Art & Design

Latest Stories

Machines at the Museum: The Ethical Dilemmas of AI-Produced Art

Written by:

Molly-Anna MaQuirl
Posted: 16-01-2024

Machines at the Museum: The Ethical Dilemmas of AI-Produced Art

This is an AI-generated image created with Midjourney by Molly-Anna MaQuirl

For many of us, being able to transform our ideas into drawings or paintings is an impossible dream. Being an artist requires talent and years of dedication to hone your skills and train that talent. That’s why when the art-generating AI Dall-E was released to the public it dominated AI news for weeks. Being able to produce art without any actual labor was a huge development for all of us.

Dall-E and its successor Dall-E 2 were created by OpenAI. Dall-E was initially released in January 2021 and is a text-to-image model, meaning that it converts written prompts into images. In order to function, the AI needs to be trained on existing images. An AI art generator is only as good as the content that it is taught with. To be effective, it needs to learn for the biggest pool of images possible.

The ethics of AI art became an immediate part of the discussion. The images that Dall-E was trained on weren’t just works that were out of copyright. Most of the art used for training was still under copyright protection or the work of creators who produce art for themselves, not for sale. It’s an important conversation that needs to be had since AI art generators are here to stay and could play an exciting role in the future of art.

Copyright and Creativity

Art enriches our lives, and we are often awed by works of art because we can recognize the amount of effort and talent that went into them. AI art generators have changed that. There have already been some highly publicized examples of AI art being mistaken for human art, including Twin Peaks’ co-creator Mark Frost’s shock that an animated trailer for his classic show was an AI product.

With the technology we have currently, it is impossible for an existing AI to actually be creative. In order to function, an AI art generator has to be trained on existing images, which it then combines into a new form to fulfill a prompt. Without the training materials or the user’s prompt, the AI can’t generate anything.

Being able to craft a perfect prompt that gets the AI to generate exactly what it is that you have in mind is a talent. It requires the ability to describe an image that doesn’t exist so exactly that an algorithm can understand and replicate it. This talent in itself should be respected and is where the real creativity behind AI art lies.

There is a further copyright issue that no one saw coming. This is the issue of whether AI-generated art can be copyrighted. In 2022, Matthew Allen won an award at the Colorado State Fair for his AI-generated artwork. He then tried to copyright the piece and discovered that he couldn’t. He has fought the decision, but the courts have been consistent: copyright can only be applied to works created by humans.

Even in his arguments to the courts, Allen had to acknowledge that AI-generated art could only fall under fair use doctrine because it transformed pre-existing copyrighted material. This case highlights just how complicated the conversation about the ethics of AI art generation will become in the coming years.

Other ethical dilemmas

The question ‘Is AI art ethical?’ extends beyond how AI is trained and the use of copyrighted works. One of the major areas of concern is surrounding identity and deepfakes. A deepfake is an advanced form of photo manipulation. Deepfakes existed before AI art generation was a part of the tech landscape, but now there is the potential for them to become even more realistic and easier to produce.

The most concerning deepfakes are those used harm people’s reputations by showing them in compromising situations or making offensive statements, and those that spread disinformation by showing authority figures sharing false narratives. Deepfakes can erode our ability to distinguish reality from fiction.

AI art generation is a part of this conversation because it can make deepfakes more convincing and more difficult to spot. However, this isn’t the fault of the AI art generator at all. Photo manipulation has been going on for as long as photography has existed, AI is just the latest tool being used.

There is also a lot to be positive about in this area. OpenAI and other companies working in the field have been proactive since the very beginning. For example, OpenAI has programmed Dall-E and Dall-E 2 to reject prompts that include public figures. It has also been trained to reject uploads that contain human faces. Both these steps hurt the ability of Dall-E to generate realistic human faces. The fact that OpenAI has still chosen to implement them shows that they take the problem of deepfakes very seriously.

Final thoughts

In the end, AI art is fun, but it is unethical except under very specific circumstances. If an artist were to feed an AI only their own artworks and then use the AI to explore how other subjects might work in their own style, that would be an ethical use of an AI art generator.

On the surface, AI art generation is a threat to the livelihood of all artists working in 2D mediums — we haven’t merged AI art generators with 3D printers quite yet! If anyone can create any image they want, why would we need real artists? Most likely this will only be fear-mongering as just because we can see the world through Google Streetview doesn’t mean everyone has stopped traveling.

AIs like Dall-E are a lot of fun to play around with and for personal use there isn’t really an issue with using them. The ethical problems arise when people try to profit off the art that they’ve “created” using AI. There are plenty of other creative ways to use AI for art that shouldn’t impact the world of real art.

