What happened
- Dessa is a full-duplex speech model that can run locally on a MacBook M4 Max.
- Dessa is trained entirely on human speech and uses emotion built into its model.
- Dessa can mimic emotions like tiredness, laughter, and crying in conversations.
Why it matters
Block's new voice model could revolutionize how humans interact with AI, offering a more natural and emotionally responsive experience. If successful, it could change the landscape of voice assistants and improve user engagement with AI systems.
The Elephant take
🐘 鼋 Block’s Dessa model is a bold attempt to make AI sound human, but it's unclear if the emotion is truly built-in or just simulated. The company claims it's a GPT3 moment for speech, but we'll have to wait and see if it's as impactful.
Who should care
- AI developers
- voice assistant users
- block company executives
What to do next
- Test the model for potential safety issues
- Evaluate its emotional responsiveness in real conversations
- Assess its performance on consumer-grade hardware
- Monitor its impact on user engagement with AI
Keep in mind
The model is not yet publicly available and may have safety concerns that need testing.