
bergentyover 2 years ago
So? It’s not a copy of her work, just her style. I see no problems here.
评论 #33427733 未加载
评论 #33427863 未加载
allturtlesover 2 years ago
I think the reason this is creating so much angst is that it&#x27;s not really just about the future of digital art as a career. It doesn&#x27;t take much imagination to see this as the vanguard of what will be coming for all jobs that involve manipulating bits on a computer, which is a huge class of jobs these days, including most of the jobs done by participants on this site.<p>The rapid development of AI may mean the near-completion of the hollowing out of the middle class. On one side the capitalists who own the machines and data that can do all the knowledge work, on the other the mass of workers trying to scrape by in menial service jobs. What will remain in the middle will be the craft trades that require dealing with complex real world physical problems and thus are still extremely hard to automate (plumbers, electricians, etc.).
bfungover 2 years ago
Before this article, I had no idea that a Hollie Mengert existed.<p>In a different light, this could be a new market of people the artist can cater to. A flesh &amp; blood original could hold sentimental value, with the AI being an ok sneak preview.
throwaway12245over 2 years ago
Thing I don&#x27;t get is people post on public internet and then somehow expect that that information is somehow protected from uses that they don&#x27;t want. If you don&#x27;t want your data mis-used, don&#x27;t post it on the internet.
评论 #33424244 未加载
评论 #33424323 未加载
Cypherover 2 years ago
They&#x27;ll be AI&#x27;s that generate new styles eventually.
zmmmmmover 2 years ago
Interestingly, she doesn&#x27;t mind the fact a model was created so much as the commandeering of her name and identity associated to it. Literally everything it produces is associated to her ... whether it&#x27;s beautiful pictures or CSAM.<p>In this respect I&#x27;m curious what right people have in general to object if someone falsely associates your name with something. It&#x27;s not even an AI question really. If I create a different kind of morally objectionable software under the name &quot;Elon Musk&quot; and represent that it embodies his style and personality, can he object to that? If I run a brothel under his name? At what point are natural names effectively implicit trademarks where the person inherits some kind of legal right to control or object about how it&#x27;s used?
sexy_seedboxover 2 years ago
Not a single mention of Greg Rutkowski in this thread?
ChoGGiover 2 years ago
Is it just me, or does her style look like Disney&#x27;s style?<p>Yes I do see that she has drawn Disney characters.
godelskiover 2 years ago
I&#x27;m an AI researcher working on generative modeling, so I want to note my bias upfront.<p>But I&#x27;m also a musician&#x2F;artist and so I find some of these conversations odd. The problem with them I see is that they are oversimplified. To get better at drawing I often copy other works. Or I&#x27;ll play a piece exactly as intended. Then I get more advanced and learn a style of someone I admire and appreciate. Then after that comes my own flair.<p>So I ask, what is different between me doing it and a machine? The specific images being shown in this article shows that Hollie is doing the same process as me and the machine. Their work is in fact derivative. There&#x27;s absolutely nothing wrong with that either. They say a good artist copies but a great artist steals. I don&#x27;t think these generators are great artists, but they sure are good ones.<p>I learn by looking at a lot of art and styles. I learn by copying. I am able to be more efficient than the machine because I can understand nuance and theory, but the machine is able to do the former much better than I can. It can practice its drawings a hundred or thousand times an hour.<p>Now there are highly unethical stuff that is actually going on and I don&#x27;t want the conversation getting distracted from. Just today a twitter account posted a demo of their work and actively demonstrated that one can remove watermarks with their tool[0]. This is bold and borderline illegal (promoting theft). There are also people presenting AI generated work as human digital paintings (we need to be honest about the tools we use). People presenting work in ways that it was not actually created is unethical. But there are other generative ethical concerns.<p>Now there are concerns about photographer&#x27;s&#x2F;artist&#x27;s rights. If I take someone else&#x27;s work and post it as my own, that is straight up theft. Even celebrities can&#x27;t post photos of themselves that were taken by others[1]. This gets muddled if I make minor changes but it&#x27;s been held up in court that the intention matters and it needs to be clear that the changes were new in an artistic manner and not a means of circumventing ownership rights. These are some of the bigger issues we&#x27;re running into.<p>A problem with these generative models is interpretation. How do you know if the image you produced actually exists in the wild or if it is new and unique? There&#x27;s been papers that show that there are privacy concerns[2] and that you can pull ground truth images out of the generator. I&#x27;d argue that this question is easier to answer the more explicit the density your model calculates. Meaning that this is very hard for implicit density models (such as GANs), moderately difficult for approximate density models (such as diffusion and VAEs), but not too bad when pulling from explicit density models (such as Autoregressive or Flow based models).<p>This is a concern that is implicit by articles such as this, but fail to actually quantify the problem here: &quot;How do we meaningfully differentiate generated images from those made by real people?&quot; I&#x27;m a strong advocate for the study and research for explicit density models, but a good chunk of the community is against be (they aren&#x27;t currently anywhere near as powerful, but there&#x27;s nothing theoretically in the way. I&#x27;d argue it is that few people are researching and understanding these models. There is a higher barrier to entry). So I&#x27;d argue that the training methods aren&#x27;t the major concern, but what is actually produced. While the generators learn in a similar fashion to me it is clear that I&#x27;d get in trouble if I was passing off a fake Picasso as a legitimate one. But it is also fine for me to paint something in that same style as long as I&#x27;m honest about it.