Thinking Machines Lab, an AI startup founded last year by former OpenAI chief technology officer Mira Murati, announced on Monday what it calls an interaction model, which essentially sounds like AI that interrupts users.
At this point, all the AI models we’ve used so far work the same way. When you speak, it listens. It responds, you listen. Thinking Machines aim to change this by building models that process input and generate responses at the same time. So it’s more like a phone call than a text chain.
The technical term for this is “full duplex,” and the company’s model, TML-Interaction-Small, claims to respond in 0.40 seconds. This is approximately the speed of natural human conversation and significantly faster than comparable models from OpenAI and Google.
However, this is a research preview and not a product. The company has not yet released it to the public. A “limited research preview” will be available in the coming months, with a broader release expected later this year.
So what should we do? I’m not sure. The benchmarks are impressive, and the basic idea that interactivity should be native to the model rather than something built into it is definitely interesting. We won’t know if real-world experiences live up to the technical claims until people actually use it.
