Last September, Mira Murati resigned from her position as chief technology officer at OpenAI, expressing a desire to "create the time and space to do my own exploration." Speculation in Silicon Valley suggested she was preparing to launch her own venture. Today, she confirmed these rumors with the announcement of her new role as CEO of Thinking Machines Lab, a public benefit corporation dedicated to advancing artificial intelligence with a focus on accessibility.
Murati identifies a significant gap between the rapid progress in AI technology and the public’s comprehension of its capabilities and limitations. Even among professionals in the scientific community, the understanding of AI remains unclear. Thinking Machines Lab aims to bridge this gap by prioritizing accessibility from the outset and committing to transparency through the publication of technical notes, papers, and code.
Believing we are still in the nascent phase of AI development, Murati argues that the competitive landscape is still open. The recent emergence of DeepSeek—boasting advanced reasoning models at lower costs—bolsters her perspective that new entrants can rival existing high-efficiency models.
Thinking Machines Lab plans to place a strong emphasis on high-end large language models, asserting that the most sophisticated models will facilitate transformative advancements, including groundbreaking scientific and engineering developments. The company intends to scale its models’ capabilities significantly, though achieving this will be resource-intensive. While funding partners have yet to be disclosed, the organization is optimistic about securing necessary investments.
Murati has gathered a talented team of researchers and scientists, many of whom previously worked at OpenAI, including Barret Zoph as the chief technology officer, and other notable figures such as Alexander Kirillov and John Lachman. The lab’s chief scientist, John Schulman, who played a crucial role in developing ChatGPT, rounds out this impressive lineup alongside experts from other top companies in the AI field.
Having established an office in San Francisco, the team has begun work on various initiatives aimed at optimizing collaboration between humans and AI—an area Murati sees as the primary bottleneck in the current landscape of artificial intelligence.
This vision hearkens back to more than three decades ago when American inventor Danny Hillis, inspired by AI pioneer Marvin Minsky, created a supercomputer called Thinking Machines, marking a pioneering effort in AI. Despite its eventual bankruptcy in 1994, the legacy of Hillis’ work resonates in Murati’s new venture.
