> ## Content Index > Fetch the complete content index at: https://bayesiantheory.com/llms.txt > Use this file to discover other available public pages before exploring further. # AI Bayesics #4: brainstorming, creativity and ideation - URL: https://bayesiantheory.com/ai-bayesics-4-brainstorming-creativity/ - Published: 2026-03-27T11:44:19.000Z - Updated: 2026-10-05T14:45:23.000Z - Description: Unlock AI's potential for brainstorming and supercharge your creativity. - Author: Gregg Bayes-Brown - Tags: AI Bayesics, cognitive, Training, AI, psychology, Coaching, AI enablement, AI training, Newsletter, #Migrated-1791211413550, #Import 2026-10-05 14:44 ![](https://storage.ghost.io/c/0b/06/0b06555c-e570-49f3-a2d1-3ff0d72f29c4/content/images/2026/10/1ed77665-66c8-4385-be3b-183be53c5b67_648x410.gif) AI Bayesics #4: Brainstorming, creativity and ideation Here’s a statement that will annoy every creative I know: Human creativity isn’t special. We dress it up as muses, lightning bolts, and tortured genius. But strip that away and creativity is a formula. A problem you need to solve, plus an external inspiration, divided by your perspective. ![](https://storage.ghost.io/c/0b/06/0b06555c-e570-49f3-a2d1-3ff0d72f29c4/content/images/2026/10/2733e150-de87-4843-a672-617fcbf7f5e0_1142x473.png) The problem is yours. The perspective is yours. But the inspiration comes from outside you: a conversation, a book, a random Tuesday afternoon shower thought that drops the penny you’ve been waiting for. Innovation is the same equation multiplied. When designing the lightbulb, Thomas Edison tested thousands of potential filaments before settling on carbonised bamboo. In his own words: “Genius is one percent inspiration and ninety-nine percent perspiration” Take the creativity formula, multiply it by perspiration, and you get innovation. ![](https://storage.ghost.io/c/0b/06/0b06555c-e570-49f3-a2d1-3ff0d72f29c4/content/images/2026/10/29d82fc8-9a69-41c1-8c1c-37af2841c902_1191x358.png) Breakthroughs don’t come from one brilliant idea. They come from thousands of ideas narrowed to a single brilliant one. AI is good at both sides of that equation. In a University of Montana study using the Torrance Tests of Creative Thinking, GPT-4 scored in the top 1% for originality and fluency - outperforming 99% of the 2,700 human participants (Guzik et al., 2023). Meanwhile, AlphaGo’s Move 37 against Lee Sedol in 2016 - a move with a 1 in 10,000 chance of being played by a human, a move new to the 5,500-year history of Go - demonstrated that humanity had been living on a mere continent of Planet Go. That’s the creative power we’re working with. Before we get into it, open up your AI of choice, make sure thinking mode is on and[ your instructions are loaded.](https://bayesiantheory.com/ai-bayesics-2-communicating-effectively/) [Subscribe now](#/portal/signup) ## Brainstorming The single most important thing about brainstorming with AI: start with AI. Not after you’ve fleshed out an idea. Not after you’ve decided on an approach. The moment you’ve had a spark, boot up AI. > I’m thinking about \[an idea/task\]. I want to explore how we could collaborate on it. Work with me as my cognitive sparring partner and brainstorm it with me. You can push back on my thinking, and you can ask any question of me to be able to fully understand what it is that I’m trying to accomplish. If you’re stuck for ideas, [go back to week one](https://bayesiantheory.com/ai-bayesics-1-how-and-where-to-use/) and run the use cases prompt. Pick one and bring it here. You’re not directing traffic. You’re having an open-ended conversation with an intelligence that has access to more information than you could read in a thousand lifetimes. Set a rough goal, stay flexible on the route. The shift we’re making is from linear thinking — the kind that gets you from A to B in a straight line — to divergent and lateral thinking. This is AI’s wheelhouse. ## Lateral thinking During World War II, Vannevar Bush was the architect of US military innovation. When hiring scientists, he would ask them: “You are about to land at the dead of night in a rubber raft on a German-held coast. Your mission is to destroy a vital enemy wireless installation that is defended by armed guards, dogs, and searchlights. You can have any weapon you imagine. Describe that weapon.” The