Last September, Mira Murati unexpectedly left her job as chief technology officer of Openai and said: “I want to create the time and space to do my own exploration.” The rumor in Silicon Valley was that she retired to start her own company. Today, she announced that she is indeed the CEO of a new public benefit corporation called Thinking Machines Lab. The mission is to develop Top-AI with a view to making it usable and accessible.
Murati believes that there is a serious gap between the rapid promotion of AI and the public’s understanding of the technology. Even sophisticated scientists do not have a thorough understanding of AI’s abilities and limitations. Thinking Machines Lab plans to fill the gap by building accessibility from the beginning. It also promises to share his work by publishing technical notes, papers and real code.
The underpinning of this strategy is Murati’s belief that we are still in the early stage of AI, and the competition is far from closed. Although this occurred after Murati started planning her lab, the rise of Deepseek claimed to build advanced reasoning models for a fraction of the usual costs, confirming her thinking that newcomers could compete with more effective models.
However, Thinking Machines Lab will compete at the high point of major language models. “In the end, the most advanced models will unlock the most transformative applications and benefits, such as enabling new scientific discoveries and breakthroughs of engineering,” the company writes in a blog post on Tuesday. Although the term “agi” is not used, Dinking Machines Lab believes that the scaling up of the capabilities of its models to the highest level is important to fill the gap it identified. Building those models, even with the effectiveness of the Deepseek era, will be expensive. Although Dinking Machines Lab has not yet shared its financing partners, it is confident that it will collect the necessary millions.
Murati’s pitch has attracted an impressive team of researchers and scientists, many of whom open on their resume. This includes former VP of Research Barret Zoph (who is now CTO at Thinking Machines Lab), multimodal research head Alexander Kirillov, head of special projects John Lachman, and top researcher Luke Metz, who left Open a few months before. The chief scientist of the laboratory will be John Schulman, an important inventor of Chatgpt who only left Openai to Anthropic last summer. Others come from competitors such as Google and Mistral AI.
The team moved to an office in San Francisco at the end of last year and has already started working on a number of projects. Although it is not clear what its products will look like, Dinking Machines Lab indicates that they will not be copy of Chatgpt or Claude, but AI models that optimize collaboration between people and AI – which Murati as the current bottleneck in the field Considered.
American inventor Danny Hillis dreamed of this partnership between people and machines more than 30 years ago. Hillis, a protégé of AI pioneer Marvin Minsky, built a superculator with powerful chips running parallel – a precursor to the clusters driving AI today. He calls it thinking machines. Prior to his time, the thinking of machines declared bankruptcy in 1994. Now a variation of his name, and perhaps its legacy, belongs to Murati.