 

Explore More News

Machines at the Museum: The Ethical Dilemmas of AI-Produced Art

Written by:

Molly-Anna MaQuirl
Posted: 16-01-2024

Machines at the Museum: The Ethical Dilemmas of AI-Produced Art

This is an AI-generated image created with Midjourney by Molly-Anna MaQuirl

For many of us, being able to transform our ideas into drawings or paintings is an impossible dream. Being an artist requires talent and years of dedication to hone your skills and train that talent. That’s why when the art-generating AI Dall-E was released to the public it dominated AI news for weeks. Being able to produce art without any actual labor was a huge development for all of us.

Dall-E and its successor Dall-E 2 were created by OpenAI. Dall-E was initially released in January 2021 and is a text-to-image model, meaning that it converts written prompts into images. In order to function, the AI needs to be trained on existing images. An AI art generator is only as good as the content that it is taught with. To be effective, it needs to learn for the biggest pool of images possible.

The ethics of AI art became an immediate part of the discussion. The images that Dall-E was trained on weren’t just works that were out of copyright. Most of the art used for training was still under copyright protection or the work of creators who produce art for themselves, not for sale. It’s an important conversation that needs to be had since AI art generators are here to stay and could play an exciting role in the future of art.

Copyright and Creativity

Art enriches our lives, and we are often awed by works of art because we can recognize the amount of effort and talent that went into them. AI art generators have changed that. There have already been some highly publicized examples of AI art being mistaken for human art, including Twin Peaks’ co-creator Mark Frost’s shock that an animated trailer for his classic show was an AI product.

With the technology we have currently, it is impossible for an existing AI to actually be creative. In order to function, an AI art generator has to be trained on existing images, which it then combines into a new form to fulfill a prompt. Without the training materials or the user’s prompt, the AI can’t generate anything.

Being able to craft a perfect prompt that gets the AI to generate exactly what it is that you have in mind is a talent. It requires the ability to describe an image that doesn’t exist so exactly that an algorithm can understand and replicate it. This talent in itself should be respected and is where the real creativity behind AI art lies.

There is a further copyright issue that no one saw coming. This is the issue of whether AI-generated art can be copyrighted. In 2022, Matthew Allen won an award at the Colorado State Fair for his AI-generated artwork. He then tried to copyright the piece and discovered that he couldn’t. He has fought the decision, but the courts have been consistent: copyright can only be applied to works created by humans.

Even in his arguments to the courts, Allen had to acknowledge that AI-generated art could only fall under fair use doctrine because it transformed pre-existing copyrighted material. This case highlights just how complicated the conversation about the ethics of AI art generation will become in the coming years.

Other ethical dilemmas

The question ‘Is AI art ethical?’ extends beyond how AI is trained and the use of copyrighted works. One of the major areas of concern is surrounding identity and deepfakes. A deepfake is an advanced form of photo manipulation. Deepfakes existed before AI art generation was a part of the tech landscape, but now there is the potential for them to become even more realistic and easier to produce.

The most concerning deepfakes are those used harm people’s reputations by showing them in compromising situations or making offensive statements, and those that spread disinformation by showing authority figures sharing false narratives. Deepfakes can erode our ability to distinguish reality from fiction.

AI art generation is a part of this conversation because it can make deepfakes more convincing and more difficult to spot. However, this isn’t the fault of the AI art generator at all. Photo manipulation has been going on for as long as photography has existed, AI is just the latest tool being used.

There is also a lot to be positive about in this area. OpenAI and other companies working in the field have been proactive since the very beginning. For example, OpenAI has programmed Dall-E and Dall-E 2 to reject prompts that include public figures. It has also been trained to reject uploads that contain human faces. Both these steps hurt the ability of Dall-E to generate realistic human faces. The fact that OpenAI has still chosen to implement them shows that they take the problem of deepfakes very seriously.

Final thoughts

In the end, AI art is fun, but it is unethical except under very specific circumstances. If an artist were to feed an AI only their own artworks and then use the AI to explore how other subjects might work in their own style, that would be an ethical use of an AI art generator.

On the surface, AI art generation is a threat to the livelihood of all artists working in 2D mediums — we haven’t merged AI art generators with 3D printers quite yet! If anyone can create any image they want, why would we need real artists? Most likely this will only be fear-mongering as just because we can see the world through Google Streetview doesn’t mean everyone has stopped traveling.

AIs like Dall-E are a lot of fun to play around with and for personal use there isn’t really an issue with using them. The ethical problems arise when people try to profit off the art that they’ve “created” using AI. There are plenty of other creative ways to use AI for art that shouldn’t impact the world of real art.