keep the overall contribution to the context window moderate as well as drawing on previous experience from in- context few-shot learning. Good Ideas Prompt + ”Here are some well received ideas for inspiration: ” Overall, we generated 100 ideas without providing examples of good ideas and another 100 after providing access to examples of good ideas. Prior work in other domains suggests that the text generated by LLMs is not distinguishable from that generated by humans (Brown et al., 2020). While we do not test this question in this study, our impression is that any particular idea generated by ChatGPT cannot easily be distinguished from those generated by our students. Do LLMs Enhance Productivity in Generating Ideas? The answer to this question is straightforward. ChatGPT-4 is very efficient at generating ideas. This question does not require much precision to answer. Two hundred ideas can be generated by one human interacting with ChatGPT-4 in about 15 minutes. A human working alone can generate about five ideas in 15 minutes (Girotra et al., 2010). Humans working in groups do even worse. In short, the productivity race between humans and ChatGPT is not even close. Still, the old saying that ideas are a dime a dozen is perhaps a tad optimistic. A professional working with ChatGPT-4 can generate ideas at a rate of about 800 ideas per hour. At a cost of USD 500 per hour of human effort, a figure representing an estimate of the fully loaded cost of a skilled professional, ideas are generated at a cost of about USD 0.63 each, or USD 7.50 (75 dimes) per dozen. At the time we used ChatGPT-4, the API fee for 800 ideas was about USD 20. For that same USD 500 per hour, a human working alone, without assistance from an LLM, only generates 20 ideas at a cost of roughly USD 25 each, hardly a dime a dozen. For the focused idea generation task itself, a human using ChatGPT-4 is thus about 40 times more productive than a human working alone. In prior work, (Kornish and Ulrich, 2011) found that a typical new-product innovation domain contains thousands of unique ideas, ranging from about 1300 ideas for narrow challenges (e.g., use of technology in the classroom) to 3000 for more open-ended challenges (e.g., new consumer products). These numbers are large enough that a human working alone or in a small group is unlikely to identify most of them. However, LLMs are so productive that a human working with an LLM might reasonably fully articulate nearly every idea in an opportunity space. That is, it may now be possible to identify essentially every idea that a very large group of individuals working in parallel might identify after working for a long time, say, days or weeks. Prior work (Kornish and Ulrich 2014, Girotra et al. 2010) showed that the idea generation process in humans is essentially stationary, so ideas 2901 - 3000 exhibit the same quality distribution as ideas 1-100.