TheChek News Flow

Demis Hassabis is tasked with keeping Google ahead on a technology that its CEO has called more profound than the invention of fire or electricity

Demis Hassabis, the neuroscientist tasked with keeping Google at the vanguard of artificial intelligence, was on a hot streak. Then the AI went haywire.

Google’s AI chatbot began angering users with biased and ahistoric responses. Hassabis wanted to make something clear: It wasn’t the intended behavior of the system his team built.

“There’s more nuances there than I think the product folks fine-tuning these things further down the line had realized,” Hassabis said in a February interview with The Wall Street Journal hours after Google stopped letting its Gemini chatbot generate images of people.

Hassabis, 47 years old, was the second-best chess player in the world for his age group as a teenager. He helped create the hit videogame “Theme Park” during a gap year before college. As head of the AI research lab DeepMind he oversaw inventions that made the covers of Nature and Science magazines.

His current challenge, by all accounts, will require even greater ingenuity. Since the viral launch of ChatGPT in late 2022, Google has been on a mission to prove it is still at the forefront of a technology that Chief Executive Sundar Pichai has called more profound than the invention of fire or electricity. It is up to Hassabis to make that happen.

“I don’t get much sleep,” Hassabis said from London, his home base for directing Google’s most critical AI research following an overhaul last year by Pichai. That change also put him in charge of employees from a California-based division called Brain that had focused on developing AI systems to improve Google products. They are now part of the renamed Google DeepMind.

After taking over the more than 2,500-person team, Hassabis pushed to accelerate development of large AI systems.

In the new assignment, Hassabis has devoted long stretches at night to video calls with Google leaders in Mountain View, Calif. He previously blocked off the time for independent work and catching up on recently published research, a practice he said he would like to resume at some point.

Hassabis likened Google DeepMind to a “relentless production line.” He has consolidated teams working on overlapping projects and mandated they have leaders from both Brain and DeepMind, said people familiar with the efforts.

The chatbot controversy from February threatens to derail those efforts to re-establish Google as the AI leader. It has also forced Google’s executives to reckon with a corporate structure that has largely kept researchers away from major product decisions.

Hassabis has long resisted leaving his hometown of London. He embraced that distance when leading DeepMind, which operated independently from Google under parent company Alphabet GOOGL 0.04%increase; green up pointing triangle. He butted heads with Google executives over his desire for DeepMind to retain that independence, said people familiar with the matter.

Today, Hassabis is more central than ever to Google’s fortunes. Some inside Google have called in recent weeks for Hassabis to have even more influence over turning his team’s research into consumer products, said one person familiar with the discussions.

“He takes his work very seriously,” said Colin Murdoch, one of Hassabis’s top deputies. “Ever since the first days of DeepMind, there was always this sense—this is an exceptionally promising technology, and we need to be careful.”

Pichai told employees in February the company would make structural changes to address Gemini’s shortcomings, without offering specifics.

Shares in Alphabet, which fell more than 4% after the Gemini controversy, have rebounded recently following news that Apple is considering using Google AI technology to power iPhone features.

A researcher and software engineer by background, Hassabis often makes public appearances wearing understated crew-neck sweaters and dots his sentences with scientific references. He idolizes the British mathematician Alan Turing, an influential figure in modern AI’s early development who shared an interest in the human brain.

“I view myself as Turing’s champion,” Hassabis said at Stanford University last year. He has long told people that he wants to build artificial general intelligence, or AGI, which can accomplish any human task.

People who have worked with Hassabis described him as a demanding, competitive leader with a record of successfully steering researchers toward ambitious goals.

Hassabis began building tech products as a teenager working on videogames. After graduating from Cambridge University with a computer-science degree, he started Elixir Studios, a company that produced a political-simulation game called “Republic: The Revolution.”

“We are re-creating a whole living, breathing country in minute detail,” he told one interviewer when promoting Republic. The game received mixed reviews, and Elixir released one more title before closing two years later.

Hassabis chased even loftier ambitions at DeepMind, which he co-founded in 2010 after completing a cognitive neuroscience doctorate at University College London.

Google purchased DeepMind for about $650 million in 2014 following a bidding war with Facebook, a deal that valued Hassabis’s stake at roughly $100 million.

Encouraged by Google leaders including co-founder Larry Page to pursue AGI with few worries about cost, Hassabis pushed DeepMind to tackle what he called grand challenges. He told employees they should think about building AGI that could invent discoveries worthy of Nobel Prizes, said people who heard the remarks.

