Nothing drives home the silliness of this early (LLM) AI frenzy better than a personal experience. Here is mine.
Over twenty years ago, when I was a student at Stanford and rather poor (think six digit student loan), I was trying to make a cheese fondue for my friends. Problem was that the original ingredients were so expensive that I had to make do with what I could find at an affordable price. So I used Edamer and and Jarlsberg instead of the horribly expensive Swiss cheeses that are usually used.
While my Swiss friends would look away in disgust, I found this amusing, and blogged about it under the title Wild West Cheese Fondue.
Fast forward. The other night, over dinner, I was talking with friends about cheese fondue and mentioned my concoction from way back then. I searched for it on the internet to show my original blog post to them. Instead I got this AI summary:
It turns out that my blog post is the only source for something like a wild west cheese fondue, and it was picked up by an AI, here Brave’s, as the authoritative source.
The authoritative tone is not me: That’s the answer engine and the rub of the story.
While I write about the wild west cheese fondue in plain English, the answer engine uses an authoritative tone that suggests that the wild west cheese fondue is some sort of well established recipe. It is not. As far as has been documented on the web, there was ever only one event in history, my friends and me in Tahoe, where a wild west cheese fondue was envisioned, instantiated, and consumed. Fortunately, we lived to tell the story.
Thankfully, whenever an AI presents something to me in full-on authoritative writing, I can now think back to this anecdote to remind myself that whatever the AI is telling me might just be a repeat of a single lonely voice, somewhere, once.
Leave a Reply