Mira Murati, the visionary co-founder of Thinking Machines Lab and former CTO of OpenAI, is spearheading a transformative approach to artificial intelligence. While many ponder the arrival of AI superintelligence with a mix of awe and apprehension, Murati champions a future where human intellect remains at the very core of this evolution.
At a time when concerns about AI-driven job displacement and concentrated corporate power are growing, Murati’s new venture offers a refreshingly optimistic perspective. She firmly believes that AI doesn’t have to automate humans out of the equation; instead, it can empower us.
Redefining AI Superintelligence: A Human-Centric Vision
“At some point we will have super-intelligent machines,” Murati shared with WIRED, acknowledging the inevitable trajectory of AI development. However, she emphasized a crucial distinction: “But we think that the best way to actually have many possible futures—good futures—is to keep humans in the loop.” This philosophy guides Thinking Machines Lab’s mission to develop AI that works *with* people, allowing them to build and customize their own frontier AI models to achieve unique goals.
This approach stands in stark contrast to the current direction of many major AI companies. Giants like OpenAI, Anthropic, and Google are largely focused on developing increasingly complex models that can perform intricate tasks, such as generating entire software applications, often with minimal human intervention beyond a text prompt. Thinking Machines Lab, however, envisions a more symbiotic relationship, prioritizing human agency and collaboration.
Introducing Interaction Models: A New Era of AI Communication
Thinking Machines recently previewed a groundbreaking development: their new “interaction models.” These sophisticated AI systems are trained to communicate with individuals naturally through a camera and microphone, moving far beyond mere speech transcription. Unlike existing voice-mode interfaces that simply feed transcribed text into a language model, these interaction models natively understand the continuous, often ‘messy,’ nature of human conversation.
This means they can grasp the subtle meaning behind pauses, interruptions, and shifts in tone, allowing them to adapt dynamically as someone clarifies a point or changes the subject. The company demonstrated these capabilities through compelling videos, highlighting a future where AI truly comprehends and responds to nuanced human communication. Alexander Kirillov, a founding team member and multimodal AI expert, underscores the unique power of these models, noting their ability to constantly perceive user actions and provide responsive, contextual information.
Empowering Users: Customization and Collaboration
The vision extends beyond natural interaction; it’s about empowerment. Thinking Machines is not alone in advocating for a more human-inclusive AI future, with other labs like Humans& also developing systems that prioritize human collaboration. Prominent economists have echoed these sentiments, urging AI researchers to build systems that empower individuals rather than replace them.
Murati, who departed her role as chief technology officer of OpenAI in 2024 to cofound Thinking Machines with a team of esteemed engineers, has secured substantial funding to pursue this frontier AI. So far, the company has introduced one key product: Tinker. Launched in October 2025, Tinker is an API that allows researchers and engineers to refine open-source frontier AI models using their own custom data, paving the way for highly personalized AI applications.
This ability to customize and personalize AI is a cornerstone of Thinking Machines’ strategy. Kirillov emphasizes that these interaction models will enable users to have AI that is “constantly there to be able to reply and give you information or search for information or use other tools,” in a way that current models cannot. The goal is to move beyond rigid, turn-based conversations toward a fluid, intelligent partnership.
The Future of Human-AI Partnership
Ultimately, Murati sees these advancements as a critical first step towards a broader vision of human-AI collaboration. “This is showing the first bet on human collaboration,” she states, highlighting the foundational principle of her company.
Her grand vision is for AI to truly amplify individual preferences and values, understanding and even predicting human intent. Thinking Machines Lab is charting a course toward an AI superintelligence that is not only powerful but also deeply intertwined with, and enriched by, the human experience.
Source: Wired – AI
