I am SUPER EXCITED to publish the 145th episode of the Weaviate Podcast with Alexis Ross, Manya Bansal, and Niloofar Mireshghallah from humans&!
What if AI was more human? Well, we are currently on a fast track towards the opposite! AI models are huge making advances in coding, math and are jev-ing their way through classification. But they don't teach, write, or perform more human tasks as well as they could be!
Persimmon is one of the most exciting releases in AI recently! It is a user model designed to predict how people respond in multi-turn conversations, including natural turn-taking and multi-person interaction. The goal is not to build a more human-sounding assistant. It is to match the distribution of human behavior, including pushback, confusion, and frustration.
This podcast discusses so many interesting topics around humans and AI. We start with an overview of Persimmon, then dive into solving the Turing Test (and why we should do so), to RL with non-verifiable rewards, theory of mind in AI, NVIDIA Nemotron 3 Ultra, and more!
Niloofar, Manya, and Alexis are three of the brightest scientists I've had the honor to host on the Weaviate Podcast. This was a super fun conversation, I learned so much personally, and I hope you find it useful as well!
YouTube: https://youtu.be/6IC9jkwiZBg Spotify: https://spotifycreators-web.app.link/e/GeFwTgRdC6b