Mira Says She Wants Her AI To ‘Keep People On The Road’

Mira Murati pa they want to create AI superintelligence. But the former CTO of OpenAI sees human intelligence as the most important part of the equation.

At a time of increasing concern over AI to remove the job and increasing the power of a few large companies, the introduction of Murati, Thinking Machines Labit offers a very different vision of technology.

“One day we’ll have very intelligent machines,” Murati told WIRED. “But we think the best way to get a better future — a better future — is to make people vulnerable.”

Murati says AI should not take humans out of the equation. The best way, he says, is to let people create and develop their own AI models, then work with them to achieve their goals.

This week, Thinking Machines previewed a new breed of AI that points toward reality including humans. The company’s “action models” are trained to communicate with a person through a camera and microphone. Unlike many existing links, these new models don’t just capture and type words, and then feed them in a language they process like a chatbot. Communication systems understand continuous, interrupted, human communication – meaning they can understand the meaning of pauses, interruptions, and changes in speech. This helps them to adapt on the fly when someone makes a point or changes the subject. The company has shown several videos showing this, although the models have not been released publicly.

Murati’s approach contrasts with how the big AI companies seem to be pursuing artificial intelligence today. OpenAI, Anthropic, Google they are creating large models that handle the most complex tasks, including writing entire software programs from scratch, through custom. This requires little help from the person.

Thinking Machines aren’t the only startups that can envision a future involving humans. Other labs, including Humans&, also aim to develop AI systems that prioritize human interaction. Some famous economists he called AI researchers and companies to develop systems in this way, focusing on empowering people rather than revolutionizing them.

Murati stepped down from his role as chief technology officer of OpenAI in 2024, joining the Thinking Machine and several prominent engineers. Thinking Machines have it raised billions of dollars building frontier AI.

However, so far, the company has released only one product. Tinkerlaunched in October 2025, it makes it possible to develop a frontier AI model using regular data. Today, it is available as an API that researchers and engineers can use to improve open source models.

Alexander Kirillov, a founding member of the Thinking Machines group and an expert on multimodal AI, meaning models that listen to voice and video as well as text, says the lab’s new communication models also have the potential to support dynamic and adaptive AI.

“The model recognizes what you’re doing and is always there to respond and give you information or search for information or use other tools,” says Kirillov. “This is something that none of the (modern) people can do. The conversion (in discussion) is determined by the least intelligent system.”

Mira says it’s all part of the big AI vision.

“This represents the first bet on social cohesion,” he explains. “Where this is going is to expand on people’s preferences and interests, AI will understand intentions and make predictions.”