How to Use Generative AI, Part 2: Prompting

When we use generative AI, the quality of output is heavily influenced by the input provided. We think of it as a the strategy of good, better, and best prompting. At the heart of effective interaction with generative AI lies the art of crafting prompts that steer the AI's responses in desired directions. Generative AI creates new data based on the 
path it takes through its own training data and 
its own model so the direction, specificity, and detail matters. Ultimately, a prompt points the model to the right place in the data cosmos.

Good prompting begins with clarity and specificity, the basic yet crucial elements that set the foundation for any successful interaction with AI. A clear prompt ensures that the AI understands the task at hand, while specificity helps in narrowing down the AI's focus, leading to more accurate and relevant outputs. For instance, rather than asking for "a story," specifying "a science fiction story set in a post-apocalyptic world" gives the AI a clearer direction, resulting in a more focused creation.

Moving from good to better prompting involves a deeper understanding of the AI's capabilities and limitations. It requires not just asking the AI to create something but doing so in a way that aligns with how the AI processes information. This means framing prompts that not only specify what you want but also how you want it presented, considering the AI's unique way of generating content based on its training. For example, if seeking business advice, a better prompt would not only specify the industry but also the type of advice—strategic, operational, financial—thus directing the AI to leverage relevant parts of its training.

The best prompting moves beyond basic clarity and specificity, incorporating an understanding of both the desired outcome and the most effective way to engage with the AI to achieve it. This level means the prompter has a keen awareness of what they want and possesses the skill to navigate the AI's capabilities. It includes the use of more nuanced language, the anticipation of potential misinterpretations, and incorporates constraints that guide the AI's creative process. For example, asking for a marketing strategy might include specifying the target audience, desired tone, and key messaging points, along with any constraints like budget or channels.

Developing the skill to craft effective prompts is a critical aspect of working with generative AI. It's about understanding what you want, knowing how to articulate that desire in a way the machine understands, and strategically using the AI's strengths to your advantage. As you become more adept at prompting, the AI becomes a more powerful tool for creation, analysis, and exploration.

Here some of examples of good, better, best prompts:

Practice getting more specific and clear and don't be afraid to create long prompts that instruct the model to do multiple things. But be aware that the behavior of applications such as ChatGPT can be variable. Be prepared to try different tacks and change language approaches.