“Because I was born a believer. If you can beat him, he’s a legend. If you only have one hand, don’t just watch a marathon. First, a marathon.”
This text is part of an experiment by Wieden+Kennedy. The agency trained a neural network with seven years of work for Nike and let it generate its own text.
It was avant-garde three years ago; imagine where it can go today. Can natural language processing become a dominance for creative advertising? Can it exceed human creations when measured in output?
How does adland sit with AI right now?
Most sources agree that there is huge potential for AI and that its market in all sectors of the world is growing at a rapid pace. While AI has been a part of advertising technology for several years, from sentiment analysis to audience clustering technology, it now plays a more central role in the creative process.
Technologies such as IBM Watson’s Advertising Accelerator use AI to create multiple personalized digital advertisements for all kinds of media requirements and audiences. The power of technologies like this enable large-scale optimization in any industry, micro-targeting demographics, psychographics, purchase triggers, customer journeys, and various KPIs such as conversion, rate video viewing and application downloads.
A fully developed case study in this area can be found in the Ad Council’s Covid Vaccine Education Initiative, conducted in the United States in 2021. The goal of the project was to increase uptake of vaccines among the population, supported by IBM’s creative AI technologies on the optimization side. . IBM’s findings as the campaign progressed were that there were four key hurdles to overcome: the safety of the vaccines, the speed at which they were made, distrust of Congress, and other theories. of various conspiracies.
Based on these barriers to adoption, IBM’s technology was able to scale messaging at scale to convince audiences to get the jab. The campaign directed 39.6% more people to GetVaccineAnswers.org than the standard creative.
Similarly and in a completely different industry, Vanguard used Persado’s NLG AI platform to deliver personalized ads on LinkedIn. The financial services industry exists in a heavily regulated advertising environment, so uniqueness is an invaluable commodity for brands in this space. Persado was able to personalize Vanguard’s LinkedIn ads and test them on a large scale similar to IBM’s software, to optimize person-to-person messaging. In an industry where front lines move in meters not miles, Vanguard was able to increase conversion to 15% through the platform.
All very exciting stuff. The problem, however, is that these technologies are essentially doing high-speed iterative A/B testing, not creative development from scratch. As revolutionary as this optimization technology is, the promised land of an AI capable of creating an idea, the genesis of the strategic and creative process, is not yet there.
Where is it? The case for GPT-3 and DALL-E
OpenAI’s GPT-3 was considered a messianic moment for the AI industry. Trained on 570 GB of text, or just under 1 trillion words, and sourced from web pages deemed to be of sufficiently high linguistic quality, AI’s ability to produce compelling and relevant writing from a short human sentence was staggering.
The ability to generate new, interesting and consistent content instantly, but at incredible speed, sounds like a strategist’s and creative’s dream at first glance. However, there are pitfalls. The first serious drawback that GPT-3 faces is a result of the language source upon which its foundation rests: the Internet.
AI, a play staged over three nights in 2021 at the Young Vic, was based on the innovative production technique of letting the cast create a live script using GPT-3 and then run the results. While the AI’s ability to generate incredibly accurate, human-sounding text lends itself well to this structure, it has repeatedly cast one of the cast’s Middle Eastern actors, Waleed Akhtar, in roles. negative stereotypes, for example, a terrorist or a violent criminal. In this sense, GPT-3’s reliance on the Internet as the source of its language model creates serious problems; it effectively acts as a mirror of the Internet’s ugly underbelly.
The second and most common problem with using GPT-3 as an ideation tool is its uncontrolled randomness and the unpredictability of generated content. While some companies, in partnership with OpenAI’s API program, have developed software that smooths out the more chaotic GPT-3 outputs to focus on media and performance-oriented tasks (similar to IBM’s program) , purely creative experiments have generated revealing results.
Adweek experimented with Copysmith, a copy-generating site with GPT-3 integration, to produce ad ideas based on brand names alone — the results, while certainly inventive, weren’t necessarily promising.
AI Publicity Idea: “How about we do something with a parachute car chase?”
Brand: Lay’s Potato Crisps
AI publicity idea: “How about we experiment with a vending machine that only accepts your shadow as payment?”
AI advertising idea: “What if we did something with a giant red button that, when pressed, would announce news?”
AI programs like DALL-E 2, Stable Diffusion and Midjourney, which specialize in generating images based on text inputs, encounter the same problem. A 2022 project led by London-based agency 10 Days sought to use Midjourney to create AI-generated campaigns around 10 corporate brands, including Ray-Ban, KFC and GymShark. The only inputs were brand names and six gender-based words to feed the AI algorithm (such as “black” or “cinematic”). The results were both impressive and shocking.