How are human beings different from, let’s say, ChatGPT?
In some ways, the answer is obvious:
- Our intelligence isn’t “artificial” (whatever that means); and
- We’re not, in the words of ChatGPT’s self-definition, “highly versatile language model[s] that [have] undergone comprehensive training using deep learning techniques on a vast array of written sources.”
Chatbots aggregate and imitate—and so do we!
The important thing to know about ChatGPT and its brother/sister bots is that they don’t think.
Instead, they sort through the vast amount of “stuff” they’ve been fed, and aggregate (combine) it into familiar patterns that imitate the sound, rhythm, and style of their source materials.
In other words, they’re copying us.
If that sounds familiar, it’s because we human beings do the same thing.
We aggregate the language patterns we’ve been fed—by our families and friends, in school and work, and from the media—and imitate them when we want to make a point without having to think it through from scratch.
Don’t believe me? See if you can identify where each of these three human-generated examples came from:
- “It was a pleasure to join [people’s names] for this outstanding symposium.”
- “Your response is urgently needed before the end of our fiscal year.”
- “Residents in the region were urged to remain indoors, weeks after similarly dangerous air choked the Northeast.”
Answers: a LinkedIn entry; a fundraising letter; a New York Times story.
Our ability to synthesize and copy langauge is a tremendous time-saver (and probably why we’re so fascinated with AI).
But it stops being useful when we all start sounding like…chatbots.
Image by
Image by Mohamed Nohassi | Unsplash