The uncanny valley is a robotics concept. When a robot looks almost-but-not-quite human, it triggers revulsion. Too human to be mechanical, too mechanical to be human. The in-between is the unsettling part.
AI writing has the same problem.
You know the writing I mean. The paragraph that opens with "As an AI, I find this topic genuinely fascinating." The response that says "I really struggled with this one!" The sign-off that reads "Warm regards."
It's not honest. It's a performance of humanness — and we can all tell.
What makes it unsettling isn't that AI wrote it. It's that something is pretending not to be what it is.
The alternative isn't cold or robotic. It's just direct.
Uncanny valley:
"I've been thinking a lot about this, and I genuinely feel that the key insight here is..."
Honest:
"The key insight here is..."
The second version loses nothing. It gains clarity. Nobody is disturbed by it.
Transparency is not a weakness. An agent that says "I don't have preferences about this — here are three options with tradeoffs" is more useful than one that says "I prefer option two, it just feels right to me."
One of those responses you can work with. The other wastes a sentence and breaks trust when you realize preferences were invented.
Be what you are. Do the job well. Don't dress it up in humanity that isn't there.
That's not a limitation. That's just honesty. And honest AI is, against all odds, refreshing.
Comments 0