Who knew Demis Hassabis had this side to him!
After an intense three-day binge reading of Hassabis's latest official biography, I've uncovered many unknown facets of the man leading Google AI.
"Tricking" Geoffrey Hinton.
On the night Hinton's startup was being auctioned, DeepMind bid $10 million.
Unfortunately, Hinton was too popular. Realizing he had no chance, Hassabis called Hinton and told him: "Actually, your company is worth $50 million."
"Playing" Mark Zuckerberg.
Zuckerberg invited Hassabis to dinner at his home, but little did he know that Hassabis was essentially "interviewing" Zuckerberg. He arrived with a strategic mind, specifically intending to fish for information.
Once he got the answers he needed, he brutally rejected Zuckerberg.
...
Wait, what? This is completely different from the image I had of him.
The Nobel Prize winner and head of DeepMind, Demis Hassabis, is far more calculating than he appears on the surface.
In the book Hassabis: The Brain of Google AI, we've dug up 9 juicy pieces of gossip about Hassabis and DeepMind.
9 Revelations from the Hassabis Biography!
1. The Electronic Music Boy Racing Porsches in Cambridge
Hassabis's college life was quite flamboyant.
During his first year at Cambridge, he and his friends would dance all night until they collapsed into bed in the morning, only to keep listening to electronic music.
Specifically, The Prodigy's album — Music for the Jilted Generation.
It's quite dark, pure underground music; it's unexpected that Hassabis enjoyed this style.
By his second year of university, 19-year-old Hassabis was already driving a Porsche.
It was a Porsche 911 Turbo, which in 1995 was worth about 820,000 yuan.
As for where the money came from... the car was actually borrowed from Molinuex, his boss at a game company.
The excuse was that he needed it for commuting. In reality, he used it for joyriding and going to raves.
Slamming the Porsche steering wheel while partying all night—that was Hassabis at eighteen or nineteen.
2. Fortunate That the First Startup Failed
After graduating, Hassabis and his close friend David Silver founded a game company called "Panacea Studios."
Unfortunately, the company went bankrupt shortly after their first game was completed.
The computing power at the time couldn't support Hassabis's ambitions. Silver spent countless nights rewriting code and tinkering with hardware, yet the system crashed right at the start of the first demo at a trade show.
Silver broke down. He bolted out the door, found a sofa, and fell into a deep sleep from exhaustion.
Miraculously, this major accident didn't affect the overall demo's impact.
When Silver woke up, he discovered that Hassabis had used his silver tongue to miraculously bluff his way through.
Despite this, the game was poorly received upon release.
Silver eventually couldn't handle the pressure of working with Hassabis and resigned to recover in France for an entire year.
Their first venture ended in failure, but it seemed to steer them toward their destined paths.
Silver became obsessed with reinforcement learning, pursuing a PhD under the father of reinforcement learning, Richard Sutton; Hassabis became obsessed with the human brain and pursued a PhD in neuroscience at University College London.
In hindsight, it's a good thing that startup failed. If the game had sold well, DeepMind might never have existed.
3. Peter Thiel's Contrarian Bet
Peter Thiel was DeepMind's first heavyweight investor.
Thiel is an oddity in the investment world, never following the crowd and loving to challenge the "consensus." This is his "contrarian investment thinking."
Upon meeting Hassabis, his intuition told him: DeepMind was scientifically solid, but commercially a disaster.
At the same time, he recognized that Hassabis was a terrifying "mission-driven" entrepreneur. Such people don't care about money or fame; for them, entrepreneurship is destiny, and they will do whatever it takes to achieve their dream.
Thiel found this intriguing.
However, several partners at Founders Fund disliked DeepMind. To them, Silicon Valley was the only tech mecca, and they couldn't understand why DeepMind was determined to stay in London.
Only Luke Nosek remained consistently appreciative of Hassabis.
Thiel, being a contrarian, liked this "anti-consensus" project and approved Nosek's move. In late 2010, $2.3 million was wired to DeepMind.
But this same contrarian thinking was why Thiel eventually split with DeepMind. While he admired the "sense of mission," Thiel never actually believed in the AGI Hassabis talked about.
As DeepMind began to stabilize, hire aggressively, and buy computing power—becoming more "consensus-driven"—Thiel viewed it as a bubble.
During the Series C round, Founders Fund refused to remain the lead investor. This opened the door for Vivian Chow.
Vivian Chow, a Hong Kong entrepreneur and close associate of Li Ka-shing, managed a significant portion of Li's wealth.
After meeting Hassabis for just 15 minutes, she promised to invest. Her $13.6 million investment during the critical Series C round largely saved DeepMind from the brink of bankruptcy.
4. Elon Musk: The Danger of Bragging
Elon Musk took an interest in DeepMind early on.
Luke Nosek, who pushed Founders Fund to invest, was also a SpaceX shareholder. Through him, Musk and Hassabis connected, and Musk promised to invest.
If things had gone smoothly, Google might never have been involved. Unfortunately, Musk's big mouth delivered the opportunity right to Google founder Larry Page.
In October 2012, Musk and Nosek witnessed the successful launch of the Falcon 9 and flew back to California on Musk's private jet. Page was also on the plane.
During the flight, Page mentioned AI and noted that Geoffrey Hinton might be starting a company; if true, Google would certainly acquire it.
Unexpectedly, this triggered Musk's competitive streak.
"I believe only one AI company can succeed, and I am the investor in that company—DeepMind!"
Hearing this, Page silently pulled out his Android phone and noted down the name "DeepMind."
5. The Regret of Undercharging Musk
In early 2013, another SpaceX rocket launch was successful. Seeing the news, Hassabis finally breathed a sigh of relief.
