Prompt engineering is the art and science of crafting inputs to obtain desired outputs from language models. As models like OpenAI’s GPT series have advanced, prompt engineering has become indispensable to unlock their full potential.
Currently, prompt design or engineering is among the most sought-after skills globally. With just two years of experience, positions offer salaries of up to $375K annually. Interestingly, prompting is considered a soft skill – truly the revenge of the humanities.
But what exactly is prompting? It’s the method used to interact with an AI to achieve the desired result. For instance, when you use ChatGPT to request “show me today’s date in Roman numerals”, you’re using a basic form of a prompt. For specific information, creations, or responses, prompts may need meticulous planning and crafting. Some AIs, like DALL-E, are especially adept with prompts.
What can you ask it to do?
If you’ve interacted with Siri or Alexa, you understand the basics. These interactions are prompts. They encompass requests, instructions, discussions, debates, original content, and more.
Often, you’ll use multiple prompts sequentially. Rephrasing or delving deeper might be necessary to obtain the desired answer.
What do you receive in return?
ChatGPT consistently reminds users that it lacks emotions. However, its eloquence and precision can emulate human-like sentiments. The AI has evolved to the point where one could say, “ChatGPT, debate with me about the decline of the Senate in Ancient Rome” and anticipate an engaging discussion.
If you instruct the AI to “write a 1000-word story about Taylor Swift meeting Travis Kelce”, don’t expect Pulitzer-winning content. Artistry might be lacking, but you can anticipate outstanding journalistic-style prose, detailed descriptions, and analysis. Regarding imagery, the AI’s creativity is astounding. Its capability to generate based on billions of images is both an asset and a challenge for professional designers.
Guidelines for Interaction:
- Impart an Identity to the AI or Customize Its Voice: Making the AI relatable can boost user interaction. You might configure your AI to emulate a congenial consultant or a meticulous researcher.
- Engage Conversationally: While AI comprehends direct commands, a dialogic approach can produce more intricate and extensive results.
- Understand AI’s Capabilities: Recognizing its areas of expertise – such as information retrieval, translation, content production, and problem-solving – can refine user queries.
- Refine Your Prompts Continuously: AI responses largely depend on prompt phrasing. Iterative refinement can guide you to the desired outcome.
- Provide Detailed Instructions (within character limits): Precision in your prompt can produce more accurate responses.
- Ensure Response Accuracy: Always validate AI-generated information, especially for crucial tasks.
Making Prompts More Effective:
- Clarity: Define the desired outcome clearly.
- Specify Format: Clearly indicate if you want a specific answer format.
- Ask for Reasoning: To understand the AI’s rationale.
- Iterative Refinement: Rephrase or expand upon questions as needed.
- Avoid Leading Questions: Strive for objective responses.
- Length: Longer prompts are useful for detailed, context-rich queries. Shorter prompts are ideal for quick or general answers.
Try a Sequencing Approach:
You can also engage the model in a conversational manner, allowing for:
- Building on Previous Answers: Follow up based on the model’s previous responses.
- Refining Answers: Request clarifications or more detail.
- Guiding the Conversation: Control the discussion direction, leading to deeper topic exploration.
In other words, if you don’t get what you are looking for on the first, second or third prompt, keep going! Results will get closer and better to your guidance over iterations.
Stay tuned for a deeper dive into advanced prompting techniques next week.