Alexander Smola, co-founder and CEO of Boson AI, discusses the path from voice agents to audiovisual AI avatars. He explains why natural conversation still poses a high bar for AI, despite recent advances, and outlines the technical tradeoffs in real-time voice, including audio tokenization, latency, model size, and inference cost. Smola also explores how adding vision raises the stakes for natural interaction and why emotional intelligence is critical for avatars that feel human. The conversation covers how agents can learn from real human feedback to move beyond impressive demos toward genuinely fluid interactions.
This summary was generated from show notes and public descriptions, not from a full transcript review. Details may contain inaccuracies.
Highlights
•
Natural conversation remains a high bar for voice AI
Even state-of-the-art voice systems are easily tripped up by small delays, awkward interruptions, or misaligned tone, breaking the conversational illusion.