Demis Hassabis' Covert DeepMind Hedge Fund Initiative to Take on Jim Simons
In a remarkable revelation from Sebastian Mallaby's new book The Infinity Machine: Demis Hassabis, DeepMind, and the Quest for Superintelligence, it has emerged that DeepMind co-founder Demis Hassabis once attempted to build a secret hedge fund operation within the artificial intelligence research company. This clandestine project, which operated without Google's knowledge or approval, was ultimately terminated by the parent company when discovered.
The Secret Trading Team and Its Ambitious Target
Hassabis quietly assembled a dedicated team of approximately 20 researchers who were specifically tasked with developing sophisticated high-frequency trading algorithms. According to Mallaby's account, which was recently excerpted in Colossus magazine, this covert operation was conducted entirely under the radar of Google's oversight mechanisms.
The ultimate objective of this secret initiative was nothing less than challenging the legendary Jim Simons and his notoriously successful quantitative hedge fund, Renaissance Technologies. Based on Long Island, Renaissance Technologies has long been regarded as one of the most secretive and scientifically-driven financial operations in history, staffed primarily by researchers and scientists operating in a classified-like environment.
For Hassabis—a five-time World Games Champion at the Mind Sports Olympiad—the appeal was multifaceted. Renaissance Technologies represented a unique challenge that combined scientific rigor with financial strategy, essentially creating what he perceived as "a game worth playing" at the highest possible level.
The Kindred Obsessive: Peter Brown's Office Lifestyle
The man Hassabis specifically wanted to outperform was Renaissance Technologies' longtime leader Peter Brown, who happened to be a pioneering figure in deep learning research having studied under the renowned Geoffrey Hinton. Brown was known for his extraordinary dedication to his work, maintaining a fold-down bed against his office wall and, by most accounts, essentially living at the office.
Hassabis recognized in Brown a kindred obsessive spirit—a scientist who had successfully transformed a cloistered, secretive operation into what many consider the most successful hedge fund in financial history. This achievement represented the benchmark against which Hassabis measured his own secret project's potential success.
Google's Intervention and the Project's Demise
Unfortunately for Hassabis and his team, Google's corporate wariness ultimately proved fatal to the initiative. The parent company discovered the unauthorized project and promptly disbanded it before it could advance significantly toward its ambitious goals. This intervention occurred despite the team's progress in developing the high-frequency trading algorithms that were intended to compete with Renaissance Technologies' sophisticated systems.
Part of a Broader Struggle for Independence
The secret trading team episode represents just one chapter in a much larger narrative that Mallaby's book explores in considerable detail: DeepMind's years-long, repeatedly unsuccessful attempts to break free from Google's corporate control. Hassabis and co-founder Mustafa Suleyman spent nearly three years advocating for a governance structure they termed "Project Mario"—a comprehensive plan to spin DeepMind out as an independent "global interest company" with a specialized 3-3-3 oversight board.
This independence push gained significant momentum when LinkedIn co-founder Reid Hoffman committed an impressive $1 billion to potentially back a complete separation from Google. However, despite these substantial efforts and financial backing, the initiative ultimately failed to materialize. Google CEO Sundar Pichai maintained just enough ambiguity in his position to prolong negotiations before eventually closing the door on the independence proposal entirely.
Hassabis' Evolving Perspective on AI Governance
By the conclusion of this extended struggle for autonomy, Hassabis had arrived at a fundamentally different conclusion regarding artificial intelligence safety and governance. He explained to Mallaby that attempting to negotiate trustless governance structures in advance represented what he now considers "a dead end" approach, primarily because "the lines always get drawn in the wrong places."
Instead, Hassabis has developed a new philosophy centered on the importance of maintaining presence within decision-making environments. He now believes that what truly matters is staying inside the room where critical decisions are made and gradually earning trust over extended periods through consistent engagement and demonstrated responsibility.
Mallaby's book leaves readers contemplating whether this represents genuine wisdom gained through hard experience or simply sophisticated rationalization of circumstances beyond Hassabis' control—a question that remains provocatively unresolved in the narrative.
