While a picture might speak a thousand words, it only takes a few words in a text box to generate a picture these days, one that might even be considered top notch artwork. Artificial intelligence (AI) is to thank for this, or perhaps to blame.
While artificial intelligence has long produced art, recent tools such as DALL-E 2, Midjourney, and Stable Diffusion, have given rise to an AI generated art boom that allows even the most uncreative among us to produce intricate, abstract, or lifelike pieces by merely entering a few words into a text box.
For some, the potential and possibilities of these AI tools to democratise craftsmanship and make creativity more accessible to everyone fills them with excitement, for others it fills them with dread and a moral panic about real artists being replaced by machines, an angle that is often pushed by the news media.
Dillah Zakbah, creative director and partner at BBH, says that while much has been written in the press from a position of AI replacing human talent, not much has been looked at or said about it from the point of view of using it as a tool.
"I think it's an exciting thing," says Zakbah. "Like paintbrushes were to the first oil painters, I think AI art is a new way of exploring new ways of thinking and execution. I don't think it's good enough for a final output yet, but I have used DALL.E a number of times for internal mock ups. It just makes it easier to explain what's in your head in a quick manner, visually."
Zakbah is not alone in her support of using AI art. Laurent Thevenet, head of creative technology, Publicis Groupe APAC & MEA, has been playing in the creative AI space for years. He believes that generative AI being a good thing or not really depends on individual perspectives and general inclination towards experimentation.
"I have found it very useful to push boundaries through 'accidental' (or unexpected) outputs generated during the exploration," says Thevenet. "It can open doors to new possibilities."
But while these AI tools are clearly helping creatives explore new possibilities, one might assume AI could also help with cutting costs and time, but this is not necessarily the case.
"In its current version, it’s not reducing art production costs as it still requires experienced creatives to direct the output," says Thevenet. "On top of that, it adds costs that didn’t exist before in the production process as the off-the-shelf services are not free and advanced usage requires renting GPU processing power from cloud platforms (hundreds to thousands of USD per month)."
Just gimmicky stuff? Can generative AI really be considered art?
Typing a few prompts into a text box for a machine to generate an image – can this truly be considered art? Many have questioned it. There is little in the way of human skill applied in this process, and you could argue that a machine produced image is irrefutably soulless.
But this is not stopping AI generated art from getting noticed or even winning awards. Last year, in a highly reported case, a video game designer in the US won first prize in a fine arts competition by entering an artwork made by artificial intelligence, which prompted stern condemnation from artists furious with the outcome.
"I don't really believe, in its current format, it can be considered art," says Guy Futcher, regional creative director, VCCP Singapore. "The technical definition of art is 'the expression or application of human creative skill and imagination' so are we suggesting that by having a human input the requests, we've fulfilled the human element of the process?"
Paul Reardon, chief creative officer of TBWA\Melbourne, was the creative lead on the Melbourne Writers Festival ‘The Art of Words’ campaign last year, for which he utilised the AI image generator Mid Journey.
"The current visuals systems produce stunning artwork, but they all require thoughtful creative prompting by a human to get them there," says Reardon. "It seems no publicly available AI system has mastered the ability to come up with a truly unique and compelling creative concept yet. The developments over the past months are truly mind blowing. But at this point, it’s still a creative tool. It’s not a creative replacement."
Thevenet believes there are two types of AI art. The first type is not refined and is usually the first output generated from a single text prompt.
"This is already used as lead images by click-bait websites that produce hundreds of pieces of content a day," says Thevenet. "In this case, this is not art. It's like buying a cheap stock photo."
The second type results from a multitude of iterations, eventually based on a custom dataset for refined direction. It would then be upscaled, sharpened and colour-corrected.
"This second type requires time and a thought process," says Thevenet. "This is art created in collaboration with an algorithm."
Thevenet adds that this second type—refined AI generation—takes time. "It’s not as simple as pressing a button once. Our APAC creatives that are now working with AI algorithms usually need a multitude of iterations to generate quality outputs. It takes as much time to do that as it would take to direct a junior talent. I feel like that effort deserves ownership."
Chris Gurney, group creative director APAC at Virtue and GPT3, believes that great art, regardless of its physical appearance, is designed to provoke or to make a statement. And that if AI-generated art can deliver this to an audience, then it should also be considered art.
"Generative AI art can be considered art. AI can be used as a tool to generate art that would be impossible to create without the technology. The possibilities are endless, and artists are using AI to explore new concepts and create art that is truly unique."
Is it ethical?
