Opinions Artificial Intelligence

What if Artificial Intelligence Is the Creative One?

, by Colin MacArthur and Heather Yang
Thanks to modern Large Language Models, recent tests show that AI is able to generate as many ideas as experienced designers. Its development, therefore, will have an impact on what to design and how to do it

Recently, we interviewed a roundtable of leaders in product design about AI. One said, "AI won't change anything. Our job stays the same, just the stuff we use is different." Another shot back, "AI will change everything." Who's right?
AI isn't new. Computers have used algorithms (sets of instructions) to solve hard problems for years. As people and businesses started using computers more, they put more data in them. The internet made digital business and the algorithms behind them widespread and useful. Computers and algorithms have taken over more tasks from people over time, showing some basic artificial intelligence.
What's new now? Computers can do more thanks to better technology and recent programming innovations. "Large language models" (LLMs) can understand human questions and answer in sensible paragraphs. They make mistakes, but many are better than humans at answering questions and summarizing texts. Businesses are using LLMs to automate customer support and help their staff find information faster. Some companies even use them to write texts for ads and websites.
In the past, creative industry leaders often thought their work couldn't be done by computers. They believed making decisions about products was too tricky for computers. But as these marketing tasks show, it's not quite that simple.
LLMs speed up parts of the design process that humans normally do. They can come up with many ideas and solutions at once. Recent tests show they can generate as many ideas as skilled designers. Though designers still need to refine their ideas, they need fewer people to come up with the same number of ideas.
Also, these models will change the products designers work on. Making computers easy to use was a real challenge in the 1980s and 90s. Now, we're in a similar time for LLM-based products. Games, apps, medical devices, kitchen appliances, and electrical infrastructure may all use these models to make decisions or give information to users. Designers need to make these products meet peoples' needs, even as the technology underneath them changes.
Old ways of designing products are still important but need tweaking. Understanding customers is still key, but companies need to ask them new questions as people's preferences change with AI. Making prototypes and testing products will change too: designers may use LLMs to simulate other LLMs. Ultimately, designers will have to steer users away from LLM's weaknesses, while preserving their flexibility and strengths.
These models may even challenge what good digital design means. They aren't deterministic. Asking ChatGPT the same thing twice may get different answers. And they aren't transparent. It's hard to be certain about how they work inside and why. This is contrary to modern design principles, which focus on predictability and transparency. In the past, if the user couldn't predict what the system could do, and understand how it worked, designers were doing a bad job. Now, designers need to get users ready for unpredictability and uncertainty – or both designers and users will lose out on the benefits of LLMs.
In short, AI won't change everything for design, but it also won't change nothing. Although LLMs are the continuation of many years of work, they pose new challenges to both how design happens, and what needs to be designed.


Bocconi University
Department of Management and Technology


Bocconi University