TL;DR
Get hardware and tech essentials delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Block demonstrated Dessa, a voice AI model trained on human speech that can listen and speak at the same time and, according to the company, run locally on a sufficiently equipped MacBook Pro. The demos drew praise for their natural-sounding conversation and concerns about fraud; Block has not announced a public release date or disclosed model benchmarks.
Block has demonstrated Dessa, a voice AI model that the company says was trained only on human speech and can listen while speaking. The demonstrations prompted praise for its natural-sounding responses as well as concerns that humanlike AI voices could make scams and deception harder to recognize; Dessa is currently a research preview, not a publicly available product.
Block says Dessa processes audio directly rather than first converting speech into text and then generating a spoken reply. That approach allows it to respond to interruptions and produce features such as laughter, pauses and changes in tone, according to the company. Block said it trained the model from scratch using a single Nvidia GB200 NVL72 system.
The company says the model can run locally on a MacBook Pro with an M4 Max chip or better. Block has not published the model’s size, benchmark results or a public launch date. The audio clips circulating so far are company demonstrations, not independent evaluations.
Block is offering early access to a small group of AI safety researchers. CEO Jack Dorsey described the conversations as humanlike, writing on X: “Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.” In the company’s demonstrations, the model laughs during a sentence and responds to a surprise with an emotional-sounding reaction.
Humanlike Voices Raise Fraud Risks
Dessa’s reception highlights a tension for voice AI: more natural conversation may make tools easier to use, but it may also make it harder for listeners to tell whether they are speaking with a person or a model. That matters in phone calls and other voice interactions, where a convincing voice could be used to impersonate someone or gain a listener’s trust. The source material does not establish that Dessa has been used in scams, or that it can imitate a particular individual.
Some reactions focused on the potential benefits of more fluid voice interaction. Binance founder Changpeng Zhao suggested the technology could help with children’s education, saying it might be useful for his GiggleAcademy project because current AI voices had been a challenge. Other commenters raised concerns about people being misled. Those reactions describe possible uses and risks; they are not evidence that the model has been deployed in either setting.
Block says speech-based AI could eventually recognize emotional signals, including distress and pain, that might not come through in a text transcript. Whether that capability works reliably is not yet established by the demonstrations. Its effects will depend on how the system is tested, shared and used, as well as safeguards that have not been detailed publicly.
Products Worth Considering
Apple MacBook Pro 14-inch with M5 Max Chip (18-Core CPU, 20-Core GPU), 48GB RAM, 2TB SSD, Space Black
Apple MacBook Pro 2021 16‑inch with M1 Max, 32 GB RAM, 2 TB SSD, Space Gray (QWERTY English)
Apple MacBook Pro 16-inch Laptop with M5 Pro, 18-Core CPU, 20-Core GPU, 48GB Memory, 2TB SSD, Space Black
As an affiliate, we earn on qualifying purchases.
From Audio Research to Preview
Dessa comes from Block, the company behind Square, Cash App, Bitkey and Proto, its bitcoin mining hardware unit. Block’s AI research lab reports directly to Dorsey and includes former employees of Toronto-based AI startup Dessa, which Square acquired in 2020.
The company’s announcement describes a model designed to generate and interpret speech without relying on an intermediate text transcript. That differs from common voice systems that turn a speaker’s words into text before producing a response. Block’s account of training and local operation is company-provided; the available material does not include independent technical testing to verify performance across devices or situations.
Block has chosen a limited research preview rather than a broad release. The company says safety researchers will get early access while it evaluates risks. Social-media responses to the demonstrations have included admiration for the voice’s conversational quality and criticism of the potential for deceptive use, but those posts do not amount to a formal safety assessment.
“Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.”
— Jack Dorsey, Block CEO, writing on X
Performance and Release Details Pending
Block has not disclosed Dessa’s model size, benchmark results or public release date. The demonstrations remain company-produced samples, and the available reporting does not describe independent tests of the model’s conversational quality, safety or ability to run on the stated hardware.
It is also unclear what restrictions will apply during the research preview, how researchers will report findings, and what safeguards Block may add before any wider release. The company has described possible ability to recognize emotional cues, but the current evidence provided does not establish how accurately it can identify distress or pain. No confirmed misuse of Dessa is reported.
Safety Researchers Begin Review
The next stated step is for a small group of AI safety researchers to receive access and assess risks. Block has not named the researchers, published a review schedule or said when findings will be shared. Any account of risks or performance beyond the company demonstrations will depend on further testing and disclosure.
Readers should watch for details on independent evaluations, safeguards and access conditions, as well as whether Block sets a public release date. Until then, Dessa should be understood as a research preview: the company has shown what it says the model can do, but broad availability and independently verified results remain pending.
Key Questions
What is Block’s Dessa?
Dessa is a voice AI model that Block says was trained entirely on human speech and can listen while generating speech. The company’s demonstrations show it handling interruptions and producing laughter and changes in tone.
Can the public use Dessa now?
No public product release has been announced. Block says it is providing a research preview to a small group of AI safety researchers.
What hardware does Block say it needs?
Block says Dessa can run locally on a MacBook Pro with an M4 Max chip or better. Independent testing confirming that claim is not included in the available reporting.
Has Dessa been used to commit scams?
The reporting does not identify confirmed scams involving Dessa. Concerns about fraud are potential-risk warnings prompted by how humanlike the demonstrations sound, not evidence of actual misuse.
What information has Block not released?
Block has not disclosed the model’s size, benchmark results or public release date. The company has also not published independent evaluation results in the supplied material.
Source: rss
