JUNE 10, 2023
BY MARK FULTON
A thought occurred to me today that I’m wondering why I didn’t think of it sooner.
Most people are using high school level and high-frequency collegiate vocabulay terms in their ChatGPT prompts, could we be missing out on an opportunity to make ChatGPT prompts that generate better results?
Could there be a performance improvement in GPT-4 responses when using vocabulary that is less commonly used in natural language but more frequently used in professional publications, papers, technical writing, and high-quality editorial publications?
Large Language Models are trained using a neural net on a massive amount of data. I can only assume that during the learning process associations are formed for high-level content and those specific words contained within it.
With this theory in hand, I set off to do some research and I’d like to share my results here.
Researching High-Level Vocabulary
I asked ChatGPT-4 the following…
/w=640,quality=80,fit=scale-down)
- Abstract
- Analysis
- Argument
- Assumption
- Bibliography
- Case study
- Citation
- Conclusion
- Context
- Data
- Debate
- Definition
- Discussion
- Evidence
- Experiment
- Findings
- Framework
- Hypothesis
- Implications
- Introduction
- Investigation
- Issue
- Journal
- Literature
- Methodology
- Objectives
- Perspective
- Procedure
- Publication
- Qualitative
- Quantitative
- Questionnaire
- Research
- Results
- Review
- Scope
- Statistics
- Study
- Survey
- Thesis
- Variables
- Insight
- Inference
- Observations
- Theory
- Interpretation
- Correlation
- Examination
- Relevance
- Inquiry
- Phenomenon
- Rationale
- Proposition
- Validation
- Limitations
- Significance
- Subjective
- Objective
- Elucidation
- Parameters
- Paradigm
- Categorization
- Synthesis
- References
- Parameters
- Editorial
- Appendix
- Scrutiny
- Protocol
- Replication
- Discrepancy
- Anomaly
- Experimentation
- Articulation
- Synopsis
- Determination
- Assertion
- Differentiation
- Annotation
- Proportion
- Manipulation
- Corroboration
- Complication
- Dissection
- Substantiation
- Generalization
- Classification
- Rectification
- Extrapolation
- Deduction
- Enumeration
- Exploration
- Permutation
- Correlation
- Specification
- Stipulation
- Comprehension
- Interpretation
- Documentation
- Elaboration
Now to Come Up With Some Useful Prompts
I then asked ChatGPT-4 to come up with some prompt ideas using these words.
/w=640,quality=80,fit=scale-down)
- "Generate a comprehensive analysis on the latest advancements in digital marketing strategies, and discuss their implications for small and medium enterprises."
- "Create a well-researched, in-depth article on the importance of data-driven decision making in contemporary business management. Include case studies to substantiate your points."
- "Elaborate on the theoretical framework behind customer relationship management (CRM) systems, including a detailed review of existing literature and discussion of potential future developments."
- "Draft an insightful business plan for a startup seeking to leverage artificial intelligence in the retail industry. The plan should include market research, competitor analysis, financial projections, and strategic planning."
- "Articulate the significance of sustainable business practices in the context of corporate social responsibility. Provide evidence-based arguments and draw conclusions about future trends."
- "Design a detailed methodology for implementing business process automation in a traditional manufacturing company. Discuss potential obstacles and propose solutions."
- "Create a thorough review of the implications of the Internet of Things (IoT) on supply chain management. Include real-life examples of successful implementations."
- "Generate an insightful and informative white paper on the effects of blockchain technology on the financial industry. Discuss the current state of affairs and extrapolate on potential future developments."
- "Present a comprehensive case study on how data analytics can be used to improve customer service in the hospitality industry."
- "Develop an engaging and educational tutorial on the use of machine learning algorithms for predicting market trends. Include step-by-step procedures and practical examples."
These prompts can be further tailored to the specific needs of your business, field, or research interests. The AI can then be used to generate a comprehensive, informed, and insightful piece of content based on the prompt.
As you can see the results are quite interesting and many seem like excellent leads to follow for creating even better prompts. I have highlighted my favorites above in bold.
I’ll admit I haven’t done extensive testing, yet. This is a theory based on my understanding of how LLMs are trained.
My assumption is that these words are associated with higher quality content and my hypothesis is that using these words within your prompts will produce better-quality output.
Optimizing Basic Prompts
Armed with these high-level vocabulary words and my 🔗 new tool AI tool SmartGPT which automatically enhances basic prompts, I’ll have no problem finding new and unique marketing research and prompts to test.
ℹ️Please note that the rudimentary prompt request used in the ChatGPT thread above is nothing like how SmartGPT functions.
SmartGPT is far and away more advanced than the above prompt.
Want to Learn Prompt Engineering?
I’m teaching everything I know in The Ultimate AI Course. There’s a heavy focus on prompt engineering and developing AI tools with no code applications.