OpenAI Core Product Head Thibault Sottiaux Reveals DeepMind's Two-Year-Old Language Model Technology
Thibault Sottiaux, OpenAI's core product head, revealed in an interview with Wired that Google DeepMind had mastered similar large language model and dialogue technologies nearly two years before the release of ChatGPT at the end of 2022, but bureaucratic conservatism at the top prevented its launch.
Sottiaux stated that the success of ChatGPT "woke him up," making it unacceptable for good technology to be "squandered" at his former employer, leading him to switch from DeepMind to OpenAI. Previously, he was responsible for infrastructure projects like AlphaGo at Google London and DeepMind, and now he reports directly to Greg Brockman at OpenAI, leading the major overhaul of ChatGPT and the integration with Codex.
Source: Public Information
ABAB AI Insight
Thibault Sottiaux previously focused on infrastructure development at DeepMind, and this revelation continues his shift from the Google ecosystem to the execution-oriented path at OpenAI. Having witnessed the success of AlphaGo yet encountering delays in large model deployment, he joins several former DeepMind talents who left due to decision-making disagreements, accelerating OpenAI's product iteration lead.
On the capital front, OpenAI is investing talent resources into the ChatGPT overhaul and Codex integration, motivated to seize the generative AI window by enhancing product reliability and agent capabilities through infrastructure expertise, concentrating resources on user scale expansion and workflow integration to secure market dominance.
Similar to multiple cases of leading technologies being "squandered" internally at Google, the AI industry is transitioning from technology reserves to competition in product execution. Sottiaux's experience highlights the impact of decision-making culture differences on competitive dynamics.
Essentially, this reflects capital concentration, with action-oriented talent shifting execution resources from conservative giants to agile platforms, causing pricing power to tilt from technology reserves to deployment speed, reshaping the generative AI market structure through rapid product iteration and forcing traditional giants to seek a new balance between bureaucracy and innovation.
ABAB News · Cognitive Law
Technology reserves profit from storage, slow execution loses the window.
Bureaucratic conservatism locks in advantages, action-oriented players profit from the market.
Infrastructure builds the foundation, product integration defines dominance.