Interview, Fireside Chat
A fireside Chat with Mati Staniszewski, CEO & Co-Founder of Eleven Labs ft. Balderton Capital
- Solves the Turing test for voice conversations in most settings within the current year and deploys a true duplex model that is reliable within the same timeframe.
- Anticipates that all different types of content will become available in other languages over the next five to ten years as part of a shift from static text to audio-based voice content.
- Projects voice agents will handle website navigation, scheduling, and appointment management, with 80% of healthcare use cases already in a production stage where users prefer AI voices.
- Plans to expand the voice marketplace to include over 5,000 community-created voices, aiming to distribute more than $5 million in compensation to enable a full range of vocal options.
- Expects reliability improvements before tackling use cases involving real-world actions, while introducing cascading verification mechanisms to distinguish between real human speech, approved AI speech, and unverified AI speech.
- Envisions the emergence of new media and entertainment experiences, including personalized interactive gaming characters and personal tutor voice agents capable of remembering user history to optimize learning.
- Forecasts that future systems will primarily utilize AI voices, requiring the public to adapt to this reality, with voice interfaces extending to mobile devices for booking restaurants and accessing customer support.
- Identifies the next 12 months as a period where cascaded models remain the preferred approach for enterprise-grade use cases requiring high reliability and safeguards.
- Projects that duplex models will be more suitable for personal assistant and companion-type applications where some hallucination is acceptable, with the split between duplex and cascaded approaches beginning to converge within the next three years.
- Plans to continue investing in research sites to improve cascaded models and explore speech-to-speech capabilities, positioning voice agents as part of a brand's identity for near-future applications.