There is a 1895 French short film called L'arrivée d'un train en gare de La Ciotat (The Arrival of a Train at La Ciotat Station in English) where a stationary camera was set up on a tripod and filmed a train, well, coming into the station. The story goes that some audiences at the time reacted strongly to it, screaming and instinctually leaning back in their seats as the train approached.

We hear these stories today and go, “oh, what simple people we were back then,” as we walk out of a $300 million movie that showed us spectacle like we’ve never seen before, only to go, “there were some plot holes, that was pretty mid.”

I can’t help but think about this story when I hear about people at Google insisting that Google’s LLM is sentient or that we’re on the verge of LLMs becoming generally intelligent in the way people are. I don’t deny these feelings may be real, but I also think we should try to zoom out a little and recognize that while these products are new and shiny today, they will soon (already?) become just another piece of technology and they’re absolutely not as “smart” as some people seem to think they are today.

We should banish “smart” from our vocabulary when talking about LLMs. They’re not smart, they’re tools that are useful in some ways and destructive in others, but we do ourselves a disservice when we call them smart.

It’s also good to remember that the people of generations past weren’t dumb, they just grew up in a different era, and the things that were new and wowed them at the time eventually became commonplace.