Many human artists are understandably concerned about their own futures as a result of these AI apps; after all, why would anyone pay for art when they could create it themselves? Intense discussions regarding the morality of AI-generated art have also resulted from these apps, as well as resistance from those who believe that they are just a high-tech version of plagiarism.
Artists have been expressing worries about the legality of AI image generators and how they can diminish the skill of drawing as part of the online campaign #NotoAIArt.
"The verdict on this is still out, because many of these generative AIs are only able to produce results because existing art (from real artists) has been fed in for them to ‘machine learn’," says Robin Lau, global digital strategist, Dentsu Solutions APAC.
"However because most AIs at this point have not reached a level of maturity where they are able to execute judgement and originality, many of these artworks contain pieces from their original sources, merging them with components or styles and strokes from other sources, some pieces even merge the signatures from the original artists. It really comes down to whether art that is derivative (yet the entire piece is original) is accepted, and where it is acceptable."
Gurney argues that art has always been influenced by development in technology, and that AI image generators are just the next inevitable step of evolution in art techniques.
"If we think about how humans moved from painting with their fingers to using pencils, paintbrushes – those could arguably be considered the latest technology during that time as well – and it's only served to expand the possibilities of what the artist can create, so long as the artist embraces instead of rejects, and learns to master and evolve alongside the development of the technology."
"Therefore AI art should not be seen as a threat to actual artists or designers," adds Gurney. "Instead, it should be seen as an opportunity for them to explore new ways of creating art and to inspire their own creative processes."
Copyright issues a blocker to adoption?
The ownership of AI generated art is currently something of a minefield and still largely an unresolved issue.
For agencies and brands, this grey area could present problems and legal issues that they might prefer to avoid by not using the technology at all.
"Clearly from a commercial point of view this is quite important for a brand to be able to protect the usage of its creative, to enable a distinctive and memorable campaign – so this could become a blocker for adoption," says Matt Law, founder of Move78.
"Firstly, the way these AI images generator systems work is by ingesting huge amounts of content, and then using that as the basis for the generative development," explains Law. "But all that content is scraped from the internet, or from shared data sets, and much of that training data is copyrighted by someone. Right now those people, copyright owners, do not benefit from any of the models trained on their data, and that is a legal minefield that is still to be addressed."
"Secondly, in terms of the IP in the content output, that will depend on the terms of service of the tool you are using, and the legal jurisdiction you reside in. However, it is typically the case that copyrighted works are created by human authors, and it is not clear where we will end up in terms of settled case law."
Will AI change the creative process?
So, will generative AI art be able to overcome its obstacles, namely rights issues and Sethical concerns around plagiarism and replacing humans, to be something that sticks around and even revolutionises the creative process?
"It already has. Watch the speed at which designers retouch an image, or editors smooth over a cut between scenes, or sound engineers remove background noise from a voice recording, and you’re seeing AI in action," says Simon Brock, general manager AU & executive creative director of Digitas. "My only advice to creatives is to be among the first to embrace it, so you can help guide AI’s influence over your own artform."
Chris Gurney, Virtue APAC, says that while the potential for AI to enhance the creative process is immense, and its use in the creative industry is likely to become more commonplace in the future, one of the biggest issues with AI systems at the moment is that the power (and therefore meaning-making) is concentrated in a tiny number of hands because of the sheer scale and investment required to build these models.
"The current systems tend to be Western and have an English-bias, which could potentially restrict and limit the stories being told and the ways of seeing that are being prioritised," says Gurney. "Diversity of thought, representation and ideas is critical to the creative process. For AI to enhance, not dilute, the creative process, effort is needed to ensure AI systems continue to learn from a broad base of inputs that is truly representative of the rich, interesting and diverse world we live in."
Zakbah believes that the way we view AI will evolve like how we viewed social media back in 2010.
"Who would have thought that somewhere to share your thoughts and photos of avocado toast would turn into a must have platform to be considered in your campaign plans?," says Zakbah. "As trend hunters, we need to be more accepting of futuristic ways of thinking and working. And as it gets more advanced, it is on our end to make it work for us."
And for others, like Thevenet, the future is already here and AI has already changed the creative process. "We have long talked about art, copy, and code being the new team required to answer briefs for the digital world. We are moving towards a Humans + Machines team structure so the creative process will be altered by that. The 'Humans' side of that team will be a combination of talents that can think above what the machines can do while still having the ability to control and plug all these creative AI solutions together."
"So, it is not a passing fad. It has been decades in the making," adds Thevenet. "Now, the hype and noise around it in our industry will probably go down as other technologies pop up but I don’t see it going away. It would be like saying that after Gutenberg invented the printing press, we would still go back to writing only."