Block's new voice model sounds eerily human (techcrunch.com)

Jack Dorsey’s Block has unveiled a new voice model named Dessa, developed in its research lab, which boasts a remarkably human-like quality in its speech, complete with emotional expressions such as laughter and pauses. This model operates locally on standard consumer hardware, such as a MacBook M4 Max, and utilizes a single Nvidia GPU rack. Unlike traditional speech models that rely on large language models for knowledge retrieval, Dessa’s insights are derived entirely from conversational data, enhancing its conversational depth and responsiveness. The significance of Dessa lies in its potential to transform user interactions with voice AI, addressing ongoing frustrations with current models that often lack emotional nuance. Block argues that existing systems overly simplify human communication by converting voice to text, stripping away vital emotional context. With its built-in emotional intelligence, Dessa could provide a more engaging and meaningful user experience, especially in sectors like finance where effective client communication is crucial. However, the model is currently not publicly accessible and Block is seeking safety researchers to ensure the technology is secure against misuse. This could position Block favorably in both business and consumer markets, potentially reshaping how voice AI is integrated into everyday interactions.