Block's new AI model Dessa was trained entirely on human speech to catch emotions text-based models miss
decoder ·
Block, the parent company of Square and Cash App, has introduced Dessa, a new AI speech model trained exclusively on human audio data without any text inputs. Unlike traditional systems that convert speech to text first, Dessa processes audio directly, allowing it to detect nuances such as emotions, sarcasm, laughter, and pauses in real time. The model supports full-duplex communication, enabling simultaneous listening and speaking capabilities. It was developed using a single NVIDIA GB200 NVL72 and can run locally on a MacBook Pro M4 Max. Block created Dessa driven by curiosity and safety concerns, aiming to address how current voice AI often ignores emotional cues. The company is currently seeking AI security researchers to test the model for unexpected risks before any public release. This development follows previous attempts by competitors like Hume AI, Kyutai, and OpenAI to enhance emotional intelligence in voice interfaces.
decoder