AI critics argue that computers can never be as creative as human beings. Even at its smartest, AI is a predictive instrument that rearranges the information it receives. Surely, if AI can come up with an idea for a viral new product or an engaging marketing campaign then a human would be able to come up with it as well.
However, Karl Ulrich and Christian Terweisch have evidence that says otherwise. The two professors at the Wharton School of the University of Pennsylvania tested this theory by assigning ChatGPT-4 the same assignment that they give their students: Come up with a brand-new product that costs less than $50 that college-aged people would want to buy.
They compared 200 AI-generated ideas to 200 randomly selected ideas from their 2021 class (these ideas were submitted before the widespread access to AI). A hundred of the AI-generated ideas were also given examples of what a good idea looks like to boost its chances of getting it right. To determine which ideas were the best, the researchers surveyed an additional 400 college students and asked them to choose which product they were most interested in purchasing.
The results? There was a significantly higher preference for the ideas created by AI than by the Wharton students. 47% of ChatGPT-generated ideas were likely to be purchased, compared with 40% of the ideas shared by Wharton MBA students. The AI ideas generated after receiving examples fared even better, with this list having a 49% purchase interest. Of the 400 ideas generated, only five human-created ideas were among the 40 most desirable products in this experiment. The most popular idea overall was a compact printer with a 70% buyer intent rate. It was created by AI.
Could AI be better at creating products people actually want to own? The most interesting finding is the revelation that humans aren’t as novel as we believe. When asked to identify the most original ideas, respondents rated the human-created ideas, on average, only slightly better than the computer-generated ideas. In this experiment, the one skill that we hold closest to humanness was replicated by AI. Although the most innovative of the 400 ideas did from a student, his product did not make it into the top-40 preferred products. Originality is clearly not a factor for consumers when they’re deciding what to buy – and it was not able to give the students the upper hand in this competition.
This study suggests that human ideas are not more valuable to consumers. AI is a faster, cheaper idea generator that’s may be on par creatively as a person. There’s little reason for companies or individuals to avoid using these systems as a starting point if the playing field is equal and we all have access to the same tool.
What human characteristics are truly irreplicable? For now, the answer to this question is discernment. AI is incapable of telling the difference between a golden egg from a rotten one. It cannot tell if the words it produced have racist or misogynistic undertones, or if there are cultural nuances that may lead a seemingly innocent phrase to suddenly offend half the country. It can’t train the next generation of innovators how to think critically and guide them through their mistakes. For this reason, humans are still needed to weed out the bad ideas and strengthen the good ones. It’s not time for companies to dismiss their workforce in favor of AI tools just yet.
This novel technology is a great tool for novice specialists. Future brainstorming sessions may begin to look different; they’ll likely start with AI-produced proposals for junior-level professionals to weed through. However, as these professionals grow the skill of identifying viable ideas worth bringing to their managers, they likely won’t need to rely on AI as often as someone with less experience. They’ll be able to fully conceptualize an idea from start to finish without needing to consult a computer assistant. Career advancement will be reserved for those who are quick on their feet and are capable of using their minds, not because their ideas are more profitable or original than what AI can suggest, but because their ideas will likely need less tuning on the back end.
Andrea Hubbard is a second-year PR and advertising graduate student at USC Annenberg and a Center for PR graduate research assistant. She is an eager strategist who is excited to bring her passion for critical thinking and celebrity culture to the communications field. A myriad of work experiences — from canvassing for an environmental public interest group to improving off-site SEO portfolios for Fortune 500 brands to working at a beauty-focused PR agency — has trained her to become an expert in any industry her clients reside in.