Imagine watching a movie trailer. There’s a space man exploring a vast white planet. He has an orange-red knit cap and the sun is shining bright. It’s an impressive trailer and it might make you want to go see the full movie. However, when you look into it you are faced with the fact the entire trailer, down to the orange pixels in the space explorers knit cap, were generated by artificial intelligence. Well, as many people might already know, this is a description of an actual video made by AI. There is no denying the remarkable images and videos that AI have been able to create, but that still does not erase the issues with it. The issues AI generated work have raised are everything from ethical conundrums all the way to legal concerns and exploitative material.
Let’s begin by establishing some facts to know about AI. While there are many different kinds of “AI”, for this discussion “AI” will simply mean any kind of generative Artificial Inteligence that can take what it’s learned and produce something ‘new’. Examples of this would be programs like ChatGPT, DALL·E 2, and Sora AI. The way these programs are able to produce something new is through being fed similar content en masse and use the patterns found in that content to create a ‘unique’ piece of content. For instance, if a generative AI model that writes essays was fed all of your pieces of writing, then you gave it a simple topic for a new essay, it could spit out an essay with a similar style to your own.
Now knowing how it works, in the simplest way, one might wonder where all the content the AI is learning from is sourced. The answer to that is quite simple as well: The internet. While at first that might seem innocent, it can be far from it. Generative AI is often being fed content without ever being given permission to do so. Image AI has been learning through artwork even when the artists have explicitly said they do not want their work training AI. Additionally, many websites have been updating their terms of service to include allowing them to sell any user’s content to train AI without first asking for permission.
Not only has AI been using art from the internet without permission to learn how to “create art,” the people that own the AI lied about it. In 2022, Midjourney’s founder David Holz admitted that they hadn’t made an effort to ask the artists the AI was learning from during an interview with Forbes. Additionally, a long list of artists who were used to train Midjourney was released. Many of these artists were infuriated that their art was not only used without their permission but also without their knowledge. While some of the artists were well-known famous artists (Frida Kahlo, Andy Warhol, etc.), there were even some current artists ranging in popularity.
AI trained on real artists–especially without their permission–will be using these artist’s styles to produce the ‘new art’. The generated art the AI can be prompted to emulate a particular artist’s style, creating a potential legal issue. While human artists will draw inspiration from artists they admire and implement small parts of their styles into their own, AI is very different. An AI is a program, completely artificial, and can’t intertwine an artist’s style with their own, they don’t even have their ‘own’ style.
While there are current open lawsuits against AI companies like Midjourney and Stability AI for copyright infringement, it will likely take some time before they are settled. However, in August of 2023 it was ruled that AI generated art cannot be copyrighted by a federal court. In short, this means that any art generated by AI can be used by anyone for any reason without asking the original poster.
Along with the legal issues AI brings up, there is an ethical issue as well. When anyone can use an AI to create art of anything they want in the style of any artist the AI has learned from, it can be used to hurt real artists. If someone were to take a distinct feature of someone’s art and add it into their own without acknowledging where they were inspired, it would absolutely violate copyright laws. Yet, when AI is doing it there are no current punishments in place. While pieces from artists like Van Gogh are in the public domain, current artists are still protected by copyright.
At the same time, there are new issues being faced since Sora AI’s launch. It’s an impressive video generative AI. All you need to do is give it a written prompt and it will generate an entire video. Though Sora AI isn’t public, it shows what the technology of AI is capable of doing now. In a few years, will people be able to trust videos anymore? In court, will video evidence be able to be written off due to the possibility of it being AI generated? The positive usages of video generative AI like Sora AI are far outweighed by the nefarious usages.
With that being said, AI still isn’t perfect. There are many ways to spot AI generated images and videos and they don’t need a professional eye to notice. Most notably, AI has a very difficult time making normal looking hands. If you see a picture and the subject’s hands have missing or extra fingers or looks wildly misshapen, there is a good chance that it’s AI. Also, a lot of the time AI will be able to create the general shape of something like a flower, but with closer inspection it’s just some smushed together shapes. The final tip to spot AI would be to look at whether you could definitely trace the picture with solid lines. Meaning, there’s no shapes that just blend into each other for no reason or blend into the background randomly.
The technology in the world is improving in vast and exciting ways, however that doesn’t mean it should harm people in its wake. From writers, to artists and now regular people, generative AI has the ability to harm and it won’t be remorseful about it. Being aware of AI and making sure not to let yourself be tricked by it is an important skill now, but many people don’t have it.
[Works Cited]
https://www.nvidia.com/en-us/glossary/generative-ai/
