Former OpenAI CTO Mira Murati has introduced her new startup. Unsurprisingly, it’s centered on AI.
Referred to as Thinking Machines Lab, the startup, which got here out of stealth at the moment, intends to construct tooling to “make AI work for [people’s] distinctive wants and objectives,” and to create AI methods which can be “extra extensively understood, customizable, and usually succesful” than these at the moment accessible.
Murati is heading up Pondering Machines Lab as CEO. OpenAI co-founder John Schulman is the corporate’s chief scientist, and Barret Zoph, OpenAI’s ex-chief analysis officer, is the CTO.
In a weblog put up shared with TechCrunch, Pondering Machines Lab wrote that whereas AI capabilities have superior dramatically, “key gaps” stay.
“The scientific group’s understanding of frontier AI methods lags behind quickly advancing capabilities,” the weblog put up reads. “Data of how these methods are skilled is concentrated throughout the high analysis labs, limiting each the general public discourse on AI and other people’s talents to make use of AI successfully. And, regardless of their potential, these methods stay tough for individuals to customise to their particular wants and values.”
Pondering Machines Lab plans to deal with constructing “multimodal” methods that “work with individuals collaboratively,” in response to the weblog put up, and that may “adapt to the complete spectrum of human experience and allow a broader spectrum” of purposes.
“[W]e are constructing fashions on the frontier of capabilities in domains like science and programming,” the weblog put up acknowledged. “Finally, probably the most superior fashions will unlock probably the most transformative purposes and advantages, akin to enabling novel scientific discoveries and engineering breakthroughs.”
AI security will likely be one other core tenet of Pondering Machines Lab’s work. The corporate mentioned that it plans to contribute to security by stopping misuse of the fashions it releases, sharing finest practices and recipes for construct protected AI methods with the trade, and supporting exterior analysis on alignment by sharing code, datasets, and mannequin specs.
“We’ll deal with understanding how our methods create real worth in the actual world,” Pondering Machines Lab wrote in its weblog put up. “An important breakthroughs usually come from rethinking our aims, not simply optimizing current metrics.”
I began Pondering Machines Lab alongside a outstanding workforce of scientists, engineers, and builders. We’re constructing three issues:
– Serving to individuals adapt AI methods to work for his or her particular wants
– Growing sturdy foundations to construct extra succesful AI methods
– Fostering a…— Mira Murati (@miramurati) February 18, 2025
Murati left OpenAI last October after six years at the company. On the time, she mentioned she was stepping away to “do her personal exploration.”
Murati got here to OpenAI in 2018 as VP of utilized AI and partnerships. After being promoted to CTO in 2022, she led the corporate’s work on ChatGPT, the text-to-image AI DALL-E, and the code-generating system Codex, which powered early variations of GitHub’s Copilot programming assistant.
Mirati was briefly OpenAI’s interim CEO after CEO Sam Altman’s abrupt firing. Altman has described her as an in depth ally.
For months, rumors have flown of Murati hiring high-profile AI researchers and staffers for an AI enterprise. Pondering Machines Lab’s weblog lists 29 staff from OpenAI, Character AI, and Google DeepMind, amongst different high companies.
Pondering Machines Lab is actively hiring machine studying scientists and engineers, in addition to a analysis program supervisor, per the corporate’s put up.
At one level, Murati was mentioned to be in talks to lift over $100 million from unnamed VC companies. The weblog didn’t verify or deny this.
Earlier than OpenAI, Murati spent three years at Tesla as a senior product supervisor of the Mannequin X, the automaker’s crossover SUV, throughout which Tesla launched early variations of Autopilot, its AI-enabled driver-assistance software program. She additionally was VP of product and engineering at Leap Movement, a startup constructing hand- and finger-tracking movement sensors for PCs.
Murati joins a rising record of former OpenAI execs launching startups, together with rivals akin to Ilya Sutskever’s Protected Superintelligence and Anthropic.