Mira Murati Cautiously Steps Back into the Spotlight

Mira Murati, a relatively unfamiliar figure on the conference scene, has been OpenAI’s CTO but rarely made public appearances on behalf of the company. Now, as the CEO of her own venture, Thinking Machines Lab, she has become even more elusive. Consequently, her recent dialogue with Bloomberg in San Francisco — her first significant media engagement in roughly 18 months — attracted notable interest, even with her cautious approach to sharing details.

The timing is quite significant. Over the past 18 months, Thinking Machines has functioned primarily behind the scenes, concentrating on capital acquisition, talent recruitment, and the development of its inaugural product, Tinker, an API designed for fine-tuning open-source AI models.

In the meantime, rivals competing for the same talent, clients, and media coverage have only grown more prominent. OpenAI, where Murati spent six years as CTO, remains a consistent presence in the news. Anthropic’s recent advancements are generating industry buzz. Furthermore, xAI, Elon Musk’s AI project, has merged with SpaceX ahead of a much-anticipated public offering, attracting considerable attention and investment. In such a landscape, maintaining a low profile is becoming less effective; at some point, generating buzz is essential to remind the market of your existence.

Murati seized the opportunity during her Bloomberg appearance to do just that. She unveiled what Thinking Machines calls “interaction models,” a novel type of AI interface that moves away from the conventional turn-based, prompt-and-response structure typical in current AI products. In discussion with interviewer Emily Chang, she elaborated on how these models are developed to manage continuous streams of audio, text, and video in 200-millisecond intervals. The aim is to capture the subtleties of human communication — including interruptions, mid-thought corrections, and pauses for reflection — in a manner closer to real-time. However, Murati stressed that this is an initial stage rather than a finished product, refraining from committing to a particular launch timeline.

She also reflected on the pivotal moment that brought her into the public spotlight: the tumultuous week in November 2023 when OpenAI’s board dismissed Sam Altman, resulting in her taking on the role of interim CEO. This event has been internally dubbed “the blip” at OpenAI. Murati conveyed a sense of clarity in her decision-making during that chaotic period, explaining that her objective to safeguard the mission and the team rendered her choices obvious, despite the exterior chaos. She indicated that the company would have “imploded” without her involvement during those remarkable five days and the following aftermath. Nevertheless, she acknowledged that a clear intent does not equate to foresight regarding consequences. In hindsight, she wished she had pushed for better information, a smoother transition plan, and greater transparency. However, she did not clearly state whether she viewed the outcome as beneficial.

When asked if she still trusts her former boss, she expertly dodged the question, shifting the focus to her overarching concern, which she reiterated multiple times: the centralization of critical decision-making in too few hands — not just at OpenAI but throughout the sector. She emphasized that her main worry centers less on the character of individual leaders (though that is important) and more on the absence of structural checks. Good individuals can make poor decisions, and well-meaning organizations can veer off course. She suggested that an excessive emphasis on individual virtue has overshadowed the need for governance.

Chang also gently inquired about the departures of several key researchers from Thinking Machines in recent months, a subject Murati has largely avoided in public and one she downplayed during their conversation. She stated that establishing a leading AI lab from scratch often compresses years of typical organizational changes into just months. While she conceded that high salaries — the nine-figure wages now prevalent in the talent competition for AI — do captivate attention, she indicated that it doesn’t tell the entire story. Amusingly, she remarked about her own competitive mindset, “When I wake up in the morning, I am not thinking about how to kill the competitor.”

Naturally, Chang asked her about the future of AI, especially regarding humans who were initially said to be empowered by AI but now fear mass job displacement and a future where AI could be utilized to create chemical weapons.

Murati, who is originally from Albania and speaks with a slight Eastern European accent, provided a balanced response. She challenged the idea of an inevitable dystopia or utopia, asserting that neither scenario is predetermined and that the current phase will significantly influence the ultimate direction of these outcomes. Nonetheless, she cautioned — reiterating a point made earlier in the interview — that if humans disengage too soon from overseeing developments, the future could be markedly different, and not in a positive way.