You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Jun 22, 2024. It is now read-only.
If anyone's discovering this through the AI community, welcome! 👋😃
If you find Please useful, interesting, or just fun to play with, ✨please✨ give this a read.
When I first found out about OpenAI and all the amazing things you can do with GPT-3, I was so excited. What inspired Please was the sheer amazingness of being able to talk to a computer and have it provide meaningful, intelligent responses. It was the same when I first got to try out DALL-E, just a few weeks before they removed the waitlist and, all of a sudden, anyone with an internet connection could generate images by talking to a machine. Isn't that the most sci-fi thing ever?
I can't remember how I felt the first time I heard artists voicing their concerns about AI image generators displacing human artists, using their work without permission, or appropriating the word "art" to describe soulless images created by the batch. Perhaps I took their concerns into consideration? Perhaps I came up with reasons and justifications why I thought they were mistaken on one point or another? I don't remember.
In any case, I didn't immediately turn my back on AI image generators, or generative AIs in general, as many have done. I kept on using DALL-E to create little bits and pieces of utilitarian artwork—D&D character portraits, mock-ups, that sort of thing—and I kept on using GitHub Copilot to accelerate my day-to-day work.
What I did do, however, is keep thinking, constantly, about the ethical implications of the production and usage of ML models trained on publically available data, and the artists who receive neither credit, nor compensation, nor the means to opt out of their work being used for profit.
At one point, (either ironically or appropriately, you decide), I opened up ChatGPT and asked the AI itself what it thought. After I had explained everything to it to the best of my understanding, the AI agreed with me that the people building these tools were wrong to use artists' work against their collective and loudly-stated wishes. At the very least, we concluded, some opt out mechanism must be provided that simply and effectively removes an artist's work from the training data set and, ideally, from any published models (the technical complexities of this notwithstanding).
I now find myself siding with artists more often than not whenever I see this topic arise. There are certain points I will concede: Arguing about whether generated images are "art" is an argument over semantics and distracts from the real issues in need of discussion. Running a preexisting AI on your own machine to generate images that would otherwise never have existed harms no one and can, and does, bring joy. But I have absolutely no patience for the "tech bros" who claim that artists are gatekeeping the ability to draw. You're punching down, not up.
Going forward, I think all future AI image generators should be trained on public domain or explicitly licensed images. I won't go so far as to say that data sets like LAION shouldn't exist, since there are applications for these data sets outside the realm of generative AI, such as building open source assistive technology, whose existence provides a greater societal good than AI image generators ever will.
I would also encourage you to stop giving money to companies that continue to wilfully disregard artists' moral rights. You're better than that.
But above all else, ✨please✨, stand with artists. Use AI for good.
The text was updated successfully, but these errors were encountered:
If anyone's discovering this through the AI community, welcome! 👋😃
If you find Please useful, interesting, or just fun to play with, ✨please✨ give this a read.
When I first found out about OpenAI and all the amazing things you can do with GPT-3, I was so excited. What inspired Please was the sheer amazingness of being able to talk to a computer and have it provide meaningful, intelligent responses. It was the same when I first got to try out DALL-E, just a few weeks before they removed the waitlist and, all of a sudden, anyone with an internet connection could generate images by talking to a machine. Isn't that the most sci-fi thing ever?
I can't remember how I felt the first time I heard artists voicing their concerns about AI image generators displacing human artists, using their work without permission, or appropriating the word "art" to describe soulless images created by the batch. Perhaps I took their concerns into consideration? Perhaps I came up with reasons and justifications why I thought they were mistaken on one point or another? I don't remember.
In any case, I didn't immediately turn my back on AI image generators, or generative AIs in general, as many have done. I kept on using DALL-E to create little bits and pieces of utilitarian artwork—D&D character portraits, mock-ups, that sort of thing—and I kept on using GitHub Copilot to accelerate my day-to-day work.
What I did do, however, is keep thinking, constantly, about the ethical implications of the production and usage of ML models trained on publically available data, and the artists who receive neither credit, nor compensation, nor the means to opt out of their work being used for profit.
At one point, (either ironically or appropriately, you decide), I opened up ChatGPT and asked the AI itself what it thought. After I had explained everything to it to the best of my understanding, the AI agreed with me that the people building these tools were wrong to use artists' work against their collective and loudly-stated wishes. At the very least, we concluded, some opt out mechanism must be provided that simply and effectively removes an artist's work from the training data set and, ideally, from any published models (the technical complexities of this notwithstanding).
I now find myself siding with artists more often than not whenever I see this topic arise. There are certain points I will concede: Arguing about whether generated images are "art" is an argument over semantics and distracts from the real issues in need of discussion. Running a preexisting AI on your own machine to generate images that would otherwise never have existed harms no one and can, and does, bring joy. But I have absolutely no patience for the "tech bros" who claim that artists are gatekeeping the ability to draw. You're punching down, not up.
Going forward, I think all future AI image generators should be trained on public domain or explicitly licensed images. I won't go so far as to say that data sets like LAION shouldn't exist, since there are applications for these data sets outside the realm of generative AI, such as building open source assistive technology, whose existence provides a greater societal good than AI image generators ever will.
I would also encourage you to stop giving money to companies that continue to wilfully disregard artists' moral rights. You're better than that.
But above all else, ✨please✨, stand with artists. Use AI for good.
The text was updated successfully, but these errors were encountered: