As ChatGPT and other generative AI tools have ballooned in popularity, many have pondered whether they are a match for some of the more human elements of work, such as creativity. After all, these tools have shown themselves to be reasonably adept at creating works of art, music, or literature.
A recent study from Harvard Business School explores whether AI could also lend itself to the kind of creative tasks we undertake in the workplace. The researchers were especially keen to understand how AI performed in the kind of open-ended problems without a clear solution that so often form a key part of the innovation process.
Creative process
The researchers compared the performance of ChatGPT to that of a group of people who were tasked with crowdsourcing the solution to a challenge. The analysis showed that both man and machine had strengths, with the humans typically coming up with novel suggestions whereas the machine was more inclined to go for more practical solutions. As has so often been the case in the past, the best outcomes tended to be when man and machine worked together.
The innovation process typically begins with brainstorming. Prior research suggests that crowdsourcing can be an effective way to generate initial ideas, but it’s often time-consuming and costly. Creative teams usually need to incentivize participants, wait for their input, and then sift through submissions to identify the most promising leads.
Many hope that generative AI can overcome these challenges as these tools can generate an unlimited number of ideas almost instantly. But how good are these ideas?
To explore this, the researchers asked people to brainstorm business ideas for a sustainable circular economy, where products are reused or recycled to create new ones. They posted a request on an online platform, offering $10 to participate and $1,000 for the best idea.
Generative ideas
They then asked ChatGPT for ideas along a similar line. They particularly asked it for ideas that would involve “sharing, leasing, reusing, repairing, refurbishing [or] recycling existing materials and products as long as possible.” The suggestions returned would then be ranked according to the environmental benefits, the potential profit from them, their feasibility, and also their uniqueness by a team of 300 evaluators, all of whom were experts in the circular economy.
The feedback was that the ideas created by the humans were generally more innovative and more “out of the box”. By contrast, the ideas produced by ChatGPT tended to err more on the feasible side. This perhaps makes sense as generative AI is trained on that which already exists.
For instance, one of the human participants suggested creating a system of interlocking bricks using waste plastic and foundry dust. This would essentially create a new construction material that would produce less air pollution. The evaluators commended the creativity of the proposal but were skeptical about how realistic it was.
One of the artificially generated responses, in contrast, was to convert food waste into biogas to use as a renewable energy source. It’s far from the most creative idea, but it’s one that could be easily implemented and start to generate a clear return on its investment.
“We were surprised at how powerful these technologies were,” the researchers explain, “especially in these early stages in the creative process.”
The ideal approach
The best approach, however, is to mirror Kasparov’s law and use AI alongside humans rather than either working independently. The researchers explain that humans should be constantly using ChatGPT to generate ideas and adapt prompts accordingly.
“We consistently achieved higher quality results when AI would come up with an idea and then we had an instruction that said: Make sure before you create your next idea, it’s different from all the ones before it,” they explain.
Adding more prompts boosted the creativity of ideas, leading to everything from waste-eating African flies to smart beverage containers that pay consumers for recycling. Based on these findings, researchers suggest a few tips for business leaders using AI to develop creative solutions.
First, asking the right questions is key. Companies should consider building an “AI-literate” workforce—one that understands what AI can and can’t do to generate the best ideas.
Second, be careful not to rely too much on AI. Over time, this could limit creativity, leading to smaller improvements instead of major breakthroughs.
Finally, think of generative AI as a partner. One approach is for humans to brainstorm ideas and then use AI to refine them, making them more valuable and practical. Another approach is for humans and AI to work together in cycles, constantly improving the ideas.
If companies can combine human ingenuity with machine practicality, then they can get the best of both worlds and start getting creative ideas with a good chance of successful implementation.
