Thinking Machines wants to build an AI that actually listens while it talks
The Mira Murati-founded startup says its ‘interaction models’ can process user input and generate output simultaneously, aiming for more natural full-duplex conversations.
Thinking Machines Lab introduced a research preview of ‘interaction models’ designed for full-duplex interaction—processing user input while generating responses. The company claims its TML-Interaction-Small responds in about 0.40 seconds, and says a limited research preview will open in the coming months with a wider release later this year.
Read the original reporting at TechCrunch.
