Block unveils Dessa voice AI that sounds human in demos, drawing awe and scam warnings
In brief
- Dessa, Block says, was trained only on human speech and can listen while speaking.
- One Nvidia GB200 NVL72 trained Dessa, Block says; the model runs on M4 Max MacBook Pros or better.
- Block supplied the demos; no independent tests, benchmarks, model size or release date have been published.
- Changpeng Zhao (CZ) cited children's education; Justine Bateman and Shiv Aroor warned of deception and scams.
What Block says Dessa can do
Dessa was trained entirely on human speech rather than text, according to Block's announcement as reported by CoinDesk. It can listen while it's speaking, so it can handle interruptions, laughter and changes in tone. Most voice systems convert speech into text before they generate a spoken reply. Dessa doesn't. It processes audio directly.
In one demo, it laughed mid-sentence. In another it reacted to a surprise with what sounded like genuine emotion (Block says those conversations were recorded live, with nobody directing the model's responses).
Block said it trained the model from scratch on a single Nvidia GB200 NVL72 computing system. The company also said it's compact enough to run locally on a MacBook Pro with an M4 Max chip or better. Block operates Square, Cash App, the bitcoin wallet Bitkey and the bitcoin mining hardware unit Proto. Its AI research lab reports directly to Dorsey and includes former staff of Dessa, the Toronto-based AI startup that Square acquired in 2020.
“Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence,” Dorsey wrote on X
What hasn't been disclosed
The evidence so far comes entirely from Block.
The audio samples are company demonstrations, not independent tests. Block hasn't disclosed the model's size, benchmark results or a public release date. Dessa is still a research preview, and the company is giving a small group of AI safety researchers access to it.
Awe, an education pitch and scam warnings
X user Alexander Hoff said he'd never heard anything artificial that sounded so human, and user Hesamation called the demo sci-fi, CoinDesk reported. Binance founder Changpeng "CZ" Zhao suggested the model could be used in education to help children learn through conversation. He wrote that it might be very good for GiggleAcademy, his education project.
Other reactions were critical. Filmmaker Justine Bateman asked why AI voices need to sound human at all and argued that the realism could be used to mislead people. Journalist Shiv Aroor raised concerns about fraud and predicted a surge in scam calls.
Block focused on a different use. The company said speech-based AI could eventually get better at recognizing emotional cues such as distress and pain, which text transcripts can miss.
Frequently asked questions
What is Block's Dessa voice AI?
Dessa is a voice AI model from Block, the payments company led by Jack Dorsey. Block said it was trained entirely on human speech rather than text and can listen while speaking, so it handles interruptions, laughter and changes in tone. It processes audio directly instead of converting speech to text first.
Can the public use Dessa yet?
No. Dessa is a research preview, not a publicly available product. Block is giving a small group of AI safety researchers access and hasn't disclosed a public release date, the model's size or any benchmark results.
Why are some people worried about Dessa?
Filmmaker Justine Bateman questioned why AI voices need to sound human and argued that the realism could be used to mislead people. Journalist Shiv Aroor raised fraud concerns and predicted a surge in scam calls.
