A new artificial intelligence research project launched by a startup founded by former OpenAI chief technology officer Mira Murati aims to make conversations with AI systems feel faster and more human-like.
The company, Thinking Machines Lab, announced a new technology called “interaction models,” which allows AI systems to process speech and generate responses at the same time. The approach is designed to move beyond the delayed, turn-based conversations commonly used by current voice assistants and chatbots.
The technology relies on what researchers describe as “full duplex” interaction, similar to the way people naturally communicate during phone calls or face-to-face discussions. Instead of waiting for a speaker to finish completely before responding, the AI can begin preparing and delivering replies while continuing to listen.
According to the company, its research model, known as TML-Interaction-Small, can respond in about 0.40 seconds. That timing is close to the pace of ordinary human conversation and is said to be faster than some existing voice-based AI systems developed by competitors.
Most AI assistants currently operate using a stop-and-start structure in which the user speaks first, the system processes the request, and the response follows after a noticeable pause. While functional, that format has often made interactions feel mechanical and less natural.
Thinking Machines Lab said future AI systems should treat interactivity as a central feature rather than an extra layer added after development. The company believes reducing delays and improving conversational flow could make digital assistants feel more intuitive in everyday use.
The project remains in the research phase and has not yet been released publicly. The company plans to introduce a limited research preview in the coming months before expanding access later this year.
Industry analysts say the announcement reflects a broader shift in the artificial intelligence sector, where companies are increasingly competing not only on the intelligence of their systems but also on how naturally they interact with users.
The race to create more lifelike AI assistants has intensified as technology firms seek to make digital tools more useful for customer service, virtual assistants, education and workplace communication.
Despite the promising benchmarks released by Thinking Machines Lab, questions remain about how the system will perform outside controlled testing environments. Researchers and users alike are expected to closely watch whether the technology can maintain accuracy, responsiveness and reliability during real-world conversations.
The announcement places Thinking Machines Lab among a growing number of firms focused on making AI communication smoother, quicker and closer to everyday human interaction.