DeepMind’s breakthroughs included AlphaGo, the first computer system to beat a professional player of the board game Go. Another technology called AlphaFold, which Hassabis has called his favorite modern application of AI, became the basis for an Alphabet-owned drug-discovery business named Isomorphic Labs that he also oversees. It struck partnerships last year with Eli Lilly and Novartis that could be worth a combined $2.9 billion.

DeepMind’s rise attracted scrutiny in the AI community. Elon Musk, an early investor, became concerned about the dangers of superhuman AI following a meeting with Hassabis in 2012, according to a lawsuit he filed recently against OpenAI.

Hassabis invited Musk to join an ethics board DeepMind had promised to establish following Google’s acquisition, according to that lawsuit. Musk, who had tried to purchase DeepMind himself, labeled the efforts a sham after attending the first meeting, according to the lawsuit.

Google spokeswoman Amanda Carl said Hassabis and Musk discussed both the potential benefits and risks of AI in 2012, and Google had a different recollection of what Musk thought following the ethics board meeting.

As the leader of DeepMind, Hassabis was protective of its independence within Alphabet. Jeff Dean, a longtime Google engineering leader who oversaw the Brain division, in recent years clashed with Hassabis about how their two teams should collaborate, said people familiar with the exchanges.

Among other issues, Dean took issue that DeepMind didn’t readily share research that could inform Brain’s work, the people said.

Carl, the Google spokeswoman, said the merger of the Brain and DeepMind teams has been very smooth. Dean and Hassabis have worked closely together for several years and continue to do so, she said.

While Hassabis steered DeepMind toward achievements like drug discovery, other AI research labs such as OpenAI were devoting greater resources to another goal that ended up grabbing the public imagination: building programs that could produce fluent passages of text in a consumer-friendly chatbot.

The launch of ChatGPT caught Hassabis and DeepMind by surprise, said current and former employees. DeepMind had developed a chatbot called Sparrow that was trained to respond more factually than similar products, but Google executives decided it wasn’t as ready for the public as a similar Brain effort called LaMDA.

When Google released its initial response to ChatGPT, a chatbot called Bard, it used LaMDA instead of Sparrow as its underlying technology. Carl said Google later used lessons from Sparrow to help build Gemini.

Hassabis has called this generation of large AI systems “almost unreasonably effective” and predicted that techniques pioneered by DeepMind will be important for building even more powerful systems.

“This is what I’ve always thought, I just wouldn’t have been able to predict the timing,” Hassabis said about the increasing consumer adoption of AI tools.

Hassabis, together with Dean and others at Google, early last year began discussing combining resources to produce an AI system that could rival OpenAI’s technology. Pichai promoted Hassabis months later, putting him in charge of one of the world’s largest collections of AI researchers.

By December, Google largely caught up to breakthroughs made by OpenAI and its primary backer Microsoft with a set of technologies named Gemini.

Soon after, Google introduced a new version that could analyze 700,000 words of text or up to one hour of video, a breakthrough that surprised many AI researchers.

Google in February renamed its chatbot and a host of other products Gemini, matching the branding of the technology produced by Hassabis’s team.

Then Gemini plunged the company into crisis. Users expressed outrage that Gemini produced images of Black Nazi-era German soldiers and, in some cases, refused to depict white people. Boosted by dozens of tweets from rival Musk, the outrage morphed into a broader backlash against Gemini’s responses to a range of queries.

Morale at Google DeepMind took a hit following the controversy, as researchers complained that the chatbot’s responses were unfairly tarnishing their work, said people familiar with the situation. Google’s initial response attempted to distance the issue from the underlying technology developed by Hassabis’s team.

Google co-founder Sergey Brin said in March the Gemini controversy appeared to result from insufficient testing, speaking during an event at a Hillsborough, Calif., mansion known as the AGI House.

The Gemini app used another technology developed by Google DeepMind, Imagen 2, to produce images. Google then layered additional software on top of the system that translated user queries to trigger more diverse outputs, among other changes.

“It’s a quite vast pipeline of things that have to happen,” Hassabis said. Google is still determining the most efficient way to organize those processes, he said.

Speaking at a conference in Barcelona days after Google pulled the chatbot’s ability to generate images of people, Hassabis said the feature would be restored in a few weeks.

“No matter what you try internally, there’s only a few hundred people testing these things,” Hassabis said in the Journal interview. “It’s not the same as having millions of users testing it and then reporting their findings.”