Conference Presentation, Keynote, Product Demonstration
Where We're Going We Don't Need Keyboards, Just Voice | Neil Zeghidour, Gradium | RAISE 2026
- Gradium, a three-year-old spin-off from Qtize, intends to leverage frontier research to develop the premier voice AI model for powering natural language interactions.
- Human-machine interaction is projected to shift from structured coding languages to natural language conversations, where speaking is anticipated to be three times faster than typing and to capture the nuance of thinking, hesitation, and mistakes.
- Voice agents are predicted to evolve from "chit-chat" models and closed-ended systems like 2011's Siri into agentic tools capable of accessing external data and performing complex operations, such as handling open-ended inquiries at drive-throughs.
- Current conversational voice latency of five to six seconds is expected to be reduced to human-level speeds of a quarter or fifth of a second through the adoption of full-duplex "bidirection model" technology, enabling interruption recovery and seamless flow.
- Robotics represents the next frontier for voice AI, requiring a complete architectural reset to address noisy environments, multiple concurrent speakers, and extended session durations beyond current phone-centric models.
- The scope of voice applications is expected to expand continuously from audiobooks into real-time customer support, enterprise tools, video game NPCs, and language learning.
- Voice AI is forecasted to eventually supersede the keyboard as the primary operating system medium via always-on, always-listening, asynchronous agents, with user input managed by a hierarchical agent structure overseeing sub-agents.
- To support these technological shifts and the opening of a San Francisco office, Gradium plans to recruit additional technical and go-to-market talent following a recent funding round.