Transcript

There’s something about machine precision that’s unhuman. You can see it, hear it, and sense it. If something’s “too perfect” that really means it’s too precise, too machine-like.

The false negatives are also interesting. For example, personally, I sense this when I hear the world’s best a cappella groups perform. In unison, they hit every note so perfectly that, to me at least, it sounds like an AI. Isn’t that odd? It’s as if too much practice makes you seem less human.

I was watching a Texas Tech football game the other day. Their quarterback flashed a complex sequence of hand signals with such fluidity, it was obvious he was extremely well-rehearsed. This prompted the sports broadcasters to mention how he is often so well-prepared that people will comment that he’s like a robot. There it is again: so much practice that it seems less human.

We see the reverse of this is also true, that imperfections can make something seem human. When we look at a hand-written note, listen to a live amateur performance, look at a painting, or read a hand-painted sign, it’s the imperfections we’re picking up on that make it seem more human. If someone accidentally does something that a machine would never do - like forget to dot an i - we pick up on the humanness of that.

We humans are imperfect. In a sense, allowing imperfections to show through is a form of honesty. In another sense, our human imperfections are simply being filtered about by precise machines.

ChatGPT has a certain writing style that’s not quite human enough. As a result, the #3 app in the ChatGPT store (as of this writing) is a “humanizer” - an extension that attempts to make AI-generated writing sound more human. (As a funny aside, apparently one of the giveaways that some text has been generated by ChatGPT is the inclusion of an em dash. As a detail-oriented technologist, I had previously been using the HTML escape code for em dash, &emdash;, manually in my writing because I like the way the em dash looks. However, all the other humans who don’t know code, won’t do that. Most people just use “-“, because it’s on the QUERTY keyboard. I have since stopped using the em dash in my writing, because I don’t want my writing to be GPT-like.)

Making an AI seem more human turns out to be a difficult software problem to solve. For humans, however, all it requires is being yourself.