Jack Dorsey’s Block just showed off Dessa, a speech model out of its research lab that sounds a little too human. It laughs, pauses, sounds tired when it says it’s tired, and it talks and listens at the same time.
What we know
- Dessa is full-duplex speech-to-speech, so there’s no text step in the middle. Block’s argument is that turning your voice into text throws away how you feel.
- It was trained only on human speech, using a single NVIDIA GPU rack.
- Block says it runs locally on a consumer machine, like a MacBook M4 Max or better.
- No public access yet. Block is asking safety researchers to test it first.
My take
Local is the part I care about. A voice model you can run on your own machine, with no cloud in the loop, is a big deal for privacy and for the local AI crowd.
But the safety pause is the right call. A voice that sounds this human is a gift to scammers. Fake “grandkid in trouble” calls are already a thing. Talk to your family now about a code word, and never trust a voice alone for money or passwords.
Have a blessed day! And Happy Hacking!
