The Shift to 'Interaction Models'

Mira Murati, CEO of Thinking Machines Lab, has emerged from a 18-month stealth period to introduce a new technical paradigm: 'interaction models.' Unlike current AI products that rely on turn-based, prompt-response cycles, these models are designed to process continuous streams of audio, text, and video in 200-millisecond intervals. The goal is to capture the nuances of human communication—such as mid-thought corrections, interruptions, and pauses—to facilitate more natural, real-time collaboration between humans and AI.

Governance and Industry Concentration

Reflecting on her tenure at OpenAI, particularly the 2023 leadership crisis, Murati emphasized that technical clarity does not equate to clarity of consequences. She expressed concern over the current industry trend where consequential decisions are concentrated in too few hands. Murati argues that the industry has focused too heavily on individual leadership character rather than structural governance. She advocates for stronger institutional checks to prevent well-intentioned organizations from drifting, noting that even 'good people' can make poor decisions in the absence of proper oversight.

Building in a Competitive Landscape

Murati addressed the challenges of building a frontier AI lab amidst intense competition for talent and market attention. While acknowledging the volatility of her own organization, she downplayed the impact of high-profile researcher departures. She maintains that her focus remains on the product and the core thesis of human-AI collaboration rather than competitive posturing. Regarding the broader future of AI, she rejected the binary of inevitable utopia or dystopia, suggesting that the current period is the critical window for determining the trajectory of the technology. She warned that human oversight remains essential, stating that 'if humans take their hands off the wheel too soon, the future will look very different, and not better.'