Learn to get the most from the system

- In the early 1990s, search engines were much different from today. Search engines like Lycos, Infoseek, or Alta Vista needed very specific prompts. If you searched for where to buy local plants, the results might include a list of trees, factories, flowers, or salads. That's because these early search engines didn't look at the context. They didn't know anything about you or your search history. Later companies like Google and Bing tried to fix that problem. They gathered as much data as they could. They created an online profile view to help the system better understand your preferences. Now, when you use a search engine, you get an entirely different result based on who you are. A manufacturing representative could get a list of local manufacturing plants. While a gardener might receive a list of nearby plant nurseries. You'll get different results from the same prompt. In many ways, generative AI systems are like early search engines. Even though their computing power vastly exceeds these early engines, generative AI systems still benefit from greater context. In the next five years, generative AI systems will likely build a much better profile of you. But for now, to get the most from these systems, you should focus on writing specialized prompts. In generative AI, this is called prompt engineering. Prompt engineering is about giving the system the context it needs to produce better results. Imagine wanting an AI chatbot like ChatGPT to help you choose between a Surface tablet and an iPad. You could type in a simple prompt, like, "What's the difference between a Surface tablet and an iPad?" The system would then give you a generic summary based on data vacuumed up into its foundation model, and that could be enough for most people. But to get something that's more specific, you'll need to create a persona. This is when you ask the chat system to look at the prompt from the perspective of a specific person. Change the prompt to,"Imagine that you're a creative professional, who spends most of the time using illustration software. What are the benefits of a Surface tablet over an iPad?" Here, you might get much more customized results. The prompt might mention software that works on one tablet and not the other. You can also try something called chain of thought prompting. It's here when you write down one task into several smaller tasks. You could start the prompt by saying, "Let's think about this step by step. Imagine that you're a creative professional that wants to work remotely. What are some of the best options?" Then when the system lists out the options, you can continue the chain of thought. You can even complete the prompt by asking the system to summarize its thinking into a table. Now, prompt engineering has gotten a lot of attention recently, and it's a good way to understand how these machines operate, but it's probably not a skill that you're going to be using in five to 10 years. Like early search engines, these systems will start to gather more data about you, but over the next few years, these simple techniques will often give you much better results when you're working with generative AI.