Jack Dorsey's Block has announced an AI model called Dessa, trained from scratch on nothing but human conversation. The company shared sample chats that show the model sounding uncannily natural, and occasionally a little awkward, according to The Verge.
Dessa is a true speech-to-speech model, meaning everything it knows came from human speech rather than written text. On its website, the Dessa team says it can run locally on a MacBook Pro M4 Max or better, and it will be given to a small group of AI safety researchers for early access.
Why Block built Dessa
The Dessa team gives three reasons for the project: curiosity, better understanding between humans and AI, and safety. They compare the idea to how babies learn to talk without ever seeing a word written down, and say they wanted to see if a model could do the same. It can, they add, though it is very hard.
The team argues that speech is the most natural interface to any intelligence, but today's voice AI wastes it by flattening what we say into text and failing to keep up with human timing. They also warn that AI is gaining power and independence while remaining nearly deaf to the emotion in our voices, a safety problem Dessa is meant to help fix.
Dessa's quirky conversation demos
To show off the model, Block shared several conversations that highlight its unusual, human-like style. In one clip posted by Dessa lead Ragavan Thurairatnam, the model plays a guessing game over voice and mostly keeps a natural back and forth, though it initially mistakes the speaker for someone named Ronaldo.
Other examples on the Dessa website include a chat about Christopher Nolan's films that drifts into an unexpected tangent about Bryan Singer's Superman movies, and a person telling Dessa about a surprise trip to Paris, to which the model reportedly answers with what sounds like teary gratitude. There is also a conversation between OpenAI's GPT-Live-1 voice model and Dessa, where Dessa comes across as the more natural of the two.
