Mira Murati steps back into the spotlight, carefully

Mira Murati may not be a natural on the conference stage, yet her role as CTO of OpenAI made her a key figure, if not always the public face. Now as CEO of Thinking Machines Lab, her presence has been even more elusive. However, her recent interview with Bloomberg in San Francisco marked her first significant media engagement in about 18 months, making it a noteworthy event, despite her cautious approach to sharing details.

The timing of her appearance aligns with Thinking Machines’ strategy of operating quietly over the past year and a half, during which they have focused on raising funds, hiring talent, and launching their product, Tinker, an API designed to fine-tune open-source AI models.

Meanwhile, rival companies have become increasingly prominent in the industry. OpenAI, where Murati was CTO for six years, frequently makes headlines. Anthropic is gaining significant attention, and xAI, Elon Musk’s AI initiative, has merged with SpaceX in anticipation of a major public offering, drawing considerable interest and investment. In this competitive landscape, maintaining a low profile has its limits; eventually, a company must assert its presence to remain relevant.

During her Bloomberg interview, Murati aimed to do just that, albeit with restraint. She introduced the concept of “interaction models” at Thinking Machines, describing them as a new form of AI interface. Unlike the typical turn-based, question-and-answer format of most AI products, these models are engineered to process ongoing streams of audio, text, and video in 200-millisecond intervals. This allows them to grasp the nuances of human interaction, such as interruptions, corrections, and pauses, in near real-time. However, Murati emphasized that this is merely an initial step rather than a completed product and refrained from providing a release date.

She also reflected on the tumultuous week in November 2023 when OpenAI’s board dismissed Sam Altman, and she assumed the role of interim CEO. Known internally as “the blip,” Murati expressed confidence in her decisions during that period, focusing on safeguarding the mission and the team, even as the situation seemed chaotic externally. She believed her involvement prevented the company from collapsing during those five days and its aftermath. However, she acknowledged that while her intentions were clear, the outcomes were less certain. In hindsight, she would have advocated for more information, a better transition strategy, and increased transparency, though she did not explicitly state whether she felt the situation resolved positively.

When asked about her trust in her former boss, Murati avoided the question, instead highlighting a broader concern she has emphasized repeatedly: the concentration of critical decisions in the hands of too few individuals—not just within OpenAI but industry-wide. Her concern is less about individual leaders’ character and more about the lack of structural safeguards. She noted that good people can make poor decisions and well-intentioned organizations can stray off course, suggesting that governance needs more attention than virtue.

Chang also inquired about the recent departures of prominent researchers from Thinking Machines. Murati, who has largely avoided discussing this publicly, minimized its significance during the interview. She explained that establishing a cutting-edge AI lab from scratch accelerates normal organizational changes. While acknowledging that lucrative compensation packages, now commonplace in the AI talent war, capture public attention, she argued they are not the sole factor. To some audience laughter, Murati quipped about her own competitive drive, saying, “When I wake up in the morning, I am not thinking about how to kill the competitor.”

Naturally, the conversation turned to the future of AI, including concerns about job displacement and the potential misuse of AI in creating chemical weapons, topics that have caused anxiety among those once promised empowerment through AI.

Murati, originally from Albania and speaking with a slight Eastern European accent, responded thoughtfully. She challenged the notion of an inevitable dystopian or utopian future, emphasizing that the current period is crucial in determining the direction AI will take. Yet, she warned that if humans relinquish control too soon, the future could be markedly different and not for the better.