Murati Previews 200-Millisecond AI Interface at Thinking Machines as She Reemerges After 18 Months
Updated
Updated · TechCrunch · Jun 5
Murati Previews 200-Millisecond AI Interface at Thinking Machines as She Reemerges After 18 Months
2 articles · Updated · TechCrunch · Jun 5
Summary
Thinking Machines says its new “interaction models” process audio, text and video in 200-millisecond intervals, aiming to move beyond today’s turn-based prompt-and-response AI.
Mira Murati framed the system as an early step toward AI that collaborates more naturally with humans by tracking interruptions, pauses and mid-thought corrections in near real time.
Her first major media appearance in roughly 18 months also served as a signal that the startup is stepping out after focusing on fundraising, hiring and its Tinker fine-tuning API.
Murati used the interview to revisit OpenAI’s five-day 2023 leadership crisis, saying the company would have “imploded” without her role but that she now would have pushed harder for more information and transparency.
She tied that episode to a broader warning that frontier AI still concentrates too much power in too few hands, arguing governance—not faith in individual leaders—will shape whether AI’s impact turns out well.
As AI models begin reacting in milliseconds, what new safeguards can prevent humans from losing control?
Can 'human-inclusive' AI thrive when key founders are poached by rivals for billion-dollar packages?
With leaders calling each other dishonest, how can the AI industry build public trust while its internal governance remains so volatile?
200 Milliseconds to the Future: Inside Thinking Machines Lab’s Quest for Seamless Human-AI Collaboration
Overview
After operating quietly for eighteen months, Mira Murati and Thinking Machines Lab have reemerged with a major breakthrough: a 200-millisecond AI interface that enables real-time, human-like interaction. During their time out of the spotlight, the company focused on raising capital, expanding its research team, and launching Tinker, an API for fine-tuning open-source AI models. This new interface moves beyond traditional, disjointed turn-based systems, aiming to make AI conversations feel seamless and natural. The result is a significant step forward in how people and AI can work together, setting new standards for the industry.