Is Marketing Creativity At Risk? Ask Generative AI
Once the stuff of sci-fi movies, in recent months artificial technology (AI) technology has morphed from a driving force behind operational automation and smart decision-making into a new domain that more widely influences human instinct, creativity and interpretation. It’s now firmly placed at the center of the conversation. Microsoft Head of AI Platforms Eric Boyd said (paywall) it’s “touching almost everything that’s out there,” and the mix of excitement and worry is palpable.
The buzz around generative tools, such as ChatGPT, DALL-E 2 and Writer AI (to name a few), has meant 2023 is a breakthrough year for the public understanding of artificial intelligence. With Microsoft investing $10 billion in ChatGPT-maker OpenAI and BuzzFeed’s stock surging after it announced plans to use this same technology to enhance its content, generative AI will clearly play a significant role going forward.
Machine learning is far from new. For years it’s been used in our industry to help tailor content to specific audiences that’s more engaging to our customers. It’s also used to analyze customer data to create customized messages that truly resonate with our audiences and identify patterns and trends from previous campaign data that can be used to improve future work. For example, last August Meta released Advantage+, which uses AI to generate multiple adverts to meet the objectives of the marketer. By running tests of potential ads, it selects which one it thinks will be most effective. One customer who implemented the tool saw a 20%-30% increase (paywall) in revenue earned from their advertising campaign versus other campaigns on Meta not run through this tool.
The sudden burst of interest is happening now because these generative AI models offer the general public a chance to see the potential up close in plain sight. As generative AI is increasingly used to produce ideas on tricky topics and enrich human creativity, the marketing industry needs to be ready to both embrace these new tools and understand their limitations.
Speed up originality and augment the creative process.
If you ask ChatGPT to create a list of ideas for a marketing campaign, the answers are generally bland and obvious. Even if you plug the words “unique” or “creative” into the question, the answer often fails to be either. However, seeing these standard, boring responses makes the process of creating something truly novel surprisingly easier, as it helps you eliminate the obvious. In a sense, generative AI can tell you what not to do.
The various possibilities it produces in seconds are another example of one of AI’s merits: It doesn’t moan about dull, repetitive tasks. It’s also unashamed to generate basic outputs that humans may not view as outside the box but are nonetheless useful. By using it to unpick the bad ideas from the good, we can use AI as a springboard toward original, inspired ideas. For anyone intimidated by a blank sheet of paper, AI can ease writer’s block when facing a tight deadline. As we know, it’s easier to edit something than to start from scratch.
This technology proves it’s the humans that have the vision, not the robots. Moving forward, high-level, truly valuable creativity will be the result of augmenting our processes with AI and maintaining emotion and human experience as the vital ingredients.
Be wary of hallucinations.
However, as the use of this technology grows, so does unwarranted confidence in the output. Yes, it is quick, but is it correct? Generative AI is only as good as its training data—which is almost two years out of date in the case of ChatGPT. As a result, generative AI is great at fabrication. This can be entirely harmless (and fun) when asking them to create a faux poem by Wordsworth or a fake painting by Rembrandt.
As Michael Wooldridge, director of foundational AI research at the Alan Turing Institute in London, commented, ChatGPT “doesn’t know what’s true or false. It doesn’t know about the world.” There’s a fundamental difference between generating answers and ideas and validating them. As its popularity rises, we must become more vigilant about its capacity to mislead. There is no getting around the need for fact-checking before you can put your faith in what the quick, obliging AI is telling you.
It’s still our job to look at the product and check if what we’re reading is true. Even ChatGPT’s prompt page warns “it may occasionally generate incorrect information.” It has no common sense—that’s where we come in. This necessity was seen recently when Google’s rival to ChatGPT, Bard AI, cost its parent company $100 billion in market value after it provided an incorrect answer in a promotional Twitter advert launching the product.
It’s a tool, not a replacement.
As Ethan Mollick of Wharton said, ChatGPT can be thought of as an “omniscient, eager-to-please intern who sometimes lies to you.” The conundrum of generative AI as it stands today is that it’s both hugely innovative and helps humans remain at the top of the pyramid. Slightly out-of-reach, unexpected ideas are the best side of the creative process—a side generative AI can’t yet reach.
Generative AI models are great to kick-start the imagination, to mine examples of what has been done before or to list industry conventions. We need to use AI as a foundation from which strategic and creative opportunities can grow.
The benefit for us marketers is its diversity. Generative AI can take in images, long-form text, emails, social media content, voice recordings and structured data. From this, it outputs new content, translations, sentiment analysis and summaries, among other forms of content. After these have been generated, they need to be carefully evaluated and edited by a human. These different elements can then form the building blocks of campaigns and applications in our industry—all produced far faster than if we started from scratch.
Undoubtedly there will come a time when these models create content that’s reliable, not just plausible. But until then, we need to use our professional experience to recognize AI’s shortcomings. Those who do so will be the most successful in putting this new kind of intelligence to work.
Metadata: Originally published on Forbes.com on 11 April 2023.
