Demis Hassabis, Father of Google's Artificial Intelligence

Part 11. The Human Side of Hassabis

31 Leadership

Kim Kyung-ran, Kim Kyung-jin

In March, in a Seoul hotel room, Demis Hassabis stood gazing out the window in a black suit without a tie. Ahead of his match against Lee Sedol 9-dan, he appeared calm on the surface, but his mind was boiling with millions of possible scenarios and worries about potential system failures. Mustafa Suleiman, an early member of DeepMind and a longtime colleague, described Hassabis in that moment as 'the eye of the storm.'

He was a leader who maintained a stillness that quieted all the chaos around him, yet held within himself a fiercer energy than anyone else. Hassabis's leadership begins with an obsession bordering on compulsion for perfection. Colleagues unanimously call him 'a man who is never satisfied.' David Silver, DeepMind's lead researcher, recalls Hassabis's exacting standards when reviewing research output.

'He is never satisfied with something merely working. He relentlessly asks, "Why does it work?", "Is this the best approach?", "What does this mean on the path to artificial general intelligence?" Those questions sometimes exhaust the researchers, but they are ultimately the driving force that pushes us beyond our limits.'

This perfectionism is deeply rooted in DeepMind's organizational culture. Hassabis rejects Silicon Valley's mantra of 'Move fast and break things.' Instead, he adheres to the scientific methodology of 'Move slowly, but verify thoroughly as you go.'

This often created friction with the speed-driven culture of Google, its massive parent company. When Google rushed to launch Bard and Gemini in response to ChatGPT's emergence, Hassabis argued until the very end that releasing products without guaranteed safety and accuracy violated DeepMind's philosophy. Some colleagues found his stance frustrating, but every time AI produced hallucinations or biased information, Hassabis's concerns were proven right.

Another defining trait of his leadership is his role as a 'shield.' He devoted all his energy to creating an environment where researchers could focus solely on their work, free from external pressures, especially the pressure to produce commercial results. It is well known that one of the conditions he set when Google acquired DeepMind in 2014 was maintaining its headquarters in London.

He wanted to secure a physical and psychological space where academic purity could be preserved, one step removed from the noise of Silicon Valley. One researcher said, 'Demis is like bulletproof glass that stops bullets for us.' At the negotiating table with Google executives, he was a cold-blooded strategist, but when he returned to the lab, he would transform back into a curious scientist, eyes sparkling as he asked, 'What happened with that idea from yesterday?'

But this style of leadership came with clear costs. His tendency to control everything and demand perfection sometimes created bottlenecks in decision-making. As DeepMind's organization grew, it became physically impossible for Hassabis alone to oversee the details of every project. Yet he still wanted to personally set the direction of major projects, and this sometimes imposed excessive workloads and waiting periods on the working-level staff.

Some former employees have confessed that DeepMind's high standards felt like suffocating pressure. 'The pride of working with geniuses was immense, but I trembled with fear every day that I might not measure up to those standards.' This confession reveals the shadow side of the 'elitist' culture Hassabis created. To his colleagues, Hassabis is not merely a boss. He is a companion who plays chess with them, debates the origins of the universe late into the night, and sometimes a mentor who displays unapproachable intuition. His leadership does not come from charismatic speeches or coercive commands. It springs from his intellectual brilliance in seeing to the heart of a problem, and from a pure passion to solve it.

His colleagues deeply believe in the vision he puts forward, 'solve intelligence to save the world,' and that is why they willingly endure his exactingness and tenacity. He is not a figure preserved in the myth of genius, but a tormented leader who pushes himself and his colleagues to the edge every day to turn that myth into reality. The compound role of researcher, executive, and entrepreneur. Demis Hassabis's business card reads 'CEO,' but his identity is not singular.

He is a world-class neuroscience researcher, an executive leading thousands of employees, and

an entrepreneur who must orchestrate massive capital and technological power. These three roles sometimes clash and sometimes complement each other, composing the complex figure that is Hassabis. First, as a researcher, Hassabis remains active on the front lines.

He is not simply a manager who receives reports. In the AlphaFold project, which earned him the 2024 Nobel Prize in Chemistry, he provided core ideas and steered the direction of the research. He still keeps up with the latest papers and enjoys debating with researchers in front of a whiteboard, scribbling equations together. For him, management is only a means to sustain research; it was never an end in itself. He personally proposed ideas such as applying the way the brain's hippocampus reconstructs memories to AI reinforcement learning algorithms, actively drawing on his background as a neuroscientist in AI research. This 'researcher's instinct' is the greatest competitive advantage that sets DeepMind apart from AI divisions at other big tech companies.

