Jack Dorsey’s Block made an AI model that talks to you like an awkward new friend
What happened
Jack Dorsey’s Block introduced Dessa, an AI conversation model trained exclusively on human dialogue from scratch. The company showcased Dessa by releasing transcripts of the AI chatting with people. These exchanges reveal a conversational flow that is natural at times but also awkward in ways real human conversations can be. For example, the AI initially mistakes the speaker’s name, calling them “Ronald,” which adds to the offbeat tone.
Why it matters
Dessa’s training approach is unusual because it focuses solely on human conversational data rather than including other text types. This makes it an interesting experiment in how AI can replicate natural dialogue patterns, including imperfections like awkward pauses or misidentifications. For operators and builders, it highlights the trade-off between smooth, polished chatbot responses and more human-like, sometimes imperfect interaction. Models like Dessa could push the market to prioritize “realistic” conversational AI that feels less scripted, which may open up new use cases in customer support or therapy where natural rapport matters.
What to watch next
Keep an eye on how Block plans to deploy Dessa beyond demo chats. Will this approach scale effectively in real-world applications, or will the quirks challenge user trust and satisfaction? Also watch whether competitors start focusing on training models specifically on conversational datasets as a way to differentiate in the crowded AI assistant market. How Dessa handles identity and context errors over time could influence development priorities around personalized AI agents.
AI Quick Briefs Editorial Desk