TL;DR for operators LLM brainstorming should not be evaluated only by how good its individual ideas look. In a product-ideation experiment, GPT-4 generated ideas with higher average purchase intent than human participants and was seven times as likely to place an idea in the top decile. Yet its idea pools were also much more concentrated around similar regions of the solution space. ...
TL;DR for operators Generative AI can produce a pile of plausible options before a human team has finished developing one. The harder question begins afterward: which option is actually interesting, which is merely competent, and when does an unexpected result deserve to change the direction of the work? Ivan Magrin-Chagnolleau’s Can an AI System Be Creative? A Critical Perspective from Art and Engineering1 makes that gap visible in a haiku exercise. Across six batches, the AI produced 60 poems that followed the requested form, yet repeatedly converged on probable structures and was weak at distinguishing its strongest outputs. The paper’s argument is therefore not simply that AI lacks novelty. It separates rapid generation from the harder capabilities of judging creative value and recognizing when an accident is significant enough to revise the original objective. ...
Design review is where creativity usually goes to become awkward. One person likes the concept because it feels original. Another dislikes it because it looks impractical. A third praises the visual polish while quietly ignoring whether the idea solves the actual problem. Then someone asks whether the AI can “evaluate creativity”, and everyone pretends the word creativity has a stable meaning. Excellent. Very efficient. ...
TL;DR for operators Creative AI systems usually fail in a painfully familiar way: ask for ten ideas, and by idea four the model is politely repainting the same wall. Change the temperature, give it a persona, ask a panel of agents to “debate,” and the system may sound busier, but the semantic spread often remains narrow. The paper behind this article argues that this is not merely a prompt-design inconvenience. It is a structural limitation of how LLMs are conditioned. ...