But as the organization grew, he had no choice but to put on the executive's coat. DeepMind, which started in a small London office in 2010, had become a massive organization with thousands of PhD-level researchers. Hassabis faced the challenge of harmonizing a freewheeling hacker culture with rigorous academic discipline.

He wanted to make DeepMind into 'the Bell Labs of the 21st century.' To that end, he invested great effort in recruiting top talent from academia, and designed evaluation systems that allowed them to tackle long-term challenges without being shackled to publication metrics. At the same time, he had to manage the expectations of Google, the parent company.

Justifying the astronomical funds poured into pure scientific research that generated no immediate revenue while running massive annual deficits required him to constantly prove DeepMind's value. The AlphaGo event and the decision to release AlphaFold's protein structure database for free were scientific achievements and, simultaneously, masterful business decisions that imprinted DeepMind's raison d'etre on the world. As an entrepreneur, Hassabis had to confront an even harsher reality.

When OpenAI's Sam Altman shook up the AI market with aggressive marketing and product launches, Hassabis had to walk a tightrope between the principle of 'responsible AI' and market demands. Because he understood the impact AI would bring better than anyone, he wrestled deeply with the tension between democratizing technology and maintaining safety controls. When Google's AI organizations merged in 2023 to form 'Google DeepMind,' he was effectively elevated to the position overseeing Google's entire AI strategy.

This meant he could no longer remain solely a researcher in an ivory tower. He had to win the speed race against competitors,

consider shareholder interests, and respond to regulatory moves by governments around the world. The tension among these compound roles imposed constant stress on Hassabis. As a researcher, he wants to pursue perfection, but as an entrepreneur, he cannot miss the timing.

As an executive, he wants organizational stability, but as an innovator, he must pursue relentless change. What is interesting is that Hassabis does not shy away from this contradiction but charges straight through it. He is a master of 'fusion,' combining disparate elements to create something new. Just as he moved between programming and game design during his days as a game developer, he strives to secure both the depth of research and the breadth of business simultaneously. He often tells his colleagues:

'We are doing science, but at the same time we are building products. These two things are not different. The best science will ultimately become the best product.'

This connects to the mission he has held for 40 years: 'Solve intelligence, then use that to solve everything else.' For him, research, management, and business are each merely tools for carrying out this grand mission. Hassabis exemplifies the 'hybrid leader' who moves fluidly among these three roles. Sometimes as a scientist in a lab coat, sometimes as a professor at a chalkboard, sometimes as a CEO in a crisp suit, he is establishing a new standard of leadership that the AI era demands.

The life structure that produces achievement: 'The Second Shift' and the routine of focus. As London's night deepens and the city's noise fades, Demis Hassabis's 'real' day begins. He calls this 'The Second Shift.' During the hours when ordinary workers rest after leaving the office, Hassabis has dinner with his family, puts his children to bed, and only after 10 p.m. does he sit back down at his desk. Then, until 4 a.m., he gives himself time for undisturbed immersion.

This routine is not ordinary overtime. It is his sacred ritual for recharging creativity rather than burning out amid the flood of tasks, meetings, and decisions that fill his intense daytime hours. If the daytime Hassabis is a 'manager' leading Google DeepMind's thousands of employees and processing countless reports as CEO, the nighttime Hassabis reverts to being a pure 'researcher' and 'dreamer.'

His study becomes a quiet laboratory during these hours. He reads the latest papers, develops ideas for technical challenges, and reviews DeepMind's long-term roadmap. During this long, uninterrupted stretch, he connects fragments of thought he could not examine during the day. Sometimes he mentally simulates the architecture of AI models as if reviewing a chess game, and other times he reads books from entirely different fields like physics or biology, seeking inspiration.

The reason he stays awake until 4 a.m. is not simply because he lacks time. He loves the particular clarity that the quiet of early morning provides. In those still hours when the world sleeps and all noise has vanished, he descends to the deepest floor of his thinking. He has said in interviews that 'most of my most important ideas come during this time.' The decisive 'divine move' in AlphaGo, or the thread that unraveled AlphaFold's structural challenges, were likely conceived in these solitary predawn struggles.

