Mira Murati on New AI Models and OpenAI Lessons: Returning to the Public Eye

Mira Murati, known as the former CTO of OpenAI and founder of the startup Thinking Machines Lab, broke her long silence in an interview with Bloomberg. Over the past 18 months, the company has focused on developing Tinker—an API for fine-tuning open AI models. However, amid fierce competition from OpenAI, Anthropic, and Elon Musk’s xAI, Murati decided to re-enter the public sphere by introducing the concept of "interaction models."

Thinking Machines’ new technology differs from traditional chatbots by its ability to process continuous streams of audio, text, and video with minimal latency of just 200 milliseconds. This advancement could enable AI to better perceive the dynamics of human communication, including interruptions and pauses. Murati emphasized that this is only the first step and did not disclose a specific timeline for the product’s market release.

In the interview, Murati also addressed the crisis at OpenAI in November 2023, when the board ousted Sam Altman and she temporarily assumed the CEO role. She acknowledged that the situation could have led to the company’s collapse but noted in hindsight that greater transparency and a clearer transition plan should have been demanded. When asked about her trust in Altman, Murati avoided a direct answer, instead raising concerns about the concentration of power among a few leaders in the AI industry.

Murati also commented on the recent departures of key researchers from Thinking Machines. She attributed this to the natural volatility of early-stage companies and noted that financial incentives are not the sole factor influencing specialists’ decisions. According to her, competition in the industry extends beyond talent wars, and the future of AI depends on how responsibly people manage the technology.