Thinking Machines wants to build an AI that actually listens while it talks
TechCrunch • • general
Key Points:
- Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, announced "interaction models" designed to enable AI to interrupt users, creating a more natural conversational flow.
- Unlike traditional AI models that alternate speaking and listening, these models process input and generate responses simultaneously, mimicking a phone call rather than a text exchange.
- The company's TML-Interaction-Small model operates in "full duplex" mode, responding in 0.40 seconds, which is comparable to natural human conversation speed and faster than similar models from OpenAI and Google.
- Currently, this technology is in a research preview phase, with a limited research preview expected in the coming months and a broader release planned later this year.