You’ve probably heard of DALL-E, or at least seen its capabilities. Released in 2021 by OpenAI, DALL-E operates as a form of creative AI. To put it in simple terms (my technological knowledge is also quite rudimentary, as a disclaimer) it learns by scouring the internet and scraping millions of images and text captions. When users plug in certain keywords, “butterfly”, for example, DALL-E generates an image from this word prompt. The generated image is produced from a cluster of various images and bits of metadata that the AI has extrapolated from the internet. I think fascination with it is only natural, seeing that we are at a point in technological history where a machine can generate images which are extremely realistic and rarely fails in that regard.
While it is an impressive feat, ethical questions are starting to arise. After the success that DALL-E saw, Boris Daymas began an AI learning model, Craiyon as an attempt to replicate DALL-E. A lot of the images Craiyon used contained watermarks that denoted that the images in question were copyrighted. This was especially seen in landscape photos, but Google had listed these watermarked images for use. Despite these images coming from sites like Shutterstock: a photograph-hosting website that has written policies against datamining and the use of robots for gathering data. Concerns about piracy, plagiarism and payment to the owners of these images have been voiced. There have been worries expressed by artists who fear they will be displaced by creative AI. I think that these concerns should be taken seriously: perhaps these fears seem hyperbolic, but freelance artists have had their digital art stolen by learning AI without being recompensated financially.
For years, I’ve seen artists who post their work online have it reposted elsewhere, or otherwise stolen. Reposting art without consent or proper crediting has its negative consequences for these freelance artists. I’ve often seen uncredited, reposted art becoming more popular than the original post that the artist made. There is a financial cost to this: they lose the potential to be commissioned if no one credits them and directs them to their profiles. It’s easy to imagine how frustrating it is when attempting to make a living from making art. Preventing it seems inevitable at times, with how vast and populated social is. People also seem entitled to art that is posted online: there’s always one person commenting on how if you post art on the internet, you should accept that it’s going to be stolen, or how it takes work to credit artists, how if you don’t want your art to be stolen, don’t post it. The list goes on. This sort of discourse has been happening before the advent of AI art. With the release of LENSA in 2018, these issues have been exacerbated. According to TechCrunch, LENSA has accrued $29 million since then. The complaint lies in the open-source library that LENSA is trained on. Multiple artists have claimed that their art has been used without their permission for LENSA’s learning. While people may argue that a machine that uses aspects from thousands of works shouldn’t be considered plagiarism, the issue is one of consent. If an artists’ work is used for AI learning, they surely had a role to play, so why shouldn’t they also be recompensated with the profits from LENSA?
There has always been debate on what qualifies as art and artists who try to stretch the parameters of it. Think Andy Warhol’s pop art, Jackson Pollock’s drip paintings. Even a few years ago, when that banana duct-taped to a canvas surfaced and sparked arguments. Everyone has an opinion formed on what is supposed to be art and what isn’t. AI art isn’t divorced from this argument. Advocates for these learning AI would certainly call it art. But for these creative AI to function, it needs to absorb information from the pre-existing creations of people. For me, autonomising the creative process seems soulless and inorganic. Is someone plugging in a few keywords, having a machine spit out an image generated from hundreds of (possibly stolen) images considered art? Should it be?
References
Cooper, Daniel. “Is Dall-E’s Art Borrowed or Stolen?” Engadget, 27 July 2022, https://www.engadget.com/dall-e-generative-ai-tracking-data-privacy-160034656.html
Silberling, Amanda. “Lensa Ai Climbs the App Store Charts as Its ‘Magic Avatars’ Go Viral.” TechCrunch, TechCrunch, 1 Dec. 2022, https://techcrunch.com/2022/12/01/lensa-ai-climbs-the-app-store-charts-as-its-magic-avatars-go-viral/.
Morgan, Sung. “Lensa, the AI Portrait App, Has Soared in Popularity. but Many Artists Question the Ethics of AI Art.” NBCNews.com, NBCUniversal News Group, 6 Dec. 2022, https://www.nbcnews.com/tech/internet/lensa-ai-artist-controversy-ethics-privacy-rcna60242.
“Terms of Use.” Shutterstock, https://www.shutterstock.com/terms#anchor_intellectual_property.
Great read! I had no idea the LENSA was making such absurd amounts of money.
Do you think this market will see more paid services reaching this level of popularity or is there hope something will be done to protect artists from future potential theft?
Hi, Conor! Thanks for reading my blog, I appreciate it. If I recall correctly, I believe Alex Cardinell, CEO of Glimpse.ai mentioned that he hopes to implement financial compensation or a sort of universal basic income for artists who have been displaced as a result of AI technology. But that’s the only explicit statement I’ve seen so far that offers any kind of recompense for artists, unfortunately. With the success of LENSA, there will probably be more creative AI that will be trained with stolen art, but I hope that more measures will be taken in the future so that perhaps they could work together or figure out parameters that ensure that work does not get stolen inadvertently.