AI news story
Mira Murati’s Thinking Machines Lab Introduces Interaction Models: A Native Multimodal Architecture for Real-Time Human-AI Collaboration
Thinking Machines Lab has introduced a research preview of TML-Interaction-Small, a 276B parameter Mixture-of-Experts model with 12B active parameters, built around a multi-stream, time-aligned micro-turn architecture that processes 200ms chunks of a
Editor's take
Thinking Machines Lab has unveiled a novel multimodal AI architecture designed for fluid, real-time interaction with humans. This 276B parameter Mixture-of-Experts model, TML-Interaction-Small, processes audio and visual inputs in 200ms micro-turns, enabling more natural and responsive collaboration.
This development addresses a key bottleneck in current AI systems, which often struggle with synchronized, low-latency multimodal understanding. By building interaction as a native component, this approach could significantly improve applications ranging from assistive technologies to more intuitive robotics, moving beyond static prompt-response cycles.
Future research will focus on scaling this architecture and its ability to handle more complex, multi-participant interactions. The effectiveness of the 12B active parameters in real-world scenarios, particularly in maintaining coherence and context across extended dialogues, will be a critical indicator of its practical utility.
Signal score: 5
This event was corroborated by 25 independent sources. The signal score weighs cross-source corroboration, recency, source weight and topic salience. How we rank stories.
Original reporting
This story summarises reporting published by MarkTechPost. Read the original article at MarkTechPost.