I recently showcased some prompt engineering techniques that I employ to obtain some pretty excellent results from generative AI services like ChatGPT and the GPT family of models. With a lot of practice, beginning with GPT-2 released in 2020, I’ve honed my skills. The group was impressed by the results generated by the prompts, especially as their own were not delivering satisfactory results.
At the end of the session, they posed an important question: “Could you share that prompt with us?”
In this situation, the group was composed of trusted colleagues and friends, so naturally, I agreed. However, the question emphasizes how people perceive large language model prompts and the need for a shift in thinking.
Prompts may appear as natural language, like a newsletter, office memo, or social media post, but they are more than just words.
Consider a meeting with a developer in which you ask them to share their product’s source code. Their response would likely be a polite but firm refusal, and depending on the context, an offer to sell their valuable intellectual property.
What is programming? What is code? Is it complex structures like R, Python, C, etc., that resemble this?
df <- read_csv(“data/analytics.04.17.2023.csv”) |>
clean_names() |>
distinct() |>
select(date,source,medium,campaign,sessions,conversions)
That’s computer code, providing instructions for a computer to achieve consistent, reliable outcomes.
What about this?
You are a marketing analyst. You know SQL, R, set theory, and tidyverse libraries and methods. You know marketing analytics, Google Analytics 4, BigQuery, and attribution modeling.
Your first task is to write code to import a CSV file using today’s date in the name, prepare the variable names to be compliant with best practice naming standards, ensure the rows of data are unique, and then subset the data into date, the source, medium, and campaign dimensions for sessions and conversions.
Is this just language? No, this is a computer program, software.
As a company, would you share your source code or Excel macros? Does your employer even allow you to share work products? Most likely, the answer is no, and you shouldn’t share trade secrets.
Prompts are code, software, and valuable intellectual property. They are not merely press releases or blog posts.
To determine what to share or not, apply this test: will the prompt save time, save money, or make money within our business context? If it meets any of these criteria, it’s wise not to share it.
Employment agreements may also impact prompt sharing. Review your agreement for intellectual property clauses to understand ownership and sharing restrictions. As a job seeker, carefully examine employment agreements and negotiate fair terms. Consult an attorney (not ChatGPT) for advice on protecting your intellectual property, especially if it saves time, saves money, or makes money.
In conclusion, prompts are not casual text to be casually shared. Treat them with care, consideration, and prudence, particularly if they save time, save money, or make money. Employers should establish clear policies on intellectual property treatment. Exercise caution before sharing prompts, just as you would before open-sourcing software your business created. You may inadvertently give away something valuable enough for others to pay for.
Master the Art of Prompt Engineering for Large Language Models: A Comprehensive Guide