Mira Murati's Thinking Machines Lab Targets More Natural AI Conversation

What Happened

Thinking Machines Lab, the artificial intelligence startup founded by former OpenAI chief technology officer Mira Murati, has released new research targeting a longstanding limitation of AI assistants: the stilted, robotic quality of machine-generated conversation. The work, reported Thursday, represents the company's first major public technical disclosure since its founding.

Background

Murati served as CTO at OpenAI from 2018 until her departure in September 2024, a tenure that included oversight of the development of GPT-4 and the initial release of ChatGPT. She announced Thinking Machines Lab shortly after leaving OpenAI, positioning it as an independent AI research and product company. The startup has operated largely out of public view since its formation, attracting attention primarily because of Murati's prominent role in the prior development of some of the most widely used AI systems in the world.

The problem the lab is addressing is widely recognised across the AI industry. Current large language models tend to produce responses that follow predictable, formal patterns, lack the natural rhythm of human speech, and handle conversational dynamics such as interruption, hesitation, and topic shifts poorly. These qualities limit the practical utility of AI assistants in real-time communication settings.

What the Research Involves

According to reports, the Thinking Machines Lab work focuses on making AI dialogue feel less mechanical by improving how models handle the flow and texture of conversation. Specific technical details disclosed publicly remain limited at this stage. The research has not yet been described in a peer-reviewed publication based on available wire report information, and the company has not announced a commercial product release date tied to the findings.

The lab has not disclosed funding figures, headcount, or its full research agenda publicly. Murati has not made detailed public statements about the company's product roadmap beyond broad descriptions of its mission to develop AI that is more capable and more aligned with human needs.

Context in the Broader AI Landscape

The push for more natural AI conversation has become a competitive focus across the industry. OpenAI released an advanced voice mode for ChatGPT in 2024 that the company described as capable of more fluid spoken interaction. Google has invested in conversational capabilities within its Gemini model family. Several startups have also entered this space specifically targeting voice and dialogue applications.

Thinking Machines Lab enters this field with Murati's technical background as a distinguishing factor. During her time at OpenAI, she was a central figure in both the research and deployment sides of the organisation. Her departure, along with several other senior OpenAI staff around the same period, drew significant industry attention.

The timing of the lab's disclosure also comes as AI companies broadly face intensifying scrutiny over the gap between marketed capabilities and real-world performance, particularly in conversational applications where users encounter limitations quickly.

Adoption and Market Signals

Separate data published this week underscores the practical stakes of improving AI usability. A report on Australian business AI adoption found that only 7 percent of firms in Australia have integrated AI into their operations, a figure researchers attributed in part to usability and trust barriers rather than access alone. While that figure pertains to enterprise adoption broadly, analysts covering the sector have noted that conversational quality is consistently cited by business users as a factor in adoption decisions.

Thinking Machines Lab has not announced enterprise partnerships, API access, or a consumer product in connection with this research disclosure.

What Comes Next

Thinking Machines Lab has not announced a timeline for publishing full technical documentation or releasing a product based on the research, and the company's next scheduled public disclosure has not been confirmed.