I Have Seen The Future!

… And It Kind of Sucks, Actually. Part 1.

Have you searched anything online recently? If you’ve used any of the biggest Search Engines you may have noticed a section at the top of the Results an “AI” “Generated” response, where links to what you were looking for used to be. It’s sleek, it portrays itself as helpful, and it’s 100% designed to only keep you from leaving the search engine page. It also has some quirks that are outright dangerous and pose several significant issues for businesses of all sizes as well as finding accurate and usable information. “Hold on,” you say as you noticed I put both AI and Generated in quotation marks as if I was being sarcastic. I was. Unfortunately before we dive into why I portray the technology in that way and why its’ corrosive to searching in general we have to dive into a little bit of surface level history and functioning of search engines. I know that sounds as fun as a root canal, but I’ll try my best to make it simple and fun!

Brief History:

The goal of searching online, is to find. Seems simple? Right. If I type into a search engine “Pizza Recipes” it wouldn’t be very helpful for me if I got results for Pool Installation companies 1000 miles from me instead of instructions on how to bake pizzas. In the early consumer-facing internet finding broadly what you were looking for was very hard, many of the “Search Engines” at the time operated far more like Phone Directories that you’d have to sift through to get to what you were searching for. It was becoming very clear that if there wasn’t some sort of procedure to connect people to what the were looking for to the information they wanted quickly and efficiently the experiment of the World Wide Web would suffer a slow death.

Knowing how the tools have changed over the years will help you understand what is happening now.

There’s a lot of yadda-yaddaing here but a handful of companies found a pretty good solution along with other innovations that made searching the internet easy and most importantly predictable. Search engines started to act less like directories and more like libraries themselves and they were the librarians. These companies made text copies of entire websites, and from there (along with other important processes we aren’t going to cover here) would make specific inferences between the content of the sites they had on file and what the user asked for. It cannot be overstated how revolutionary this was, it made the modern internet possible.

A lot of digital ink has been spilled recently over the “AI” tools these companies but very few outlets are really explaining what all of it actually means and assume a baseline amount of information that the average user likely does not have.

So What Is AI? What About LLMs?

Very basically “AI” as many know is Artificial Intelligence, which is a very wide variety of technologies from spell check, to mapping best routes for GPS systems, among other functions. The products being advertised generally are what are called “LLM” or “Large Language Models.” What do LLMs do? Well, they require a vast amount of human created content to train on. Training in this sense is simply copying and indexing information. Based on that content then can guess what the next best word is based on a probability model to attempt to answer a prompt. Is it smart? No, it has no understanding of what it is doing: it cannot perform basic analysis, evaluation of sources, weigh likeliness of accuracy and strength of the material it has been trained on, nor how to judge or parse veracity when two bits of its’ training material is in direct conflict. Nor can it determine there is a gap between what it is being asked about and a gap in training data.

In other words, it can just do its’ best guess on which regurgitation of data it has in simple conversational English is closest. It is just a parlor trick. But it has dangers that other outlets aren’t fully covering or haven’t fully pieced together coherently.

Which brings us to the main topic: Major Search Engines using LLMs under AI branding which will be cover in depth in the next post so stay tuned and contact us anytime if you need professional consulting in the changing search landscape.

Part 2 available here.