Thinking Machines wants to build an AI that actually listens while it talks

Thinking Machines Lab, an innovative AI startup launched by former OpenAI CTO Mira Murati, has unveiled a groundbreaking concept known as interaction models. This new approach suggests that AI can engage in conversations more dynamically by being able to interrupt speakers, much like a natural dialogue. Traditionally, AI interactions follow a linear pattern: a user speaks, the AI listens, and then responds. Thinking Machines is challenging this norm by developing a model capable of processing input and generating responses simultaneously—akin to a phone conversation rather than a text exchange. This capability, referred to as 'full duplex,' allows for a more fluid communication experience. The company’s model, dubbed TML-Interaction-Small, boasts a response time of just 0.40 seconds, mirroring the pace of human conversation and outpacing similar offerings from major players like OpenAI and Google. However, this technology is still in the research phase and is not yet available to the public. A limited research preview is anticipated in the coming months, with broader access scheduled for later this year. While the technical specifications are promising, the true test will come when users can experience the system firsthand. The concept of embedding interactivity into AI models as a fundamental feature, rather than an added capability, raises intriguing possibilities for the future of human-AI interaction. Only time will tell if the practical application aligns with the ambitious claims made by the company.