<p>The nuance here really matters and I think we need to not lose sight of that. This is a complex topic and I would like to hear other views. But I&#x27;m not interested in mic drops or unmovable positions. I don&#x27;t think anyone has the right answer here and to solve it we must get a lot of different view points. So do you agree or disagree with me? I especially want to hear from the latter.<p>[0] <a href="https:&#x2F;&#x2F;twitter.com&#x2F;ai_fast_track&#x2F;status&#x2F;1587475575479959559" rel="nofollow">https:&#x2F;&#x2F;twitter.com&#x2F;ai_fast_track&#x2F;status&#x2F;1587475575479959559</a><p>[1] <a href="https:&#x2F;&#x2F;collenip.com&#x2F;taylor-swift-entitled-say-photographers-rights-belong&#x2F;" rel="nofollow">https:&#x2F;&#x2F;collenip.com&#x2F;taylor-swift-entitled-say-photographers...</a><p>[2] <a href="https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2107.06018" rel="nofollow">https:&#x2F;&#x2F;arxiv.org&#x2F;abs&#x2F;2107.06018</a>
评论 #33428901 未加载
bagelsover 2 years ago
Software is next.
rhyn00over 2 years ago
Is AI the problem here, or is this story a symptom of a bigger problem?<p>This is a somewhat long comment, but just thought I would share my thoughts.<p>The concern is that someone who has worked hard to create art wants use it to leverage it to obtain security in their life, and they are at risk of losing that leverage with new AI technology. Well, first of all, plenty of people work hard their whole lives for low compensation because their skills are not unique, and often don&#x27;t have the time&#x2F;capital&#x2F;energy to build new skills even if they wanted to.<p>I question the way we live our life, there is a lot of suffering in the world because we are focused on taking from each other (in some cases using violence, in other cases laws). In the first world most people are expected to spend their childhood preparing to work away their adult life to earn minimum compensation enriching someone else, and in the third word most people are in extreme poverty, which the first world is happy to exploit for labor, resource extraction, brain drain, strategic military advantage, ect.<p>Artificial intelligence is about to make a lot of things that required difficult and unique training trivial. Making many more people redundant, and no longer unique&#x2F;needed even if we do end up creating copyright laws enforcing that nobody can use someone&#x27;s work to train the AI. Do we really think that we won&#x27;t become redundant just because it doesn&#x27;t train on our art&#x2F;code? So like the artist here, we should all be wondering what is to come.<p>I think reason people are so worried is because we have it in our heads the way to conduct trade, and ourselves is by extracting value from one another. Instead of a culture of giving gifts without guarantee of reward, fostering relationships and caring for and understanding each other, we are trying to take from each other to ensure our needs&#x2F;wants are met. The ultimate way to achieve security under this way of life is to create a dependence, become the owner of the assets&#x2F;capital that can be used to take value&#x2F;labor from others want&#x2F;depend on what we have.<p>We act like we have only a few choices when it comes to how to live, religion, capitalism, communism, socialism, and so on. But we have the choice to try and understand the people we meet in life, and reach out to those who we don&#x27;t know. Knowing each others needs&#x2F;wants we can help each other out. We have the ability to be generous when we can, foster relationships, so that in our time of need we might be helped, by someone that cares (because we&#x27;re there for them too). It&#x27;s just that we listen to the advertisements telling us if we want to be happy, to buy more and more things (possibly at determent of our own health&#x2F;well being), we listen to leaders that would have us go to war with each other, we listen to society telling us to focus on making more money, exchange labor for wealth, and use the wealth to obtain security&#x2F;happiness.<p>What if instead we believed the most important thing was to live with balance with each other and with nature, and to communicate with each other to see how we can help. We do after all live in this world together, and when there is an imbalance we see the effects like poverty, stress&#x2F;anxiety, addiction, theft, violence, wars, exploitation, hate, not to mention the obscene amount of time we spend working (40+hr&#x2F;wk for 40+ years) to enrich someone that doesn&#x27;t care about us.<p>Are we not living in a backwards way? Most people are dependent on a few, by culture, or by force (think of how this force is created - if you don&#x27;t obey those with power&#x2F;authority, someone else who does will threaten you with violence or revoke privileges). If someone breaks a law, or threatens our livelihood, we&#x27;ll just take away their livelihood (or what little they had to begin with) - how well has this worked out?<p>With the advanced technology to communicate more easily than ever before, and provide what we need&#x2F;want with much less labor&#x2F;skill as before, and the wealth of knowledge available, it&#x27;s time realize, that by giving to each other and helping one another, we will all be better off. Well, maybe not the extremely wealthy and powerful powerful people - but even they don&#x27;t get to live in peace because someone is always trying to take their spot - think about all the wars fought because one authority is threatened by another, or all the companies trying to hold their grip in the face of more desperate innovate companies.<p>I&#x27;m not saying doing something different is easy, or straightforward, but it might be better. I think we don&#x27;t need to ask anyone&#x27;s permission to do it, we just start doing it within our own networks - building up each other, making new relationships, looking out for one another.<p>If you want your security to come from laws and by taking instead of giving, how long that will last, just how secure is that anyways?
