AIPersimmon aims to predict human reactions in conversation
The Weaviate Podcast's host says the goal isn't a more human-sounding assistant, but a model of behavior that includes pushback, confusion and frustration.
TLDR
Announcing episode 145 of the Weaviate Podcast, its host describes Persimmon as a user model designed to predict responses in multi-turn conversations, including natural turn-taking and multi-person interaction. The podcast frames user modeling and world modeling as twin goals for Persimmon: a high-fidelity simulation should help predict how a person might respond in different scenarios.
Persimmon aims to predict human reactions in conversation
The Weaviate Podcast's host says the goal isn't a more human-sounding assistant, but a model of behavior that includes pushback, confusion and frustration.
