🤖PolyAI CTO: Voice AI Hasn't Had Its ChatGPT Moment
TL;DR
PolyAI CTO Shawn Wen says full-duplex voice models exist, yet voice AI still lacks a ChatGPT moment. Otter's Alex Gay adds that transcription errors flow into every downstream action, so ASR accuracy is the real bottleneck.
PolyAI CTO Shawn Wen says full-duplex voice models exist, yet voice AI still lacks a ChatGPT moment. Otter's Alex Gay adds that transcription errors flow into every downstream action, so ASR accuracy is the real bottleneck.
Key Points
Wen: the next hurdle is making reasoning fast enough for natural conversation
ASR models often miss important keywords, which undermines context capture
Otter is building digital twins for meetings and wants to notify everyone in a chat when recording starts
Wen says enterprise calls should make clear that callers are talking to an AI
Remarks come from a HumanX conference session; investors are putting billions into voice AI startups
Why It Matters
Latency and keyword accuracy are the two numbers to test in any voice-agent pilot. Disclosure norms are also firming up before regulators force them.
Quick Facts
Comments
Be the first to comment
Enjoyed this article?
Get it daily. 7am. Free. Reads in 5 minutes.
Join 3,564 builders reading daily.