Block unveils Dessa, a speech-only AI model trained entirely on human conversation
Jack Dorsey's Block says its new Dessa model learned to talk from human speech alone and can run locally on a high-end MacBook Pro, with early access going to AI safety researchers.
By News Media House Staff ·
Block, the company led by Jack Dorsey, has announced an AI model named Dessa that it says was trained "from scratch on nothing but human conversation." To demonstrate the system, the company has published recordings of exchanges between the model and people, some of which come across as slightly awkward or uncanny.
One sample was shared by Ragavan Thurairatnam, who leads the Dessa project. The back-and-forth in that exchange sounds mostly natural, although the model at first mistakes the speaker for someone called "Ronaldo."
The Dessa website hosts several more examples. In one, the conversation turns to Christopher Nolan's films, and Dessa veers into an unexpected tangent about Bryan Singer's Superman movies. In another, a person tells the model about taking it on a surprise trip to Paris, and the site says Dessa "then answers with what sounds like teary gratitude." A further clip pairs Dessa with OpenAI's GPT-Live-1 voice model, and in that exchange Dessa sounds more natural.
According to the Dessa team, the system is a "true speech-to-speech model, and everything it knows came from human speech." They also say it can run locally on a MacBook Pro M4 Max "or better."
The team lists three motivations for the project: "curiosity (aka fun), better understanding between humans and AI, and safety." On the first point, they draw a comparison with children: "Babies learn to talk without ever seeing a word written down. We wanted to see if a model could too. It can. It's just very hard."
The developers also argue that current voice assistants waste the potential of speech. In their view, speech is the most natural interface to any intelligence, yet today's voice AI "flattens what we say into text and can't keep up with human timing."
The safety argument centers on emotion. The team says AI systems keep gaining power and independence while remaining "nearly deaf to the emotion in our voices." They offer a scenario of someone in serious pain while a voice AI has no idea and carries on regardless, and describe Dessa as "a step toward fixing" that problem.
Block plans to give "a small group of AI safety researchers" early access to the model. The announcement did not set out a wider release schedule.
Source: The Verge
