AdvertiseDonate
LatestRevere resident: 97-year-old Maria O’Callaghan dies after six decades of marriage
Digital Newz Live · 126 Brookline Avenue, 2nd Floor, Boston, MA 02215

Block Trains Speech AI Model On Human Audio Alone

The company led by Jack Dorsey says Dessa learns language without reading text and runs locally on M4 Max Macs.

By Carolina Wood ·

Illustration for: Block Trains Speech AI Model On Human Audio Alone
Illustration generated for this article, not a news photograph.

Block has released an artificial intelligence model called Dessa that processes and generates speech without using written text. The company, led by Jack Dorsey, states the system was trained entirely on human conversation. The release includes audio samples of interactions that range from natural to slightly awkward.

Ragavan Thurairatnam, who leads the project, shared one demonstration clip. In that exchange, the model initially mistakes the speaker for a person named Ronaldo before the conversation flows more smoothly. The Dessa website hosts additional examples. One recording features a discussion about Christopher Nolan films that drifts into a tangent regarding Bryan Singer’s Superman movies.

Another sample involves a user describing a surprise trip to Paris. The site describes the model’s response as sounding like teary gratitude. A fourth clip pairs Dessa with OpenAI’s GPT-Live-1 voice model. In that specific interaction, the source notes that Dessa sounds more natural than in other examples.

The development team describes the system as a true speech-to-speech model. They claim every piece of knowledge within the model came from human speech. The team says the software can run locally on a MacBook Pro M4 Max or better hardware. They cite three reasons for the project: curiosity, improved human-AI understanding, and safety.

The team compares the approach to how babies learn to talk without seeing written words. They state that creating such a model is very difficult. The developers argue that current voice assistants flatten speech into text. They contend this process causes the AI to lose track of human timing. The safety argument focuses on the model’s ability to detect emotion in the voice.

The team notes that many AI systems remain nearly deaf to emotional cues. They describe a scenario where a person in pain talks to a voice assistant that does not realize the urgency. They position Dessa as a step toward fixing that disconnect. Block plans to grant early access to a small group of AI safety researchers. The announcement did not include a wider release schedule.

More Tech · Report a business · Got a Story?

Get Digital Newz Live in your inboxHeadlines from coast to coast. Free, a few times a week, unsubscribe any time.

Related stories

Get Digital Newz Live in your inboxHeadlines from coast to coast. Free, a few times a week, unsubscribe any time.