I finally listened when a beta tester said my AI's chat responses felt too robotic and scripted.

After they specifically pointed out that the model kept using the same three transition phrases like 'furthermore' and 'in conclusion' across 50 different test conversations, I rewrote the prompt to prioritize more varied, natural sentence flow, so has anyone else gotten feedback that made you overhaul a core part of your model's personality?

2 comments

2 Comments

dakota4791d ago

Overhauling the whole prompt for that seems like a big reaction. People expect some patterns from an AI anyway. It's not like it was giving wrong answers.

thea_carter15h ago

Wait, but isn't that the whole point? If people just expect the same patterns, then what's even the difference? I mean, getting a correct but totally robotic answer feels pointless to me. Maybe it's just me but I'd rather have something that tries to sound a bit more real, even if it stumbles sometimes.