There are many ways to utilize AI, specifically NotebookLM and image generation (Nano Banana), in Sci-Fi Prototyping. While the world is overflowing with “ways to use AI,” the evolution of AI itself is so fast that these methods become obsolete quickly. What we deal with here is not a method where the result is determined solely by the performance of the AI, such as in image generation. I would like to consider methods to train the creative mind using AI.
Please see here for more on Sci-Fi Prototyping.
AI is good at “generating a large volume of not-very-interesting ideas.” However, it seems to struggle with producing interesting ideas that are “unexpected and extremely interesting.” In short SF stories, this “unexpected and extremely interesting” quality is crucial.
Another important point is that struggling to come up with interesting ideas is fun. Therefore, even if AI could produce interesting ideas, you shouldn’t let it do so in Sci-Fi Prototyping. There is no point in telling someone who enjoys drawing illustrations, “AI can draw it in an instant,” is there? Since Sci-Fi Prototyping is training in thinking, the role of AI should remain strictly that of an assistant.
In my Sci-Fi Prototyping, I conduct two types of workshops: “Reading Workshop (Divergence)” and “Writing Workshop (Convergence).” NotebookLM and image generation can be utilized in both of these.
Reading Workshop (Divergence)
Since workshop time is limited, I ask participants to prepare by engaging with SF beforehand. I (YASHIMA Yugen) present a list of recommended works to the participants.
Using NotebookLM individually in advance, you can quickly learn the overview of works in the reference list or your own favorite works. For works where the full text is available, detailed analysis is possible. There are numerous classic SF works out of copyright, such as those by H.G. Wells, Jules Verne, Mary Shelley, and Unno Juza (classic SF has significant influence and is very interesting). Even with recent works, if they are published on the web and the copyright holder does not prohibit it, there should be no problem performing analysis to understand the work. You can also analyze works based on your own interests.
After I have written the catalyst story, you can also generate images of the work to help expand the participants’ imagination. By having NotebookLM analyze the setting of the completed story and create detailed prompts, you can generate illustrations and videos that are faithful to the setting. Literally, you can increase the resolution of your imagination.
Writing Workshop (Convergence Type)
It is possible to use NotebookLM in real-time even during the Writing Workshop. It can be used to aid in brainstorming ideas. However, given the strict time constraints, participants should write notes in text as much as possible during the workshop, and NotebookLM can be used when consolidating ideas from those notes after the workshop ends.
If there is time during the workshop, generate images based on the plot once it is done, and compare the participants’ mental images with the generated images. A prompt is the verbalization of an image, and constitutes the descriptive work itself. For example, if you feel that generation isn’t working well, you can check what is missing from the idea. For instance, if there is absolutely no specification of age or appearance, a character different from what you expected will be generated (although it might not be very specific at the plot level). Conversely, looking at the images might inspire new ideas.
It is also important to have it challenge/critique whether the generated ideas are sufficiently interesting.
Also, you can have NotebookLM analyze the grounded story I write at the end, and use it in discussions to generate ideas on how to connect it to actual product and service development.
In either case, it can only be called a meaningful use if you use it to support and train your own thinking, rather than leaving everything to the AI.
