There is no denying the dramatic impact that artificial intelligence has had over the last few months. Platforms like ChatGPT have seen users experience a gambit of emotions: excitement, amusement, skepticism, and for some terror. We are now starting to see AI enter the realms of photography, art, and music. The question is: are we ready?
The possibilities that lie before humanity with AI are seemingly endless. However, one of the critical challenges we face is ethical use. AI has rapidly encroached into creative areas, leaving a trail of controversy along the way. This was first seen in the rise of text-to-image platforms but has evolved now to impact the music world and even photography awards.
The rise of text-to-image platforms has gone through a rapid rise over recent months. Open AI rolled out the second iteration of their text-to-image solution, DALL-E in July 2022. The results of the AI saw people creating entertaining, and sometimes terrifying, images. While the images were clearly created by AI combining images together, the concept was sound and hinted at what the future would look like.
Plenty of text-to-image solutions have arisen over the last year, and now some, like Midjourney, can create photorealistic images. These images are now becoming so convincing, that a recent Sony photography competition was won by someone who submitted a ‘photo’ made in Midjourney. AI images are not permitted in competitions, meaning this image managed to successfully fool the judges. The person who submitted the image declined the prize and admitted to the use of AI.
This photography competition issue does help to highlight the big threat: AI is evolving so quickly it is becoming indistinguishable from reality, and real creativity. Just look at the hysteria a few weeks ago when an image surfaced of Pope Francis wearing a stylish puffer jacket. It was an image that seemed unbelievable but at the same time possible. In truth, it was an image created in Midjourney, but it captured the attention of many. These photorealistic fakes can pose genuine threats to the understanding of truth and reality. However, this is nothing new.
As Adobe’s Photoshop grew in popularity, there was an influx of news articles fearing the rise of misinformation. Photoshop offered a way for people to easily fake images and make them look almost real. In an article from 2010, one person wrote that “faking a photo is easier today than ever before”. The fear of misinformation has been prevalent with the rise of new technologies but has been exacerbated over the last 20 years. The biggest difference is the ease with which people can now do similar things that were possible in Photoshop. To create a highly realistic fake image, people needed a lot of time and a deep understanding of how to use Photoshop. Now, though, people simply need access to platforms like Midjourney, a reasonable understanding of the kind of prompts needed to get the best result, and an idea.
It is no longer just images that are creating misinformation. This trend is starting to appear in the music world now as well, in fact, some of the biggest AI controversies are emerging in the music industry. The proliferation of ChatGPT has seen people using it to write lyrics in the style of major artists. From there, people have been able to use the growing AI-generated text-to-speech, or AI-powered audio to audio to platforms to make songs that sound like particular artists. Both Eminem and Drake have recently seen AI-generated songs released, copying their style. Neither artist has welcomed this, with subsequent lawsuits directed at the ‘creators’ of these AI versions, and versions of the songs being pulled from all streaming and social media platforms.
The biggest threat to creativity is the way AI is being trained in world and pop culture. This is what helped Midjourney create an image of the Pope in a puffer jacket. This is what helped ChatGPT to write lyrics in the style of certain artists, and how audio tools can replicate the sound of these artists. All of this leads to the question of intellectual property and ownership.
For visual artists, there has been an outcry to stop text-to-image platforms from using these artists’ designs in machine learning. These images are often taken without any recognition, or financial reward for the artists who have had their creations used to build the AI models. This is further compounded by users of text-to-image platforms, who can ask for an image in the style of almost any artist. This is dangerously close to plagiarism if it isn’t already.
While some artists are beginning to take legal action to block this, perhaps the highest profile is Getty Images. Getty is currently suing Stability AI, the creators of the popular text-to-image AI platform Stable Diffusion, for ‘unlawfully’ scraping millions of images from their site to feed into the Stable Diffusion AI learning. The evidence is clearly obvious to many Stable Diffusion users. This is because Getty watermarks all their images, and the Stable Diffusion AI was trained to understand that the Getty watermark should exist on all images.
There are currently no regulations in place to address this, but there is some hope that companies might be starting to look to protect artists. Adobe recently announced they would be launching their own text-to-image platform, called Firefly. The vision of Adobe is to pay artists to submit their designs to Firefly to help the AI learn and develop. But the key difference with Firefly, compared to all other solutions, is that Adobe would look to pay artists for their designs and pay any time their designs were used for any output from Firefly. The amount artists would be paid is yet to be confirmed, but at least it is a step in the right direction for artists to protect their work.
While fear dominates many, there are, in fact, some artists who are in full support of AI and any likeness of them being used. Canadian musician, Grimes, recently tweeted that she supported anyone using AI to generate music with her voice. Grimes went so far as to say that she would split any royalties 50/50 on any successful AI-generated song using her voice. This royalty split is the same as her deal with anyone she collaborates with on any music she makes. Grimes is the first to embrace the homage AI can give, but only time will tell if she represents the future opinion or remains a minority.
Everything has evolved incredibly quickly in the AI space. While users are struggling to keep up with the changes and innovations, lawmakers and regulators are dramatically falling behind. AI is rapidly moving into creative spaces, lifting, shifting, and stealing intellectual property as it looks to understand and improve its own output. We are frantically approaching a precipice with AI and creativity. Will we let AI consume all artistic creations and recreate them at will, will we be able to protect artists, or at least provide them some level of financial reward for true creativity? Or will creativity become mass-produced through artificial intelligence?