Last September, Mira Murati unexpectedly left her job as Director of Openai Technology, saying: “I want to create time and space to make my own exploration.” The rumor in Silicon Valley was that she was withdrawing to start her own business. Today, she announced that she is indeed the CEO of a new public service company called Thinking Machines Lab. Its mission is to develop a first -rate AI in one eye to make it useful and accessible.
Murati thinks that there is a serious difference between the rapid advancement of AI and the understanding of the public of technology. Even sophisticated scientists do not have a good understanding of the capacities and limits of AI. The laboratory of reflection machines plans to fill this gap by strengthening accessibility from the start. He also promises to share his work by publishing technical notes, articles and real code.
Under the rest of this strategy is the conviction of Murati that we are still in the first stages of the AI, and the competition is far from closed. Although this happened after Murati began to plan his laboratory, the emergence of Deepseek – who claimed to build advanced reasoning models for a fraction of the usual cost – declares his thought that new arrivals can compete with more models effective.
The laboratory of reflection machines will compete on the top of the end of large languages. “In the end, the most advanced models will unlock the most transformative applications and advantages, such as the activation of new scientific discoveries and engineering breakthroughs”, writes the company in A blog article Tuesday. Although the term “act” is not used, Thinking Machines Lab believes that the increase in capacities of its models at the highest level is important to fill the gap it has identified. The construction of these models, even with the effectiveness of the Deepseek era, will be expensive. Although Thinking Machines Lab has not yet shared its financing partners, he is convinced that he will increase the necessary millions.
Murati’s field has attracted an impressive team of researchers and scientists, many of whom opened on their curriculum vitae. These include the former vice-president of Research Barret Zoph (who is now CTO at Thinke Machines Lab), the director of multimodal research Alexander Kirillov, the special project manager John Lachman, and the best researcher Luke Metz, who left AI open several months earlier. The laboratory -in -chief scientist will be John Schulman, a key inventor of Chatgpt who left Openai for Anthropic only last summer. Others come from competitors like Google and Mistral IA.
The team moved into 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 to what its products will look like, Thinking Machines Lab indicates that they will not be copies of Chatgpt or Claude, but models of AI which optimize collaboration between humans and AI – that Murati considers the current curl in the field.
The American inventor Danny Hillis dreamed of this partnership between people and machines over 30 years ago. Protected from the pioneer of the Marvin Minsky, Hillis built a super computer with powerful fleas in parallel – a precursor to the clusters who execute AI today. He called it machines thinking. Before his time, machines thinking declared bankruptcy in 1994. Now, a variation of his name, and perhaps his inheritance, belongs to Murati.
