Jack Dorsey’s Block made an AI model that talks to you like an awkward new friend

Jack Dorsey’s Block has taken a bold step into the realm of artificial intelligence with the introduction of Dessa, a conversational AI model that aims to replicate the nuances of human dialogue. Unlike many existing models that rely on vast datasets scraped from the internet, Dessa has been trained exclusively on human conversations, allowing it to engage users in a manner that feels more personal and relatable. The model's development was spearheaded by Ragavan Thurairatnam, who showcased its capabilities through a series of interactions that highlighted both the strengths and the occasional awkwardness inherent in its design. For instance, during a demonstration, Dessa mistakenly referred to a user as “Ronaldo,” showcasing the model's quirky personality traits that can emerge during conversations.

The goal behind Dessa is to create an AI that can function as a conversational partner, particularly in scenarios where users might feel isolated or in need of companionship. By focusing on speech-to-speech interactions, Block aims to bridge the gap between human and machine communication, making AI more accessible and engaging for everyday users. This approach marks a departure from traditional text-based models, which often struggle to capture the subtleties of spoken language, including tone, inflection, and pacing. As AI continues to evolve, Dessa represents a significant step towards creating more relatable and human-like interactions with technology.

Key facts

Who's involved

Block, Inc., the company founded by Jack Dorsey, is the primary entity behind the development of Dessa. The project is led by Ragavan Thurairatnam, who has been instrumental in shaping the model's conversational abilities. Block is known for its innovative approaches to technology, and Dessa is a continuation of its efforts to integrate AI into everyday life, particularly in ways that enhance human interaction.

Background

The landscape of conversational AI has seen rapid advancements over the past few years, with models like OpenAI's ChatGPT and Google's Bard setting high standards for natural language processing. However, these models often rely on vast datasets that include a wide variety of text sources, which can lead to responses that feel less personal or contextually aware. Dessa's unique approach of training solely on human conversations aims to address this gap by fostering a more intimate and relatable interaction style.

Previous AI models have struggled with the subtleties of human conversation, often resulting in stilted or robotic exchanges. Dessa, on the other hand, is designed to engage users in a manner that feels more like chatting with a friend, albeit one that may occasionally misinterpret social cues or context. This blend of familiarity and awkwardness is intentional, as it reflects the natural imperfections of human conversation, making the AI feel more approachable and less intimidating.

How to read the numbers

While Dessa's performance metrics have not been publicly disclosed, the focus on full-duplex speech-to-speech interaction suggests a significant advancement in conversational AI capabilities. This feature allows users to engage in back-and-forth dialogue without the limitations of traditional text-based models, potentially leading to richer and more meaningful interactions.

What you can do with it

• Engage in casual conversations to alleviate feelings of loneliness.
• Use Dessa as a practice partner for improving conversational skills.
• Explore its unique quirks and responses to better understand AI interactions.
• Integrate Dessa into applications aimed at enhancing user engagement and companionship.

What we're watching

As Block continues to refine Dessa, the next milestone will be the release of user feedback and performance evaluations. Observers are particularly interested in how users respond to the model's occasional awkwardness and whether this enhances or detracts from the overall experience. Additionally, the potential for integration into various applications, such as mental health support or social engagement tools, will be closely monitored.

The introduction of Dessa marks an exciting chapter in the evolution of conversational AI. As developers and users alike explore the boundaries of what AI can achieve in terms of human-like interaction, the implications for companionship and social engagement are profound. The journey of refining Dessa will not only shape its future but also influence the broader landscape of AI communication technologies, paving the way for more intuitive and relatable AI companions.