Having gone to bed at 4 a.m., he rises around 10 in the morning. It is a late start compared to most, but his brain is already warmed up from the previous night's intense thinking. He arrives at the office and begins his full workday in the afternoon.

His schedule is sliced into minute-by-minute blocks, yet he always joins meetings with a clear mind. The nighttime contemplation provides him the intellectual stamina to endure the day. Hassabis's lifestyle pattern reveals his unique conception of time. Even without always wearing watches on both wrists (though he sometimes wears a smartwatch and an analog watch simultaneously), he is acutely sensitive to the passage of time.

He seems like a man who constantly urges himself forward. Perhaps the thought that 'life is short, and the secret of intelligence to be unlocked is too vast' is what keeps him from sleeping. This 'Second Shift' is also his secret to maintaining balance. Even when crushed under the weight of corporate management during the day, at night he can escape again into the ideals of science and the mysteries of the universe.

This cycle keeps him from burning out. For him, staying up all night is not labor but intellectual play and a process of recovery. Through these hours, he briefly sets down the heavy burden of CEO and rediscovers the curious, sparkling gaze of the boy who first encountered a chessboard at age four.

Hassabis's great achievements are the product not only of genius-level talent, but of this solitary, intense nightly routine, layered one on top of another, day after day.

Curiosity and boundary-crossing fusional thinking. The key to understanding how Demis Hassabis's mind works lies in 'the absence of boundaries.' He refuses to become an expert in just one field, instead exploring knowledge by dancing along the border lines of multiple disciplines, a true polymath. His resume looks as if it stitched together the lives of three different people.

A chess prodigy ranked second in the world, a programmer who created a game that sold over ten million copies, and a neuroscientist who published papers in world-leading journals on memory and imagination. These disparate experiences melted into one inside the crucible called DeepMind. His fusional thinking began on the chessboard in childhood. Chess taught him the combination of logical reasoning and intuition.

The ability to simulate the future through calculation, the psychological insight to read an opponent's intentions, and the mental management skills to maintain composure under enormous pressure all became core assets when he later ran a company and solved hard problems. As a teenager, working with Peter Molyneux at the game studio Bullfrog to create 'Theme Park,' he learned to implement this simulation ability in computer code. Designing artificial intelligence where thousands of park visitors moved according to their individual desires, he came to hold the fundamental question: 'What is intelligence?'

This question led him to the world of neuroscience. At Cambridge and UCL, he explored how the human brain stores memories and how it uses those memories to imagine new futures. His finding that patients with hippocampal damage cannot imagine the future was named one of Science magazine's top ten scientific breakthroughs of 2007.

This research did not end as merely a neuroscience achievement. Hassabis applied this discovery directly to AI algorithms. The way DeepMind's reinforcement learning AI learns from past data (memory) and finds optimal strategies in unfamiliar environments (imagination) is modeled on the mechanisms of the human brain. The ability to translate the principles of the biological brain into digital code is the essence of Hassabis's fusional thinking.

His intellectual curiosity does not stop at science and engineering. He holds deep interest in history, philosophy, and the arts. He believes Renaissance humanism must meet modern technology.

When pondering AI's impact on humanity, he poses not only technical safety questions but philosophical and ethical ones. His question, 'How will this intelligence we are building reflect human values?' is one that could not arise without a grounding in the humanities. Through reading, he converses with thinkers across the ages and strives to graft their wisdom onto present-day AI development. Hassabis's fusional thinking is also reflected directly in DeepMind's hiring and organizational culture.

DeepMind employs not only computer scientists but physicists, biologists, neuroscientists, ethicists, and even ecologists from diverse backgrounds. Hassabis encourages them to communicate not in each other's specialized jargon but in the common language of 'problem-solving.' He firmly believes that 'innovation happens at the collision points of different disciplines.' The AlphaFold project succeeded precisely because machine learning specialists looked at a structural biology problem from an entirely fresh perspective.

For him, curiosity is not mere interest but something akin to a survival instinct. He sees everything in the world as connected. The three-dimensional structure of a protein, the opening moves on a Go board, and the physical laws of the universe may wear different appearances, but behind them all lies a shared principle of processing information and finding optimal states.

