Your new best friend just arrived

Block’s AI Model: Dessa

Block presents Dessa as a speech-to-speech model trained exclusively on speech. Unlike voice assistants that convert speech to text and lose vocal information, Dessa aims to support more natural, interactive conversation.

Demonstrated Features

Dessa responds with conversational timing, including interruptions and interjections. It can maintain a conversational thread, remember details, laugh, and express emotion. The video demonstrates casual conversation, jokes, singing, and a yes-or-no guessing game.

Limitations

The presenter says Dessa can sound convincingly human but is not especially smart; its guesses can be obvious or incorrect. The demonstrations suggest more natural voice interaction, not human-level understanding.

Broader Implications

Chris Norlund warns that increasingly human-sounding AI could encourage people to form attachments to AI companions and intensify parasocial relationships. He also raises concerns about an “economy of attention and time” and about who shapes public discussion as AI and recommendation algorithms become more influential.

Human Experience and Culture

Norlund contrasts AI-generated interaction with personal experiences and human-made art. He argues that everyday memories and artists’ interpretations preserve cultural history and a sense of humanity.

Video Format and Sources

No tutorial is provided. The video offers informal commentary and showcases several product demonstrations.

Chris Norlund is the presenter and commentator. Other sources include Block’s description of Dessa and demonstration clips featuring the AI model.