A new study from researchers suggests that large language models (LLMs) like ChatGPT can generate higher quality product ideas much faster than humans. The findings could have major implications for innovation and product development.
The research compared 3 sets of product ideas for the college student market: 200 from students, 100 from ChatGPT without examples, and 100 from ChatGPT after providing high-quality examples. On average, ChatGPT’s ideas scored higher on consumer purchase intent surveys, indicating higher perceived quality. The best LLM ideas also surpassed the best human ideas.
Remarkably, ChatGPT achieved this higher quality while being vastly more productive. A single person can generate hundreds of decent ideas per hour with ChatGPT, whereas humans are limited to around 5 ideas per hour when working alone.
The combination of higher quality and productivity suggests LLMs could be transformative for innovation. The study’s authors argue LLMs’ lack of judgment is a strength here, increasing variance and allowing more extreme highs. With enough ideas, the best LLM ideas rise above humans’.
The findings imply ideation may shift from idea generation to evaluation/selection. Humans still appear better at assessing feasibility and selecting the most promising concepts. But LLMs can rapidly map the space of possibilities.
Product innovation is one area, but LLMs could enhance creativity in other domains like marketing, design, even art. Automating rote work is typical for AI, but excelling at creative ideation is counterintuitive.
Of course, care is required to avoid inappropriate or nonsensical outputs. And further research should examine similarity of LLM ideas. But with prompt engineering and example fine-tuning, LLMs may soon become indispensable creative partners.