Archive — US — techcrunch
Thinking Machines wants to build an AI that actually listens while it talks
Thinking Machines Lab, founded by ex-OpenAI CTO Mira Murati, announced a research model, TML-Interaction-Small, designed for 'full duplex' AI interactions that mimic natural conversations. It processes input and generates responses simultaneously, achieving a response time of 0.40 seconds. Although impressive, the model is currently not available to the public, with a limited preview expected soon.
Key points
- Thinking Machines Lab is founded by Mira Murati, former OpenAI CTO.
- The new model, TML-Interaction-Small, allows simultaneous processing and responding, known as 'full duplex'.
- It boasts a response time of 0.40 seconds, comparable to human conversation.
- The current model is a research preview and not yet available for public use.
- A limited research preview will be released in a few months, with broader availability later in the year.
Why it matters
- This innovation aligns with the 2025 IT trends focusing on enhanced AI interactivity and real-time processing.
- Improvements in natural language processing have been a key area of development, reflecting ongoing interest in making AI more conversational.
- The approach mirrors ongoing developments in conversational AI, indicating a shift from traditional input-output models to more interactive systems.
Thinking Machines' approach to AI interaction represents a significant step toward more natural and engaging human-AI communication, potentially reshaping user experiences in future conversational applications.