pov: you're tired of having low IQ while using AI

pov: you're tired of having low IQ while using AI Case Study: 4 Dimensional Prompting By Engineer Girlfriend (Written May 20, 2025) AI has uncovered a level of productivity that was unimaginable a couple of years ago. However, despite how powerful LLMs are, it is easy to quickly hit a wall, especially when generating abstract or creative content. This limitation is rarely due to the lack of capabilities from the AI model - the issue lies with the prompting. The following is a framework on how to increase the complexity of your prompting capabilities to extract increasing value from AI today. 1D Prompting This is where everyone begins: "Create a beautiful painting". Image creator Prompt Create a beautiful painting. A simple command input with a directly correlated output. The output is expected to be the final product imagined by the prompter, however it rarely is. The issue lies with the simplicity of the prompt. A simple prompt leads to a simple output. 1D prompting is useful for 1 dimensional tasks such as "Respond to this email" or "Summarize this article". 2D Prompting The next level of prompting is unlocked with technical syntax. Artists create better AI art, engineers generate more correct code, writers create higher quality excerpts. Today, professionals have an edge when using AI because they have already been trained in the manual craft pre-AI. They are aware of the industry lingo used, which is the data LLMs are trained on. Technical syntax takes you closer to the final line you desire. For example, a professional photographer can write a highly detailed prompt to generate an image most don't even know is possible with AI. AI Prompt subject an attractive young woman Detailed description of the subject with blue eyes lying face down on the bed, Photographer industry synthx description in the style of Raimond Gifs, light white and light amber, curvaged edges, the un-named aesthetic, timeless beauty, goodpose, sunrays shine upon it -- ai model specific syntax --no freckles --chaos 65 --ar 1.2 --profile yrurxpc2 -- stylize 750 --v 6.1 3D Prompting 3D prompting enters into the realm of "chain of thought". Most complex tasks are not executed by a single command. Even the human brain, arguably a much more complex "LLM", breaks down complex tasks into smaller modules. When AI approaches a complex task as a list of small tasks, the cumulative output is significantly higher quality than a simple prompt. At a simple level, chain of thought can be envisioned as a linked list. One task leads to the next, in a serial fashion. The complex task of "Write a research paper on the implications of AI art on the future of artists" could be broken down into the following simpler steps: 1. Research the capabilities of AI art generation. 2. Interview artists on their thoughts about AI art. 3. Draft a 3 page essay with high level concepts and citations. 4. Flesh out the essay with content. 5. Submit the essay for review from peers. However, a true chain of thought follows a tree structure of tasks. The first level of tasks can be broken into a sub tree of simpler tasks. By breaking down tasks until they are at the optimal atomic level for AI generation guarantees optimized outputs from the AI. The above linked list of tasks could be modified into the following tree structure: 1. Research the capabilities of AI art generation. a. Research the most popular AI models for art generation i. Generate a set of AI examples from each AI model 1. Generate a range of art types (traditional painting, graphic illustrations, photography). 2. Generate using a constant set of AI prompts to directly compare the outputs of each AI model. b. Research AI prompts used by professionals on AI prompt websites. i. Compare the outputs of AI art with different variations of prompt anatomy. 2. Interview artists on their thoughts about AI art. a. Interview a spectrum of AI artists to get differing opinions. i. Interview current students vs seasoned professionals. ii. Interview artists who use AI vs those who do not. b. Organize a facilitated debate between two artists who have different opinions. 3. Draft a 3 page essay with high level concepts and citations. a. Organize materials created from above steps. b. Layout the structure of the essay (introduction, question, hypothesis, research, conclusion). 4. Flesh out the essay with content. a. Write context and context for each section of the essay. b. Final review of the essay for grammar quality, typos, etc. 5. Submit the essay for review from peers. a. Collect a list of peers to get feedback from. b. Send the essay to peers for review. c. Summarize feedback from the peers. d. Edit the essay incorporating feedback. e. Finalize the essay. 4D Prompting The final dimension of prompting takes the above tree structure and turns it into a directed graph with the injection of AI. The AI injected acts as a bar raiser for the prompts that define each task. In 2D prompting, we discussed the necessity of technical syntax for high quality prompting. However, it is not possible for us humans to be aware of all existing technical syntax. As an engineer, I am not aware of industry lingo used for photography or fiction novel writers. This inhibits my ability to generate AI photography or AI writing. This is where the AI of prompting helps the prompter enhance human written prompts with AI. In a single injection node, the AI enhancer needs 3 pieces of information: 1. The original prompt used 2. The AI output created by the prompt 3. Feedback on the output (what's good & bad). Given this information, the AI enhancer edits the prompt to bring the output closer to the desired goal. At a simple level, this cycle of enhancement can be done manually by a human. However, true 4D prompting is achieved by using an "AI evaluator" to generate (3) Feedback on the output (what's good & bad). This enhancing cycle is then done on a loop until the AI evaluator deems the output acceptable. Warning: letting your AI loose on cycled chains can cause expensive AI bills! Proceed with caution. Ending notes You made it to the end! Thank you for reading and hope you got some value out of this. If you have any thoughts/feedback/questions, would love to hear them. Reach out to me on my Twitter (@enggirlfriend)