Mira Murati Launches Thinking Machines Lab To Make AI More Accessible (19459000)
In September last year, Mira Murati abruptly left her position as chief technology officer at OpenAI. She said, “I wanted to create the space and time to do my own exploring.” The rumor was spreading in Silicon Valley that she was leaving to start her own business. Today, she announced that she is indeed the CEO of a public benefit corporation named Thinking Machines Lab. Its mission is the development of top-notch AI, with a view to making it useful and available.
Murati thinks there’s a gap between the rapidly advancing AI technology and the public’s knowledge of it. Even the most sophisticated scientists are unsure of AI’s capabilities. Thinking Machines Lab aims to close this gap by integrating accessibility from the beginning. It also promises to publish technical notes, papers and actual code.
Murati’s belief, that we are in the early stages AI and the competition is not closed, is the foundation of this strategy. Murati’s belief that newcomers could compete with more efficient models is vindicated by the emergence DeepSeek, which claimed to build advanced reasoning model for a fraction the cost of the normal models.
Thinking Machines Lab, however, will compete at the top end of large language model. The company wrote in a blog that the most advanced models would unlock the most transformative benefits and applications, such as enabling new scientific discoveries and engineering advances. Though the term “AGI,” is not used, Thinking Machines Lab is convinced that scaling up the capabilities of their models to the highest possible level is essential to filling the gaps it has identified. Even with the DeepSeek-era efficiencies, it will be expensive to build those models. Though Thinking Machines Lab isn’t yet revealing its funding partners, it is confident that it can raise the required millions.
The pitch of Murati has attracted a team of scientists and researchers, many of whom are OpenAI alumni. Barret Zoph, former VP of Research (now CTO at Thinking Machines Lab), Alexander Kirillov as head of multimodal research, John Lachman as head of special projects, and Luke Metz who was a top researcher when he left Open AI a few months ago are among the team. John Schulman will be the lab’s chief scientific officer. He is a key ChatGPT creator who joined Anthropic last summer after leaving OpenAI. Others are from competitors such as Google and Mistral AI.
In late 2018, the team moved to San Francisco and has already begun working on a number projects. Though it is not yet clear what the products will look, Thinking Machines Lab has indicated that they won’t copy ChatGPT or Claude but AI models which optimize collaboration between humans with AI, which Murati believes to be the current bottleneck. Danny Hillis, an American inventor, dreamed about this partnership between humans and machines more than 30 years ago. Hillis, a protegee of AI pioneer Marvin Minsky built a supercomputer with powerful chips running simultaneously–a precursor to the clusters used today for AI. He called it Thinking Machines. Thinking Machines, a company that was ahead of its time in 1994, declared bankruptcy. Murati is a variant of its name and perhaps its legacy.
