“The future of computing is conversational…every interaction with a computer is already a conversation of sorts, just using buttons, keys, and pixels to translate human thoughts into machine-readable code.” -Mustafa Suleyman, Founder of Inflection AI
In this post, we will investigate:
How to talk to AI, aka “prompt engineering”
The best ways to get your desired output from Generative AI systems like ChatGPT
Benefits of being able to work with AI systems
How to Talk to AI
As we move toward a future with increasingly capable AI systems, it will be necessary for educators and students to be comfortable and well-trained in communicating with machines. Unfortunately, this isn’t quite as simple as a human conversation, considering that machines understand and learn differently than we do. Modern tools like ChatGPT and Bard have made this process much more manageable, akin to typing in a Google search request. However, to get the most out of AI’s functionality, there are some guidelines and tips that users should be aware of. Subsequently, a new field is emerging known as “prompt engineering.” This involves the art of communicating with artificial intelligence systems in a language that they understand. It consists of structuring requests in a format and sequence that the AI will comprehend and thus increases the probability of a successful output. In short, prompt engineering skills facilitate greater success with AI and should be taught in educational programs designed to prepare students for the future of work.
How to Get Your Desired Output from Generative AI Systems
While there is much to learn about AI and prompt engineering (much more than can be covered in this post), here are some general tips for getting started with ChatGPT and similar systems.
First, tell the AI system what you do want, not what you don’t want.
In general, telling an AI system what not to do tends to be much less effective than telling it what to do.
Be specific and detailed.
Commanding ChatGPT to “write a marketing plan” is not going to be nearly as valuable as instructing it to “Write a ten-page marketing plan about X product with the following main sections: Executive Summary, Unique Selling Proposition, Industry Overview, Target Market, etc.” In other words, be as descriptive as possible.
- Provide examples when applicable.
Providing specific examples of what a desired output should look like will always be more instructional for the AI than not doing so. It will result in a better outcome. However, given the project's requirements or deliverables, users may need to research and/or experiment with the input/format to use.
- Separate instructions and content using ###
When doing so, instructions go first, and content/context comes next in triple number symbols surrounding the text.
For example, instead of saying: “Summarize the benefits for an economy of increasing GDP,” an improved prompt/wording would be:
“Outline the following in a detailed bullet point list.
Text: ###Explain the benefits for an economy of increasing GDP###”
While the output may only differ modestly with basic information requests, more sophisticated and complex tasks will be met with greater success using this format.
Benefits of Working with AI
While the applications associated with AI systems are vast, having the ability to communicate and interact with AI meaningfully is just as important. As machine learning capabilities continue to grow, so will the need for people with the skills to extract the most from AI. Prompt engineering will undoubtedly reveal new purposes for humans and AI to collaborate further.
Currently, app store-style marketplaces are being discussed surrounding newfound and innovative uses of AI. Some have suggested that AI systems will soon operate independently in real-time with real users, making decisions with minimal input once trained. Who knows what the next phase of capable AI will look like? But one thing is sure: reliance between humans and AI will only increase and expand into the future.
How can you improve your results with AI by incorporating prompt engineering?
What benefits can you envision from learning about the tips shared above?
What is one task or application that you are personally hoping to automate using AI?
OpenAI. (n.d.). Best practices for prompt engineering with OpenAI API. Retrieved November 25, 2023, from https://help.openai.com/en/articles/6654000-best-practices-for-prompt-engineering-with-openai-api
Prompt Engineering Guide. (n.d.). Prompt Engineering Guide. Retrieved November 22, 2023, from https://www.promptingguide.ai/
Suleyman, M., & Bhaskar, M. (2023). The Coming Wave. Crown Publishing Group.