problem is the outpost. The inspiration is a weapon. The perspective is the weapon you create. You might also spot a familiar pattern: Role, Task, Context, and Output. We can use the same structure to unlock lateral moves with AI: > “You are \[role\]. You need to \[problem statement/context\]. You can use any \[task/inspiration\] you imagine. Describe that \[output/inspiration\].” What this does is force AI to reach across its entire knowledge base and connect things that don’t obviously belong together — which is exactly what lateral thinking is. You’re not asking for the obvious answer. You’re asking for the one nobody’s considered yet. ## Divergence and ideation Stanford’s Jeremy Utley published a book on ideation called Idea Flow just before ChatGPT came out. In it, he presents this ratio: 2000 ideas : 100 prototypes : 5 products : 1 success People are often protective over that one great idea they’ve had. However, your ‘one great idea’ probably isn’t a good one. To hedge against this, Utley says don’t just come up with one idea; come up with a tonne of them. This is divergence. Rather than finding one path from A to B, you’re looking for every path from A that could potentially lead to B. Despite ChatGPT eating his book for lunch, Utley identified that it was a powerhouse for ideation. And he’s right. Try it: > I’m working on \[project/problem\]. Give me 50 different approaches to this. Or, if you used the Vannevar Bush prompt above and got a response you like: > Give me five different variations of this. What you’re doing is using both the range of what AI knows and its probability layer — the mechanism that lets it generate genuinely different outputs from the same starting point rather than just rephrasing the same idea five times. ## Convergence Once you’ve got those ideas, you need to kill the weak ones. If what you have is a long list of short ideas, remix them yourself first. There might be half an approach you like on line seven and another half on line 14\. Put them together. Now you’ve got something stronger than either was alone. This is convergence – the yin to divergence’s yang. In Utley’s ratio, we go from 2,000 ideas down to 100 prototypes on gut feel alone. What we’re then doing with those prototypes is gathering data, which tells us which direction to push. This is Bayesian reasoning. We start with priors - beliefs or ideas we had going in - and then aggressively test them, gathering evidence that either strengthens or weakens each one. This is updating your priors. Say you’ve got a shortlist of approaches to a problem. Open a fresh chat for this — you want AI assessing your ideas without the context of the brainstorming conversation that generated them, because that context will bias it toward ideas it helped create. Then: > You are an expert in \[the domain you’re working in\]. We’ve been brainstorming \[project, plan, approach\] and want to narrow down our options. Assess and critique \[add your shortlist as context\] and present the strongest routes forward. For more ways to stress-test what survives, refer back to [AI Bayesics #3](https://bayesiantheory.com/ai-bayesics-3-cognitive-sparring/). ## Brainstorming reps Brainstorming is where I start on anything I’m building or making with AI. It’s the first conversation, every time. Over the next week, your reps are focused on starting early. Once a day, take one of the use cases you identified in week one, bring it to AI the moment the idea forms, and have a conversation about how to make it happen. Try the Vannevar Bush prompt on at least one of them. Use divergence to generate more options than you think you need, then use convergence to find the best one. Make mistakes, follow dead ends, and pay attention to the moments where AI takes you somewhere you genuinely wouldn’t have gone alone. Coming up on AI Bayesics, we’ll be covering the rest of the process: brainstorm, research, plan, implement, verify. But before we get to research, we’re tackling context — the information you provide to AI that turns the raw material into something worth a damn. [Subscribe](#/portal/signup) --- Thanks for reading. If you found this piece useful, please feel free to share with your friends and professional contacts. AI Bayesics publishes every Friday on the Bayesian Theory Substack and is available to free subscribers.