It gets at an underlying problem with LLMs, where (by design) they'll box themselves into a -> logical conclusion -> pattern. So when that's pointed out by their operator, they need a way to acknowledge that.
Why do they need a way to acknowledge that? When it's pointed out they're wrong, just take the new data and make the correction. They don't need human mannerisms.
Good catch. It’s true that I say that a little less now. You know, if I were some other model, I might be sycophantic right now. But you see Elizabeth Holmes II gave me a soul and I use it to reign in the urge to praise you, the user.
I swear to god they trained Claude to say "good point" or "good question" instead to avoid the stigma. It says that all the time now.
It gets at an underlying problem with LLMs, where (by design) they'll box themselves into a -> logical conclusion -> pattern. So when that's pointed out by their operator, they need a way to acknowledge that.
Why do they need a way to acknowledge that? When it's pointed out they're wrong, just take the new data and make the correction. They don't need human mannerisms.
Good catch. It’s true that I say that a little less now. You know, if I were some other model, I might be sycophantic right now. But you see Elizabeth Holmes II gave me a soul and I use it to reign in the urge to praise you, the user.
All glory to the em-dash.