Mira Murati Wants Her AI to ‘Keep Humans in the Loop’
Original reporting by Wired
Mira Murati, the former chief technology officer of OpenAI, remains committed to the audacious goal of building AI superintelligence. Yet, her new venture charts a path that radically departs from the prevailing industry trend, one she believes is essential for humanity’s best possible future. At a time when anxieties about AI's potential to displace jobs and centralize power are escalating, Murati’s startup, Thinking Machines Lab, offers an optimistic, human-centric vision: AI that doesn't automate us out, but rather keeps us firmly in the loop.
Murati left OpenAI in 2024 to co-found Thinking Machines, and this week, her team previewed a tangible demonstration of their philosophy: new "interaction models." Unlike current voice interfaces that simply transcribe words, these pioneering models are trained to natively comprehend the full, messy spectrum of human communication—deciphering nuances like pauses, tone shifts, and interruptions. This enables them to adapt dynamically, truly understanding intent and fostering a collaborative partnership between human and machine. This approach stands in stark contrast to the trajectory of many big AI companies, which largely focus on developing highly autonomous systems that require minimal human intervention. Thinking Machines, however, is making an explicit "first bet on human collaboration," aiming to amplify individual preferences and values with AI that instinctively understands and predicts human intention. It's a powerful counter-narrative, suggesting an empowering future where advanced AI genuinely works *with* people, not around them.
Mira Murati's Thinking Machines Lab presents a compelling and increasingly vital counter-narrative to the dominant trajectory of AI development. By prioritizing human integration and dynamic collaboration over automation, her vision for "interaction models" suggests a fundamental rethinking of how superintelligence might emerge and function. This approach, which allows AI to natively understand and adapt to the nuances of human communication, moves beyond simple command-and-response systems, instead fostering a symbiotic relationship where technology genuinely amplifies individual intent and values rather than dictating them.
The broader implications of this human-centric philosophy are profound, directly challenging prevailing anxieties about AI's potential to displace jobs and centralize power among a few large entities. Should Thinking Machines and similar ventures succeed, they could democratize access to powerful AI models, enabling individuals and smaller organizations to customize and steer advanced systems for their specific needs, rather than merely consuming generic, pre-trained behemoths. This paradigm shift offers a path toward AI not as a replacement for human intellect, but as an intimate cognitive partner, capable of enhancing human creativity, problem-solving, and agency across every domain. In embracing the messy, adaptive nature of human interaction, Murati’s work not only seeks to build superintelligence but endeavors to ensure that this future intelligence is inherently diverse, personalized, and ultimately, deeply human.