Weaviate Podcast #145: humans& team on making AI understand people
CShorten30 · x · 2026-09-22
Weaviate Podcast #145 hosts Alexis Ross, Manya Bansal and Niloofar Mireshghallah of humans&. Key points: AI models are racing ahead at coding and math but lag at human tasks like teaching and writing; they introduce Persimmon, a user model that predicts how people respond in multi-turn conversations including natural turn-taking.
More from AGI Musings
- Evolvent AI cofounder: China leads in pretraining architecture innovation, lags ~6 months on post-training data — vista8 · 2026-09-22
- Students need both tech-free reasoning spaces and AI-enabled creation, not consumption — import_jmr · 2026-09-22
- Anthropic CEO: If AI can crack Navier-Stokes, why is alignment intractable? — robleclerc · 2026-09-22
- The Lab That Learns: Connecting structure-aware models with reasoning AI to make every experiment smarter — AllThingsApx · 2026-09-22
- Yacine: your Linux dotfiles could just be words, with apps built by AI on demand — yacineMTB · 2026-09-22
- Moral Singularity: Unpredictable AI Values Bode Poorly for Humanity, Argues Faggella — danfaggella · 2026-09-22