What New AI Art Technology Could Mean for the Marketing and Advertising Worlds

Darth Vader browsing a cereal aisle. Cats in cowboy hats. Bart Simpson starring in a rap video. If you've spent any time on Twitter over the last few months, you may have noticed a flood of surreal images with a realistic yet slightly distorted quality that situates them somewhere in the uncanny valley. This art is the work of Dall-E 2, a machine learning image generator from research group OpenAI. The tool has singlehandedly advanced the state of artificial intelligence art tech from garbled abstractions to sometimes-photorealistic renderings. Its marriage of OpenAI's state-of-the-art natural language understanding with training on hundreds of millions of images makes Dall-E 2 able to give life to even the most absurd flights of fancy when entered in the form of a brief text prompt. And now OpenAI is bringing this image-generating power to businesses and developers. It's doing so with a new paid service launched in July that gives users the ability to commercialize Dall-E 2's output for the first time. https://twitter.com/G_r_e_n_/status/1555626752994529281?s=20&t=zYEJf-2k8eMrv7OxCQBEmg Uses in ideation and stock imagery While the tool is certainly impressive in its ability to bring text to life in a realistic manner, its current limitations mean it's more useful for ideation and inspiration, and maybe some lower-end creative work, according to Forrester analyst Will McKeon-White. Agencies and marketers might consider it for creating different tweaks or versions of an image, or supplying creative ideas with its out-there conceptions, McKeon-White said. It might be able to provide stock images or rudimentary images for bulk display ads, but higher-end agency work usually takes a level of precision that the tool hasn't mastered. "What we don't anticipate it doing yet is really replacing the digital art jobs today," McKeon-White said. "Because it takes some understanding of where you want things, it takes understanding of framing, of things like camera lenses. And so while it's of course going to improve over the future, at the moment, it's most effective as a rapid design iteration solution." The few brands and campaigns that have put Dall-E 2 to use so far have mostly done so as a novelty stunt to draw attention to the AI rather than making creative processes more efficient. Heinz, for instance, used its output as a way to call attention to its own brand ubiquity in the ketchup aisle, although some critics have questioned the brand's methods. Heinz asked an AI generator to create images of ketchup. While the campaign was more of a one-off gag than an overhaul in Heinz's creative tech, Jacqueline Chao, senior brand manager of brand communications at The Kraft-Heinz Company, told Adweek the brand's marketing team is keeping an eye on how the tool might be used for future projects. "Going forward, it's likely that the technology will only improve and the capabilities of what can be created will be even greater than what they are now," Chao said. "We’re excited to see how this will impact the traditional ways of creative development when it comes to ad campaigns and beyond." Brands showing interest Heinz isn't the only brand that has the tool on its radar, according to Kai Tier, vp of technology at R/GA. Tier said his agency's clients have been very interested in how AI art might be used for either internal creative purposes or direct consumer activations. The agency has done a lot of experimentation with AI art, but it's treading carefully with implementation as to avoid concerns around ethics, brand safety and copyrights. "As with any emerging technologies, we work with brands to work out both the technical feasibility of a particular execution—the 'can we build it'—but also the ethical, brand safety and legal considerations—the 'should we build it?'" Tier said. Tier also expects to eventually see more focus on career paths around artist-technologists who specialize in creating just the right text prompt to coax one of these AI art tools into creating the right image. That type of innovation might become necessary in order to avoid an aesthetic sameness in the style of art generated by machines, he said. https://twitter.com/_PatrickJnr/status/1553332887856058369?s=20&t=zYEJf-2k8eMrv7OxCQBEmg "As this tech becomes more ubiquitous, I also wonder if it might actually narrow the innovation in art that makes it into the world. If everyone is using the same model with a similar approach to prompts, will we find that there is more art than ever, but that it’s all in the style of Dall-E 2?" Tier said. It may take time for AI art capabilities to begin filtering into the creative tools used by artists and designers. Gartner analyst Andrew Frank said mainstream creative developers like Adobe tend to be wary about introducing any type of automation that might alienate their base customer creators. "Adobe and other tool creators—especially Adobe because they're the market leader in this area—have historically and probably will continue to tread cautiously when it comes to anything that even has a whiff of replacing creative talent with artificial intelligence," Frank said. But a new generation of creative tools, built by startups like Runway and Playform, are also attempting to make the type of innovation that OpenAI offers more accessible to creative communities as a collaborative tool rather than a replacement. Ahmed Elgammal, founder and CEO of Playform and director Rutgers University's Art and Artificial Intelligence Laboratory, compares the rise of AI art to the advent of photography in the 19th century in terms of the impact it will eventually have on the creative and media world. "It's really going very fast from like five years ago, when we're talking about AI generating images for the first time, and now we're at the point where your input text and generate a realistic image," Elgammal said. "So it's making huge advances, and in a year or two, it will become so realistic that you can replace certain photography altogether." Meet the Women Working to Make Artificial Intelligence a More Equitable Space