qullover 2 years ago
So she is claiming the flat corprate style, but with more saturation, as her own? Thats ballsy. It&#x27;s a style for sure, but one based on looking oversimplified and generic imo. No suprise its easy to emulate, and a common averaging result. Its not as likely an actually distinct style, unless massively popular, would have this issue. That said, the greater question is no less valid; what elemens of style do we own, and how will ownership manifest moving forward when many new works use ai images as a starting poit, which are then based on other artists work, some of them being ai generated. Will it become a feedback loop sooner, or later, and what will that look like?
评论 #33423976 未加载
评论 #33424066 未加载
评论 #33424022 未加载
评论 #33424092 未加载
评论 #33424001 未加载
bugfix-66over 2 years ago
Systems like Copilot and Dall-E and so on turn their training data into anonymous common property. Your work becomes my work.<p>This may appeal to naive people (students, hippies, etc.), for whom socialist&#x2F;communist ideas are attractive, but it&#x27;s poison in the real world because it eliminates the reward system that motivates most creative work. People work hard for credit or respect, if they&#x27;re not working for money.<p>Ask yourself, why does the MIT License (<a href="https:&#x2F;&#x2F;opensource.org&#x2F;licenses&#x2F;MIT" rel="nofollow">https:&#x2F;&#x2F;opensource.org&#x2F;licenses&#x2F;MIT</a>) contain the following text?<p><pre><code> Copyright &lt;YEAR&gt; &lt;COPYRIGHT HOLDER&gt; The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. </code></pre> These systems are a mechanism that can regurgitate (digest, remix, emit) without attribution all of the world&#x27;s open code and all of the world&#x27;s art.<p>With these systems, you&#x27;re giving everyone the ability to plagiarize everything, effortlessly and unknowingly. No skill, no effort, no time required. No awareness of the sources of the derivative work.<p>My work is now your work. Everyone and his 10-year old brother can &quot;write&quot; my code (and derivatives), without ever knowing I wrote it, without ever knowing I existed. Everyone can use my hard work, regurgitated anonymously, stripped of all credit, stripped of all attribution, stripped of all identity and ancestry and citation.<p>It&#x27;s a new kind of use not known (or imagined?) when the copyright laws were written.<p>Training must be opt in, not opt out.<p>Every artist, every creative individual, must EXPLICITLY OPT IN to having their hard work regurgitated anonymously by Copilot or Dall-E or whatever.<p>If you want to donate your code or your painting or your music so it can easily be &quot;written&quot; or &quot;painted&quot;, in whole or in part, by everyone else, without attribution, then go ahead and opt in. Most people aren&#x27;t so totally selfless.<p>But if an author or artist does not EXPLICITLY OPT IN, you can&#x27;t use their creative work to train these systems.<p>All these code&#x2F;art washing systems, that absorb and mix and regurgitate the hard work of creative people must be strictly opt in.<p>I say this as a person who writes deep-learning parallel linear algebra kernels professionally.<p>We&#x27;ve crossed a line here.
评论 #33423847 未加载
评论 #33423889 未加载
评论 #33423870 未加载
评论 #33423668 未加载
评论 #33423985 未加载
评论 #33424111 未加载
评论 #33423666 未加载
评论 #33423762 未加载
评论 #33424254 未加载
spoonjimover 2 years ago
This is so clearly a copyright violation that I am hoping for a big lawsuit soon so the courts can confirm it.