Block Unveils Dessa: An AI Voice Model Trained Only on Human Speech That Listens While It Talks

Block Unveils Dessa: An AI Voice Model Trained Only on Human Speech That Listens While It Talks

Block, the payments company founded by Jack Dorsey, has unveiled Dessa, a voice AI model it says was trained from scratch on nothing but human conversation — no text and no pretrained models. Unlike most voice assistants, Dessa is full-duplex: it listens and talks at the same time, so it can cut in, backchannel with a quick yeah or got it, and react mid-sentence.

Because it learned from real conversations rather than written text, Block says Dessa hears and produces the things text pipelines throw away: emotion, sarcasm, laughter, sighs and pauses. The company says it trained the model on a single Nvidia GB200 NVL72 system, using hundreds of times fewer words than large language models need, and that it runs faster than real time locally on a MacBook Pro with an M4 Max chip or better.

Dessa is not publicly available yet. Early access is going first to a small group of AI safety researchers, with a public waitlist open, and Block has published no parameter count, dataset size or formal benchmark results — so independent experts cannot yet verify the claims. Block's argument is about safety as much as conversation: today's voice AI, it says, is nearly deaf to distress in a caller's voice, and an assistant that cannot hear emotion cannot respond to it properly.

Source: read more

Comments

Popular posts from this blog

Fairbairn's Walk-Off Field Goal Seals Wembley Thriller as Texans Edge Jaguars 34-31

Anthropic IPO: Public S-1 Filing Expected Today Ahead of $2 Trillion Listing

Netflix Beats Earnings but Slashes Ad Forecast — Stock Tumbles After Hours