Standfirst://Recognising the chasm between hype and reality

Like you, I was awed and a bit scared when OpenAI rolled out ChatGPT earlier this year. All at once generative AI burst onto the scene, making all sorts of promises (and prompting many to fear the loss of their jobs in the process). We were asked to picture a world in which generative AI would allow anyone with an idea but no skills to write a screenplay, along with the musical score to go with it. In fact, generative AI is one of the reasons why Hollywood writers are on strike today.
 
Closer to home, generative AI would replace what Paragon Digital Services teams do daily, setting up campaigns, analyzing results, optimizing audiences and targeting criteria in real time, as well as generating reports for clients.The challenge with these claims? To put it simply, they exceed the capabilities of generative AI. Before we go down a spiral of hysteria, let’s establish what generative AI actually is. As it happens, this is the perfect question for ChatGPT, and so I asked it:
 

 
The most important sentence in ChatGPT’s answer is “It uses deep learning models to generate content that mimics human-like patterns or styles.” Let’s dive into this a bit.
 
Generative AI, like all machine learning, must be trained on data. In the case of generative AI, that data is all the human ideas that have been created in the past. ChatGPT will predict what someone might say based on what people said in the past. And while it’s absolutely true generative AI can produce “unique and creative outputs,” it can’t guarantee that any of those outputs will be true or accurate as it has no way to verify its responses against external sources.
 
ChatGPT will be the first to say as much:
 

 
ChatGPT Can’t Keep Up with the Speed of Advertising
 
The other challenge with generative AI is that it can’t keep up with our fast-paced business environment. Let’s say you’re a brand manager for a home mortgage company, consider how much has changed since the pandemic years. In September 2021, mortgage rates were as low as 1%, and everyone, it seemed, was in the market for a new home, along with furniture to furnish them. It was a boom time.
 

 
What’s more, the Great Resignation was frontpage news, and every business, including warehouses, were offering signing bonuses to attract new employees. Now, less than two years later, we may or may not be in a global recession, mortgage rates are skyrocketing, and mass layoffs are the news of the day.
 
Marketers and advertisers must keep up with the speed of the market, and working with their teams, must be in a position to pivot as events in the world change. But generative AI is somewhat stuck in the past, as ChatGPT freely admits:
 

 
No one in marketing, product planning, advertising or just about any role should rely on data that was created prior to September 2021. ChatGPT is completely unaware of the changes in the media world that have occurred since its cutoff date. It’s ludicrous to assume it can provide guidance when planning ad campaigns that are relevant today.
 
To say that generative AI will replace human beings in campaign set up, monitoring, delivering and reporting on campaigns is equally ludicrous. Any strategy and report it delivers will be based on what people said in the past, and not on current conditions! That’s not to say generative AI will play no role in advertising. ChatGPT can help the ad creative side come up with three or four alternatives to ad copy, for instance. It can help copywriters distill a 75-word description to a 50-word one (although if you try this, expect to do a lot of editing).
 
The problem with the hype we hear about generative AI is that we’re expecting it to do what it hasn’t been designed to do. It can generate incremental improvements in ideas that already exist, to be sure. The giant leaps that represent true innovation or a truly novel idea, however, are still the domain of humans. We still need humans to create the germ of the idea, whether it’s identifying an audience to target, the best messages to engage them, the best places to reach them, and how best to interpret campaign results.
 
This is why Paragon Digital Services isn’t planning to replace any of our team members with seats of ChatGPT 4. And to be honest, I doubt OpenAI would recommend we so do.
 
AI Already Drives Much of 
 
Pundits tell us all the time that AI won’t eliminate work, it will simply change the way we work. We already live this reality within the digital advertising space. Google Search, programmatic, all of the tools we use daily leverage AI to execute campaigns at scale.
 
AI automates a great deal of analysis and decisioning, but AI still requires a human to supervise it, less it focuses on goals that are outside of the marketer’s objectives. This is why Paragon Digital Services teams still analyze results daily, and tweak targeting criteria and strategy on an as-needed basis to assure campaign KPs are met.
 
We are grateful that AI can assess millions of campaign impressions in a day, of course — we couldn’t scale digital advertising without it. But we must also be cognizant of the fact that not every decision that AI makes will be the right one for a campaign, and only a human can recognize when that mistake has been made and address it.
 
As an industry, we succeed with things like programmatic and dynamic creative optimization because we know the opportunities and limitations of AI, and adapt to them. We’ve allowed it to change the way we work for the better, and don’t expect it to do more than it was designed to do. If we want to use generative AI successfully, we must allow it to follow the same trajectory, and stop expecting it to do things it was never designed to do.
 
David Tyler is president, global sales and partnerships at Paragon Digital, a Dentsu International company