Mira Murati Startup Presents AI with Real-Time Communication

Thinking Machines Lab, a company founded by the former CTO of OpenAI Mira Murati, recently revealed a significant advance in Artificial Intelligence (IA) technology with the TML-Interaction-Small model. Based on 'full duplex' technology, the system allows AI to process information and generate responses at the same time as the user speaks, bringing the interaction of a telephone conversation closer together and eliminating typical breaks from online assistants. According to information disclosed by the company, the model was trained with millions of hours of human dialogues to learn speech patterns and natural interruption.

How does TML-Interaction-Small revolutionize AI with real-time communication?

TML-Interaction-Small can process the speech of the user in real time and respond continuously, without waiting for the sentence to end. This is possible thanks to 'full duplex' design, which allows simultaneous bidirectional communication. The AI can interrupt the user when necessary, making the dialogue more dynamic and human. In internal tests, response latency was reduced to less than 200 milliseconds, a milestone for the area. The technology also deals with speech overlaps, identifying when the user wants to intervene or just take a break.

Why is this real-time AI a milestone for developers?

The ability to process and respond in real time opens new borders for AI applications. In customer service, the model can handle complex questions without waiting for the customer to complete the sentence, speeding up technical support and sales. In education, it allows virtual tutors who respond to doubts while the student explains a problem. In the area of health, it can assist in rapid screening, hearing symptoms and suggesting actions simultaneously. Developers can integrate the model via API, with documentation available on the startup website, which also offers a sandbox for testing. For more details on how to implement this technology, check out our full release guide.

FAQs about real-time communication AI

What is full duplex technology?

Full duplex is a communication approach that allows simultaneous sending and receiving of data. In AI, it means that the system can listen while speaking, without the need to switch between listening and response modes.

How does Thinking Machines Lab plan to use this technology?

The company plans to integrate TML-Interaction-Small into virtual assistants, customer service systems and accessibility tools. Also studies versions for robotics and devices IoT, where real-time communication is crucial.

What is the potential impact of this technology on the AI industry?

By eliminating the barrier of sequential response, 'full duplex' technology can make interactions with AI as natural as human conversations. This should accelerate the adoption of assistants in contexts that require quick responses, such as call centers and emergency applications. Experts point out that advancement can influence the design of future language models, which will prioritize processing simultaneity.

With the TML-Interaction-Small, the Thinking Machines Lab not only demonstrates the feasibility of continuous communication with AI, but also sets a new standard for man-machine interactions. The company has already announced partnerships with development platforms and plans to make the model available on a commercial scale later this semester.