
Artificial Intelligence is everywhere I am on the internet and January has been a month of learning to use it. This post includes my experiences with some of the artificial intelligence tools, including Rufus on Amazon, AI Overview in Google Search, and ChatGPT. The brief comment I would make about any of these tools is we must change how we interact to be effective. This isn’t your parents’ Google search any more. Without doubt, AI made my life better. We must ask better questions.
WordPress uses artificial intelligence on its help screen. The paradigm is simple. Define your role and frame what you want. For example, “I am a site admin and don’t have a lot of programming experience but I’d like to set up a new site and transfer my domain to it.” WordPress AI frames its response in terms of the request, often using the same language. This is ultra simple and important to every AI platform. That is a key learning point.
My main learning this month has been to ask any artificial intelligence tool better questions. Google and other search engines have trained us how to use them for decades. The old ways of entering a few related nouns or a simple phrase do not serve us as well going forward. Because AI has been trained on an enormous portion of human-written text, part of our queries must include minimal framing of questions. For example, I wanted to use a photograph as the basis for ChatGPT to render it in the style of Claude Monet impressionism with oil paints. It did a reasonable job of doing so. This kind of role-defining for our AI interface seems subtle at first, but more so it seems fundamental to the new approach needed to maximize our value.
Amazon sells stuff and uses an AI platform named Rufus. Even here query framing matters. The same type of role playing is important, yet roles are likely similar for everyone — we mainly visit Amazon to buy stuff. I asked Rufus, “Based on last year’s purchases, what are my buying patterns?” It listed Brand Loyalty, Shopping Style, and Household Profile. It identified me as someone who uses the account to shop for myself, incorrectly identifying me as a single-person household, which surprised me, since my spouse and I have linked accounts. Rufus also identified me as “price conscious but quality-focused” because I bought some Made In cookware. It also noted I am an active cook, based on buying Mexican oregano, canning jars and rings, and the aforementioned cookware. I likely used Rufus the least of the AI platforms mentioned.
With the broad database inherent in large language models like ChatGPT and Google’s AI Overview, our queries must include a way of paring potential answers down. To make our intent clear, state our goals for the tool, and most importantly set constraints. One of my favorite constraints is to write “I have 30 minutes to work on this so give me the top 3 findings,” or something similar. If I know something about what I am querying, I mention that as well. AI can provide its reasoning, and there’s no harm in asking for it.
I am still learning, yet with the long discussions I have with ChatGPT, the tool remembers what was previously said within a single chat. This is something I tend to forget when my follow up query is a week or two after the initial one. One evident thing is I need a better skill set when it comes to querying AI tools. Eventually, better AI queries will become part of a standard tool box for using artificial intelligence.










You must be logged in to post a comment.