I would posit that "smarter" models like character.ai or the recent llama finetunes are potentially healthier because they are more deductive, and less submissi

I would posit that "smarter" models like character.ai or the recent llama finetunes are potentially healthier because they are more deductive, and less submissive. They will push back. And bots like this are a good temporary resource for socially isolated people in bad situations. This is still a worrying trend though. And while people have been whining about getting "sucked into" fiction for centuries, some of the commercial bots really are dangerous because they are designed to suck you in.