Get App
Download App Scanner
Scan to Download
Advertisement

Jack Dorsey's Block Unveils Dessa AI Model That Laughs, Chats With...A Few Awkward Moments | Listen In

Dessa is designed to preserve those elements, allowing conversations with pauses, changes in volume and laughter that can begin before a speaker finishes talking.

Jack Dorsey's Block Unveils Dessa AI Model That Laughs, Chats With...A Few Awkward Moments | Listen In
Jack Dorseys Block Unveils Dessa, A Voice AI Model Trained On Human Speech
Photo: Dessa/On X

Jack Dorsey's Block has announced Dessa, a voice artificial intelligence model trained on human conversation, sharing recordings that invite listeners to hear exchanges ranging from natural banter to awkward misunderstandings. The company says Dessa learned to hold conversations from audio alone, without text or pretrained models. It describes the system as a speech-to-speech model whose knowledge comes entirely from human speech.

Block argues that voice systems which convert speech into text discard information about emotion, timing and delivery. Dessa is designed to preserve those elements, allowing conversations with pauses, changes in volume and laughter that can begin before a speaker finishes talking.

The team describes Dessa as a full-duplex model, capable of the overlapping interaction that characterises human conversation. According to its website, the model decides what to say, when to speak and how to sound, including its voice and personality.

Film Chatter, Paris Plans And A Mistaken Identity

The demonstrations also reveal some unusual moments. In a recording shared by Dessa lead Ragavan Thurairatnam, the model initially mistakes its conversation partner for someone named Ronaldo, despite otherwise maintaining a mostly natural exchange.

Other examples include a discussion of Christopher Nolan's films that veers into an unexpected tangent about Bryan Singer's Superman films. In another recording, a person tells Dessa about a surprise trip to Paris, prompting a response the team describes as sounding tearfully grateful.

The website features a conversation between Dessa and OpenAI's GPT-Live-1 voice model.

Block says the clips are live recordings without anyone steering the model. Subtitles were added afterwards. The team also says Dessa can increase its volume when asked to speak up, a behaviour it says was not explicitly taught.

ALSO READ | 'AI Agents Could Outnumber Humans': Mike Walsh Explains India's Big Opportunity

Learning To Listen, With Safety

The project's stated motivations are curiosity, improving understanding between humans and AI, and safety. Its developers point to babies learning to speak without reading as an inspiration for testing whether a model could learn through listening alone.

They also argue that an AI system's inability to recognise distress in someone's voice could create safety problems, particularly as such systems gain more power and independence. Dessa can run locally on a MacBook Pro with an M4 Max chip or better, according to the team.

Block says a small group of AI safety researchers will receive early access. The team has also raised research questions about how speech-trained models acquire facts and languages, and how they learn from information provided during conversations. It presents those questions as areas for investigation into the model's capabilities.

ALSO READ | Bitchat Removed From India App Store: Apple Cites MeitY Demand Under Section 69A

Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.

Newsletters

Update Email
to get newsletters straight to your inbox
⚠️ Add your Email ID to receive Newsletters
Note: You will be signed up automatically after adding email

News for You

Set as Trusted Source
on Google Search
Add NDTV Profit As Google Preferred Source
Listen to the latest songs, only on JioSaavn.com