Thinking Machines: Ex-OpenAI CTO’s new AI startup

 

Mira Murati Launches Thinking Machines to Democratize AI

Former OpenAI CTO Mira Murati has unveiled Thinking Machines, a new AI research and product company focused on making AI more accessible, customisable, and collaborative.

A Mission to Make AI More Inclusive

Thinking Machines aims to bridge key gaps in the AI landscape by ensuring these technologies are not just powerful but also adaptable and widely understood. The company is committed to democratising AI, enabling users to tailor AI tools to their unique needs.

“We’re building a future where everyone has access to the knowledge and tools to make AI work for their unique goals,” the company states.

Addressing AI’s Key Challenges

Despite AI's rapid advancements, knowledge remains concentrated within a select few research labs, making it difficult for broader communities to innovate. AI systems also lack customisation, limiting their real-world application. Thinking Machines aims to address these issues by creating open, customisable, and capable AI systems that foster both innovation and accessibility.

The company plans to achieve this by leveraging intellectual openness, cutting-edge infrastructure, and advanced AI safety practices to empower both researchers and end users.

A Team of AI Pioneers

Murati has assembled a team of leading scientists, engineers, and technologists who have previously worked on groundbreaking AI tools, including OpenAI’s ChatGPT, Character.ai, PyTorch, and OpenAI Gym. Their expertise positions Thinking Machines to make a substantial impact in the AI space.

A Human-Centric Approach

Unlike companies focused solely on autonomous AI, Thinking Machines prioritises human-AI collaboration. It embraces an open research culture, sharing research papers, technical insights, and code with the wider AI community.

“Scientific progress is a collective effort,” the company emphasises. “By collaborating with the broader research community, we can advance AI for the benefit of all.”

The company is also investing in multimodal AI systems that integrate text, video, and imagery, allowing seamless interaction between humans and AI. This personalisation will enable AI to support diverse fields, from scientific research to creative industries.

Building Strong AI Foundations

While many AI startups rush to deploy new systems, Thinking Machines is taking a deliberate and methodical approach. Its strategy rests on two key pillars:

1. Intelligent AI Models—Developing frontier AI models capable of breakthroughs in programming, scientific discovery, and engineering.
2. High-Quality Infrastructure—Creating efficient, secure, and user-friendly platforms that enhance AI usability and scalability.

The company is also focusing on advanced multimodal capabilities, integrating language, imagery, and sensory data to build AI that better understands and interacts with the world.

Ethical AI Through Product-Driven Learning

Thinking Machines is committed to AI safety and ethical innovation by integrating research with product development. This iterative approach ensures AI remains both practical and responsible.

Key AI safety principles include:

• Maintaining a high safety bar while preserving user freedoms.
• Sharing best practices for secure AI development.
• Supporting external research on AI alignment through open access to code, datasets, and models.

The company recognises that true AI progress requires rethinking objectives, not just optimising existing metrics. By focusing on real-world impact, Thinking Machines aims to develop AI that delivers tangible societal benefits.

A New Era for AI

Thinking Machines represents the next chapter for Mira Murati, whose leadership helped shape some of OpenAI’s most successful projects. With a team of top AI innovators, a commitment to openness, and a long-term vision, the company seeks to challenge the status quo in AI development.

By prioritising collaboration, accessibility, and responsible innovation, Thinking Machines aims to empower industries and individuals to harness AI’s full potential—on their own terms.