Re-learning How to Search in the Age of AI
Like everyone else, I’ve been watching and playing with AI. I’ve even decided to pay for Chat GPT because I feel like it adds enough utility for me and saves me enough time. One of the things that’s made obvious is that, at the very least, my search habits and skills are going to have to change tremendously.
As a bit of background, I’ve been working with the web since more or less the beginning. The first time I accessed the web was through a telnet session over modem to a university and web pages were rendered one line at a time. As a knowledge worker and technology person, I pretty early on learned how to use the tools available to find solutions to problems. I started with Usenet, then moved to chat groups, mailing list archives, and Deja news, which Google consolidated into a searchable web archive. Then, we had tools like Stack Overflow and Google which helped search these things. All this time, I became very talented at pulling things out of search engines, finding the right terms to use, and that sort of thing.
However, search had a very particular way of working and required you to be very careful about the terms you selected. If you couldn’t find what you were looking for, it did not help to be more descriptive about the concept you are trying to reach. Adding search terms after the first few rarely had any effect to change the results.
Now, enter large language models, AI, and ChatGPT (and soon Bing AI, Bard, etc). Suddenly, the skills for finding information are changing rapidly. It is actually easier for a large language model (LLM) to take as much context as possible and then predict what the best chain of response words would be to answer your question. The better and more detailed your question is, the better the answer you’re going to get.
This is a huge change from needing to limit yourself and your terms and power searching being about how to work with Google’s algorithms in a way where you could tease things out, and then in the end, have to do a bit more digging and research. Now, you can be more explicit about what you’re trying to find, understand, and get as results.
I can give a bit of a personal example here. I’ve been interested for quite a while now (a few years) in companies that are more stable and not chasing massive amounts of growth and thereby providing more dependable employment for people. I thought of these as “sustainable” companies. Due to my search skills, I kept stumbling on the term “sustainable” due to how popular it is regarding renewables and I never really could find better terms. It wasn’t until last fall when I was reading an obituary for Herman Daly that I discovered the term “Sustainable Development”. This more precise term opened up a whole new world of ideas in the search engines.
Contrast this experience with ChatGPT, which I just asked “What is it called when a company tries to be relatively stable over time and limit growth so that it doesn’t have to make drastic changes when markets go up and down?” And got the answer:
“The approach you’re describing is often referred to as a “steady state” or “steady growth” strategy. This involves deliberately limiting the rate of growth in order to maintain a stable level of operations and avoid the need for drastic changes in response to market fluctuations…”
While this is not exactly where I wanted to get in a single step, it is far, far closer than having to sift through results about renewable energy.
This is going to be a big sea change for people like me who were raised on low-context type search systems, but it’s also going to make information more accessible because people want to find information in a conversational way. It’s how we were brought up in terms of learning and interacting with each other. These new methods of interacting with these models, while there are pitfalls, are much more natural, and that’s a good thing. But it means people like me will need to retrain ourselves to be more aware when we’re interacting with computers. But that’s not a bad thing and I’m ready to learn.
