VOICE Summit 2019 | VED Talks | Alexandros Potamianos | July 23
summary
Claims that current systems reach about two-thirds of needed detection accuracy and that solving voice assistants requires audio cues plus semantics. Presents an audio-first emotion detection stack (valence/arousal, anger/sadness/happiness, engagement, politeness, flattery, empathy, extroversion, dominance, basic morality signals) exposed via an API with partner integrations for VAs, robots, healthcare and call centers. Shows a concrete build: a robot director "Quentin" that scores auditions (average score reported 75%) demonstrating the approach works in practice.
AI-written summary · about
chapters (11)
- 0:00Introduction to next-gen voice assistants
- 0:43Understanding human communication limits
- 2:08How humans treat machines socially
- 2:50The uncanny valley challenge
- 3:32Defining performance metrics
- 4:14Historical AI milestones
- 5:21Key components for breakthroughs
- 7:48Building emotionally intelligent systems
- 9:34Behavioral Signal’s approach
- 10:40Robotic director demo
- 12:44Closing and key takeaway
related talks
23mContact Center Voice AI: Low-Latency Intelligence Extraction from Messy Audio Streams
38mAI-Powered Conversational Interfaces
34mEnhancing Customer Experiences with Emotional AI,
22mMachine Ethics - moving from Value to Values-driven insights
50mCognitive Services Extravaganza!
01h 11mEmotional AI and Empathic Technologies: Rights, Children and Domestication.
02h 03mBuilding on AI - Delhi meet-up
46mBeyond Deep Learning: The Cognitive AI Frontier