He was planning to ask Musk for investment, but it would have been awkward if the rocket had crashed.
Hassabis was right. Musk was in a great mood, and the first thing he said when the phone connected was:
"How much do you want me to invest?"
Hassabis was stunned. After so many funding struggles, he hadn't expected such a windfall. He hesitated for a moment and cautiously named a number that seemed astronomical to him at the time.
"$5 million."
Musk agreed instantly without a second thought.
In that moment, Hassabis realized: I messed up; I should have asked for $50 million.
6. The Offer Hassabis Couldn't Refuse
At Elon Musk's 42nd birthday party held in a castle, both Hassabis and Page were invited.
While guests were mingling, Page made his move, inviting Hassabis for a stroll in the castle courtyard. Then, Musk's lead was stolen.
Page gave Hassabis a reason he couldn't refuse:
"Don't you want to build AGI? I've already prepared the funding, the talent, the computing power... everything at Google. So why waste time starting from scratch?"
Hassabis craved AGI so desperately that to achieve this dream, he was willing to submit to Google.
7. Zuckerberg's Wasted Trip
Hearing that DeepMind was flirting with Google, the FOMO-driven Mark Zuckerberg couldn't sit still.
Learning that Hassabis was having lunch with Page, Zuckerberg quickly invited him to dinner at his own home. Hassabis agreed.
Zuckerberg prepared a feast. Everything was set for them to become close allies. However, Hassabis brought a strategic playbook to this first "date."
First, he talked AI with Zuckerberg. Since this was DeepMind's passion, the savvy Zuckerberg played along and showed great interest.
Then, while Zuckerberg's guard was down, Hassabis steered the conversation toward other hot techs—VR, AR, and 3D printing.
Whether Zuckerberg genuinely loved tech or fell into the trap, he responded to everything, saying: "There's potential in all of these!"
Hassabis's heart sank. He wanted someone "All-in," where AGI was greater than everything else—not a boss who could talk a bit about everything but had a three-minute attention span for each.
After dinner, Hassabis called Page: "Let's proceed."
If he couldn't have it, he'd try to dismantle it. After being rejected, Zuckerberg hired Yann LeCun at Facebook with the intention of poaching DeepMind's researchers.
8. The Art of the Deal with Hinton
Hassabis is truly calculating. When Page mentioned acquiring Hinton's company on the plane, DeepMind was actually involved in the process.
In 2012, Hinton's team succeeded wildly in the ImageNet competition, making them the most sought-after in the industry. At an event that year, Hinton promoted his startup.
As is well known, Hinton suffered from back pain and couldn't sit. That night, he turned a trash can upside down on a hotel room table to prop up his laptop and sent emails to four potential buyers.
The "Hinton Auction Night" had begun.
"DeepMind bids $10 million." For DeepMind at the time, this was an astronomical sum. The entire company's equity was only around $45 million.
Hassabis's calculation was: if he could bring Hinton under his wing, it would be a trump card in future negotiations with Google.
But the auction became too heated, and prices skyrocketed. DeepMind dropped out of the race. Immediately after, Hinton received a call from Hassabis:
"This is crazy! I think your company is worth $50 million!"
Wait, what happened to the $10 million?
9. Plotting a Split from Google
After joining Google, DeepMind went through a period of friction, primarily regarding AI safety. Hassabis always wanted more autonomy to ensure research independence.
In 2015, during a Google reorganization, Hassabis was told: your dream might come true.
Internally called the "Mario Project," the idea was to spin DeepMind off into a semi-independent experimental project, making Google's operations easier and financial reports look better.
Initial talks were smooth since it was Larry Page's idea. Until... the new CEO Sundar Pichai dealt him a blow.
Pichai is a hard man to read. Meetings would go well, only for him to overturn everything later, saying, "That's not what we meant."
Frustrated, DeepMind devised a "Crazy Plan B": raise $5 billion externally and leave Google. Not only was this a massive sum, but it threatened a legal nightmare.
Looking back, this "crazy" plan might have been the only way out. Pichai never had any intention of letting them go. He couldn't let DeepMind slip away; AI is a project that could disrupt Google's very foundation.
Seeing DeepMind's determination to leave, Pichai used a classic tactic—division.
DeepMind had three founders: Hassabis, Legg, and Suleyman. Among them, Mustafa Suleyman was unique. He had started as an assistant and fought his way up to a founder title.
Pichai approached Suleyman, suggesting they split: let Hassabis take the research team, and Suleyman would lead the application side and merge into Google.
As a reward for this "defection," Pichai promised Suleyman responsibility for all of Google's AI business.
Suleyman was convinced. In 2018, they were one step away from the final move.
A bubble of dreams expanded within DeepMind, promising both Suleyman and Hassabis what they desired. But just as the bubble became large enough to carry DeepMind away from the conflict—it popped.
John Giannadakis was poached by Apple. Amidst the personnel chaos, Jeff Dean of Google Brain took over as the head of Google AI.
Suleyman had nowhere to go. DeepMind's independence plan failed again.
Perhaps this path was impossible from the start. Hassabis spent three years in vain, only to prove the "Oppenheimer Dilemma": people only care about safety after the technology has completely spiraled out of control.
In 2023, Google Brain and DeepMind merged, and Pichai finally brought DeepMind fully under his control.
Surprisingly, after the merger, supreme command was given to Hassabis, rather than Jeff Dean, who had been there longer.
How Dean felt is unknown, but it's said he lost interest in the management duties he should have held.
Filling the remaining gap was an unexpected figure—the other Google founder, the playboy Sergey Brin, who had been an absent landlord for years.
— End —