[kravens_last]

"Uncanny valley" is actually the opposite of what you mentioned.

The uncanny valley hypothesis predicts that an entity appearing almost human will risk eliciting cold, eerie feelings in viewers. For CGI and things that have visual queues, the minute differences in movement and facial muscle contractions contribute tremendously to our emotional responses to the "realness" of the thing.

ChatGPT has no such thing as it's virtually impossible to differentiate between an AI bot's writing and a human being's writing. It simply typically reads as concise and grammatically correct, if generic, university educated writing.

I actually WISH that there was a "tell" that allowed for an "uncanny valley" like experience, but chatGPT is as close to real as "real" is.

[smeej]

I disagree. I find it easy to tell the difference between chatGPT responses in conversations and human interactions. It does mimic humans, but in a stilted way that's specific and consistent.

I'm not sure if this could be connected to the way I learned to understand human behavior, by extensive deliberate pattern-matching with my organic intelligence. I was led to believe I had no hope of understanding people intuitively, so I should study the patterns in how humans interact and learn to mimic them as well as possible.

That made it easy to see who was reacting intuitively (i.e., acting like a human naturally) and who was doing what I was doing (i.e., "running 'human' in emulation mode," as Elon described having Asperger's).

From my vantage point, chatGPT is clearly and obviously not reacting like a human, which does make it feel uncanny to watch it try.