Block’s latest voice model strikingly mimics human speech.
Image Credits:screenshot via Revolution.Social (opens in a new window)
Block Launches Human-like Voice Model Dessa
Jack Dorsey’s company, Block, has made significant strides in the realm of voice technology with the recent release of its innovative voice model, Dessa. This model, developed in Block’s research lab, operates locally and is designed to mimic the nuances of human conversation, showcasing natural expressions such as laughter, pauses, and even cries. According to Block, Dessa has been trained entirely on human speech data and operates on a single NVIDIA GPU rack.
Understanding the Full-Duplex Functionality
Dessa is a full-duplex voice model, meaning it can engage in simultaneous speaking and listening, similar to OpenAI’s GPT-Live model launched earlier this year. Typically, voice models rely on large language models for information retrieval, but Block asserts that Dessa derives its knowledge from its conversational training data, allowing it to engage users in more meaningful and contextually relevant dialogues.
Jack Dorsey expressed the company’s belief that voice is the most natural medium for interacting with artificial intelligence. He stated, “We created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person… full of emotion, nuance, pauses, and a natural cadence.”
Challenges with Existing Voice Models
Block critiques current voice models, arguing that they often sound mechanical and can be annoying due to the way they process speech generation. Many existing systems, such as ElevenLabs v4 and Fish Audio Drama 3, employ voice tags to evoke emotional responses in their speech. In contrast, Block claims that Dessa inherently incorporates emotion into its design. However, it remains ambiguous whether this emotional delivery is influenced by predefined tags or developed through its training process.
“We believe interacting with current voice AIs can be frustrating,” Block stated in a recent blog post. “These systems convert voice into text and lose a significant amount of critical information that text alone cannot convey. The most advanced AIs cannot perceive emotional cues, which we consider unacceptable, leading to subpar user experiences and potential hazards.”
Demonstrating the Versatility of Dessa
To bolster their claims, the researchers at Block have shared numerous examples showcasing conversations with Dessa, where the voice assistant demonstrated varied emotional expressions, tones, and delivery styles. In one instance, the bot conveyed feeling tired from work, accurately reflecting the emotion in its tone. Other scenarios included moments showcasing emotional depth, such as when the bot expressed appreciation for the comfort offered by the user, almost reaching a breaking point in its voice.
Ragavan Thurairatnam, another notable figure in the company, emphasized the novelty of Dessa, stating, “We made the first AI model you can have a realistic conversation with that was trained entirely on speech! Dessa is a full-duplex speech-to-speech model that can run on consumer GPUs. This is a GPT-3 moment for speech.”
Accessibility and Future Testing
Block has indicated that Dessa can operate on consumer-grade hardware, such as the MacBook M4 Max or superior systems. However, the model is not publicly available yet. The company has released a call for safety researchers to test its functionalities, aiming to explore scenarios in which voice agents could potentially mislead users.
The Business Impact
As a finance-focused company, Block seeks to leverage Dessa not just for consumer interaction but also to enhance its business offerings. Square, a subsidiary of the company, has already integrated voice AI for handling orders. With a dedicated voice model like Dessa, Block can expand its range of applications, improving both business and consumer experiences while reducing overall costs.
By establishing a locally operating, emotionally responsive voice model, Block not only attends to the growing need for more natural conversational interfaces but also positions itself competitively in the tech landscape, providing advanced solutions to small businesses and beyond.
Conclusion
Block’s introduction of Dessa marks a pivotal moment in the evolution of voice interactions. By focusing on creating a model that understands the emotional nuances of communication, they are setting new standards for AI-human interactions. Through responsible testing and further exploration of this technology, Block may very well redefine how we engage with artificial intelligence in our daily lives and transactions, paving the way for more human-centric design in the future.
Thanks for reading. Please let us know your thoughts and ideas in the comment section down below.
Source link
#Blocks #voice #model #sounds #eerily #human
About The Author