Hassabis ceaselessly crosses boundaries in pursuit of this universal principle. His fusional thinking is a powerful instrument that refuses to confine the world's complex problems to a single category, instead surveying them from all directions to find solutions. This is what shows that DeepMind's mission was never empty rhetoric. The documentary 'The Thinking Game' (2024): Tribeca Film Festival premiere and Hassabis's inner world. In June 2024, DeepMind's logo appeared on the screen at the New York Tribeca Film Festival. It was the moment director Greg Kohs's documentary 'The Thinking Game' received its world premiere.

Over five years of embedded access inside DeepMind, this film was not a corporate promotional video. It was a record that examined, as if under a microscope, the anguish and elation of people who challenged the nearly impossible mission of 'solving intelligence,' and the inner world of the man at the center: Demis Hassabis. The film unflinchingly shows the grueling periods of failure hidden behind the glamorous success story. The early obstacles of the AlphaFold project, launched to open an era in which AI leads scientific discovery, receive substantial screen time. Scenes of the research team hitting dead ends in despair, and Hassabis pacing the conference room with a tormented expression, testify that genius-level achievement is never the product of coincidence or inspiration alone. On screen, Hassabis

simultaneously shows the face of a confident CEO and that of an explorer feeling fear before the unknown. Through this film, audiences witness the motivation seated deep inside Hassabis. Why he so obsessively digs into the nature of intelligence, and what 'understanding' means to him, are laid bare. The film intercuts footage of him as a childhood chess prodigy, his game developer years, and his journey founding DeepMind and joining Google. Throughout, Hassabis expresses his unshakable conviction that 'science is the greatest tool humanity has ever invented,' and that AI will sharpen this tool further to solve humanity's hardest problems: disease, the climate crisis, energy. But the film does not rest in optimism alone.

It also contains scenes of Hassabis sounding warnings about AI's potential dangers, insisting that this powerful force must remain within controllable bounds. Tribeca audiences left the theater carrying both the wonder of technological achievement and the weighty question of human responsibility in handling it. For Hassabis, this film was not a trophy to flaunt his accomplishments, but a sincere confession, an attempt to share with the world the beauty and gravity of the act of 'thinking,' to which he has devoted his entire life.

In the film, he says: 'If we come to understand intelligence, it will be the most decisive moment of humanity understanding itself.' Connection to the 'AlphaGo' documentary (2017). 'The Thinking Game' is the spiritual sequel to and narrative conclusion of the documentary 'AlphaGo,' released in 2017 to enormous global resonance. Both were directed by Greg Kohs and both center on DeepMind as their subject, but their aims and moods are markedly different. If 'AlphaGo' was a tense sports drama focused on the 'contest' between human and AI, 'The Thinking Game' is an intellectual expedition about 'collaboration' between human and AI, expanding the horizons of science.

In the earlier film, the match against Lee Sedol 9-dan delivered the urgency of a zero-sum game with a clear winner and loser. AI was depicted as a fearsome entity surpassing humans, an adversary to be overcome. In this new work, by contrast, AI appears as a reliable partner that compensates for the limitations of human scientists and solves the challenge of protein structure prediction, a problem unsolved for 50 years.

The thread connecting the two documentaries is Hassabis's vision. In the final scene of 'AlphaGo,' rather than reveling in victory, he said, 'Now we must apply this technology to solving scientific problems.' 'The Thinking Game' is the proof that this declaration became reality. If AlphaGo was a victory in the closed system of the game of Go, AlphaFold is a victory in the infinite, complex open world of biology.

In terms of filmmaking technique, the two works are also linked. As in the first film, the director translates complex AI technology into visual language accessible to a general audience. But whereas 'AlphaGo' showed the battle of moves unfolding on the two-dimensional plane of a Go board, 'The Thinking Game' visualizes the dynamic structures of proteins twisting and folding in three-dimensional space, expressing the mystery of life as art.

The evolution of Hassabis as a character is another compelling point of comparison. If the Hassabis of 2016 was a spirited challenger eager to prove his technology to the world, the Hassabis of 2024, a Nobel laureate and knight, shows the mature bearing of a thinker concerned with technology's social responsibility and the future of humanity. If 'AlphaGo' asked, 'Can AI beat a human?', 'The Thinking Game' asks, 'How far can AI and humans go together?'

These two documentaries will endure as historical records of the decade-long trajectory of Hassabis and DeepMind, and as Parts One and Two of a grand narrative heralding the arrival of the age of artificial intelligence.

A portrait of Demis Hassabis