Skip to main content
2h ago

Block Unveils Dessa: An AI Model with Awkward Yet Natural Conversations

You must listen to the conversations between humans and Block's new Dessa AI model to understand its capabilities.Jack Dorsey’s Block announced the la

2 min read9 views5 tags
Originally reported bytheverge

You must listen to the conversations between humans and Block's new Dessa AI model to understand its capabilities.

Jack Dorsey’s Block announced the launch of Dessa, an AI model trained entirely "from scratch on nothing but human conversation." To demonstrate the model's capabilities, Block shared examples of human-AI chats that are occasionally awkward and uncanny.

For instance, Ragavan Thurairatnam, the lead of the Dessa team, shared a specific interaction. While the dialogue flows naturally, the model initially confuses the speaker with someone named "Ronaldo."

A video demonstrates Dessa engaging in a voice-based guessing game:

The Dessa website highlights additional interactions, such as a discussion on Christopher Nolan films where the model unexpectedly pivots to Bryan Singer’s Superman movies. Another example involves a user sharing a surprise trip to Paris, receiving a response described as "teary gratitude." The site also compares Dessa to OpenAI's GPT-Live-1 voice model, where Dessa appears more natural. Clips from these conversations can be found in the accompanying video.

The Dessa team describes the model as a "true speech-to-speech model" where all knowledge originates from human speech. It is optimized to run locally on a MacBook Pro M4 Max or more powerful hardware.

The team states they developed Dessa for three reasons: curiosity (or fun), improving human-AI understanding, and safety.

Just as babies learn to speak without written words, the researchers wanted to see if a model could learn speech purely from audio. They found that it can, though achieving this is significantly difficult.

The team argues that speech is the most natural interface for intelligence. They claim current voice AI wastes this interface by flattening speech into text and failing to match human timing.

They further explain that because AI is gaining independence, it remains deaf to the emotions in our voices. This creates a safety issue where an AI might not recognize pain or distress. Dessa is presented as a step toward resolving this.

Finally, the Dessa team announced they will provide early access to the model for a select group of AI safety researchers.

#AI News#Block#Dessa#Speech-to-Speech#Voice AI
ES
Editorial StaffEditor

The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.

View all posts
Reader feedback

What did you think of this story?

User Comments

Filter:
No comments yet. Be the first to comment!
Continue reading
View all news