EA should avoid using AI art for non-research purposes?
My strongest reason for disliking AI-generated images is that so often they look tacky, as you aptly said, or even disgustingly bad.
One of the worst parts of AI-generated art is that sometimes it looks good at a glance and then as you look at it longer, you notice some horribly wrong detail. Human art (if itās good quality) lets you enjoy the small details. It can be a pleasure to discover them. AI-generated art ruins this by punishing you for paying close attention.
But thatās a matter of taste.
What Iām voting ādisagreeā on is that the EA Forum should have a rule or a strong social norm against using AI-generated images. I donāt think people should use ugly images, whether theyāre AI-generated or free stock photos. But leave it up to people decide on a case-by-case basis which images are ugly and donāt make it a rule about categorically banning AI-generated images.
I am trying to be open-minded to the ethical arguments against AI-generated art. I find the discourse frustratingly polarized.
For example, a lot of people are angry about the supposed environmental impact of AI-generated art, but what is the evidence of this? Anytime Iāve tried to look up hard numbers on how much energy AI uses, a) itās been hard to find clear, reliable information and b) the estimates Iāve found tend to be pretty small.
Similarly, is there evidence that AI-generated images are displacing the labour of human artists? Again, this is something Iāve tried to look into, but the answer isnāt easy to find. There are anecdotes here and there, but itās hard to tell if there is a broader trend that is significantly affecting a large number of artists.
Itās difficult to think about the topic of whether artists should need to give permission for their images to be used for AI training or should be compensated if they are. There is no precedent in copyright law to cover this because this technology is unprecedented. For the same reason, there is no precedent in societal norms. We have to decide on a new way of thinking about a new situation, without traditions to rely on.
So, if the three main ethical arguments against AI-generated art are:
-It harms the environment -It takes income away from human artists -AI companies should be required to get permission from artists before training AI models on their work and/āor financially compensate them if they do
All three of these arguments feel really unsubstantiated to me. My impression right now is:
-Probably not -Maybe? Whatās the evidence? -Maybe? I donāt know. Whatās the reasoning?
The main aesthetic argument against AI-generated art is of course:
-Itās ugly
And I mostly agree. But those ChatGPT images in the Studio Ghibli style are absolutely beautiful. There is a 0% chance I will ever pay an artist to draw a Studio Ghibli-style picture of my cat. But I can use a computer to turn my cat into a funny, cute little drawing. And thatās wonderful.
Iām a politically progressive person. Iām LGBT, Iām a feminist, I believe in social justice, Iāve voted for a social democratic political party multiple times, and Iāve been in community and in relationship with leftists a lot. I am so sick of online leftist political discourse.
I am not interested in thinking and talking about celebrities all the time. (So much online leftist discourse is about celebrities.)
I donāt want to spend that much time and energy constantly re-evaluating which companies I boycott and whether thereās a marginally more ethical alternative.
I donāt want every discussion about every topic to be polarized, shut down, moralized, and made into a red line issue where disagreement isnāt tolerated. Iām sick of hyperbolic analogies between issues like ChatGPT and serious crimes. (I could give an example I heard but itās so offensive I donāt want to repeat it.)
I am fed up with leftists supporting authoritarianism, terrorism, and political assassinations. While moralizing about AI art.
So, please forgive me if I struggle to listen to all of online leftistsā complaints with the charity they deserve. I am burnt out on this stuff at this point.
I donāt know how to fix the offline left, but Iām personally so relieved that I donāt use microblogging anymore (i.e., Twitter, Bluesky, Mastodon, or Threads) and that Iāve mostly extricated myself from online leftist discourse otherwise. Itās too crazymaking for me to stomach.
My strongest reason for disliking AI-generated images is that so often they look tacky, as you aptly said, or even disgustingly bad.
One of the worst parts of AI-generated art is that sometimes it looks good at a glance and then as you look at it longer, you notice some horribly wrong detail. Human art (if itās good quality) lets you enjoy the small details. It can be a pleasure to discover them. AI-generated art ruins this by punishing you for paying close attention.
But thatās a matter of taste.
What Iām voting ādisagreeā on is that the EA Forum should have a rule or a strong social norm against using AI-generated images. I donāt think people should use ugly images, whether theyāre AI-generated or free stock photos. But leave it up to people decide on a case-by-case basis which images are ugly and donāt make it a rule about categorically banning AI-generated images.
I am trying to be open-minded to the ethical arguments against AI-generated art. I find the discourse frustratingly polarized.
For example, a lot of people are angry about the supposed environmental impact of AI-generated art, but what is the evidence of this? Anytime Iāve tried to look up hard numbers on how much energy AI uses, a) itās been hard to find clear, reliable information and b) the estimates Iāve found tend to be pretty small.
Similarly, is there evidence that AI-generated images are displacing the labour of human artists? Again, this is something Iāve tried to look into, but the answer isnāt easy to find. There are anecdotes here and there, but itās hard to tell if there is a broader trend that is significantly affecting a large number of artists.
Itās difficult to think about the topic of whether artists should need to give permission for their images to be used for AI training or should be compensated if they are. There is no precedent in copyright law to cover this because this technology is unprecedented. For the same reason, there is no precedent in societal norms. We have to decide on a new way of thinking about a new situation, without traditions to rely on.
So, if the three main ethical arguments against AI-generated art are:
-It harms the environment
-It takes income away from human artists
-AI companies should be required to get permission from artists before training AI models on their work and/āor financially compensate them if they do
All three of these arguments feel really unsubstantiated to me. My impression right now is:
-Probably not
-Maybe? Whatās the evidence?
-Maybe? I donāt know. Whatās the reasoning?
The main aesthetic argument against AI-generated art is of course:
-Itās ugly
And I mostly agree. But those ChatGPT images in the Studio Ghibli style are absolutely beautiful. There is a 0% chance I will ever pay an artist to draw a Studio Ghibli-style picture of my cat. But I can use a computer to turn my cat into a funny, cute little drawing. And thatās wonderful.
Iām a politically progressive person. Iām LGBT, Iām a feminist, I believe in social justice, Iāve voted for a social democratic political party multiple times, and Iāve been in community and in relationship with leftists a lot. I am so sick of online leftist political discourse.
I am not interested in thinking and talking about celebrities all the time. (So much online leftist discourse is about celebrities.)
I donāt want to spend that much time and energy constantly re-evaluating which companies I boycott and whether thereās a marginally more ethical alternative.
I donāt want every discussion about every topic to be polarized, shut down, moralized, and made into a red line issue where disagreement isnāt tolerated. Iām sick of hyperbolic analogies between issues like ChatGPT and serious crimes. (I could give an example I heard but itās so offensive I donāt want to repeat it.)
I am fed up with leftists supporting authoritarianism, terrorism, and political assassinations. While moralizing about AI art.
So, please forgive me if I struggle to listen to all of online leftistsā complaints with the charity they deserve. I am burnt out on this stuff at this point.
I donāt know how to fix the offline left, but Iām personally so relieved that I donāt use microblogging anymore (i.e., Twitter, Bluesky, Mastodon, or Threads) and that Iāve mostly extricated myself from online leftist discourse otherwise. Itās too crazymaking for me to stomach.