The head of DeepMind warns Elon Musk: If there's a problem with AI, going to Mars won't help either.

Demis Hassabis is a tech mogul who hasn't been talked about enough.

After Google merged DeepMind and Google Brain in April 2023, he has been at the helm of the newly born "Google DeepMind" and become a member of Google's core power circle.

Since then, Gemini has continuously iterated its models and been fully integrated into the Google ecosystem. It has become the most powerful large - scale model, and its user base has skyrocketed with the support of the Google ecosystem. Meanwhile, Google DeepMind continues to make breakthroughs at the academic forefront. AlphaFold 3, which was successfully launched to the public last year, inherits the "Pairformer + diffusion model" architecture of its predecessor and can predict the structure of protein complexes (including DNA, RNA, ligands, etc.), providing new tools for drug discovery and molecular biology research. As of May 2025, its research has been published in Nature with over 4000 citations.

At the end of last year, Hassabis became one of the recipients of the Nobel Prize in Chemistry because of the profound impact of AlphaFold on the field of biochemistry.

In a recent exclusive interview, he said that the revolution brought about by AI is 10 times greater and 10 times faster than the Industrial Revolution and will have a profound impact on our society.

He has half - Chinese ancestry. His mother is a Chinese - Singaporean, and his father is Greek. In a family full of artistic atmosphere, he, who was obsessed with chess and computers, was once an outlier in the family.

How did he transform from a genius chess player into a key figure in Google's AI business, a global tech giant, in 49 years? Why was DeepMind founded, and why did it join Google? What exactly are his views on AGI now?

Moreover, does he have a pessimistic or optimistic vision for the "post - AGI era"?

Recently, The Guardian conducted an exclusive interview with Hassabis. The following is a compilation and translation by Facing AI. We added questions and integrated relevant content.

What's it like to win the Nobel Prize and become an "AI ambassador"?

"I felt I had to speak out publicly."

Receiving the trophy from the King of Sweden, Hassabis felt it was like a dream. He said he isn't very good at enjoying the moment. "I've won awards before, and I always wondered 'what's next'. But this award is truly special. It's something he's dreamed of since childhood.

Becoming an "AI ambassador" is something Hassabis never expected.

He said bluntly:

"If I could have had my way, I would have kept it (AI) in the lab for a longer time, doing more things like AlphaFold and maybe even curing cancer. But that's how AI is. It also has its own advantages.

It's great that everyone can access the latest AI and experience it firsthand. In fact, it's beneficial to society, helping it to standardize and adapt to it, and also facilitating government discussions about it... I felt I had to speak out publicly, especially from a scientific perspective, about how we should handle this issue, think about the unknown, and make it more understandable."

How did he get into the field of AI?

"I was kind of an outlier in my family."

Chess led Hassabis onto the path of thinking. Between the ages of 4 and 13, Hassabis represented the English youth team in competitions:

"Competing at such a young age has a crucial impact on how your brain works. Many of my ways of thinking have been influenced by the strategic thinking and stress - handling in chess."

On paper, Hassabis' background hardly predicted his future. His family has a more artistic atmosphere:

"My father just finished writing a musical after retirement, which was staged at an art theater in North London. My sister is a composer, so I was kind of an outlier in my family." His family wasn't poor, but they weren't super - rich either. He attended several public schools in North London and was also homeschooled for a few years.

He said he also felt a bit out of place at school, but he seemed to know his goals clearly. His childhood idols were scientific pioneers like Alan Turing and Richard Feynman. He used the money he won from chess to buy early home computers, such as the Sinclair ZX Spectrum and Commodore Amiga, and learned programming:

"In the late 1980s, few people were interested in computers. A group of us used to do some hacking, make games and other things, and later this became my next career after chess."

As early as the 1990s, the gaming industry began to use AI.

At the age of 17, Hassabis programmed the hit game "Theme Park", in which players need to build a virtual amusement park. The game would react according to the players' actions. If you placed a snack stand too close to the exit of a roller - coaster, the virtual tourists would start vomiting.

Why did he found DeepMind, and why did it join Google?

"What if something goes wrong with AI? Then landing on Mars won't help you."

Hassabis studied computer science at the University of Cambridge and then obtained a doctorate in neuroscience from University College London. In 2010, he co - founded DeepMind with Shane Legg, also a post - doctoral neuroscientist, and his former classmate Mustafa Suleyman.

Hassabis said that DeepMind's mission is simple:

"Solve the problem of intelligence first, and then use it to solve everything else."

DeepMind quickly caught the attention of Silicon Valley. In 2014, the team demonstrated an AI that could learn to play Atari video games, such as Breakout, without any prior knowledge.

Some well - known tech giants today also became interested in it, including Peter Thiel (an early investor in DeepMind), Google, Facebook, and Elon Musk.

Hassabis first met Musk in 2012. During a lunch at SpaceX's factory in California, Musk told Hassabis that his top priority was to land on Mars.

But Hassabis had a different opinion: "(Mars is) a backup planet for Musk in case something goes wrong (on Earth). I don't think he thought much about AI at that time."

Hassabis pointed out the flaw in his plan. "I said, 'What if something goes wrong with AI? Then landing on Mars won't help you because if we get there, AI can obviously reach there easily through our communication systems or something.' He just hadn't thought about this problem. So he sat there for a minute, said nothing, and just thought, 'Well, this might be true.'"

Shortly after that, Musk also became an investor in DeepMind.

In 2014, Google acquired the company for £400 million, and Musk switched his support to OpenAI.

Hassabis recalled that what prompted DeepMind to join Google wasn't money or hardware. Google founders Larry Page and Sergey Brin, like Hassabis, are computer scientists, and they "ultimately see Google as an AI company," Hassabis said. He has also used products like Gmail and Google Maps.

"In the end, I think Google's mission - to organize the world's information - is a really cool mission."

How did ChatGPT affect DeepMind?

"When you're too close to the technology, you might not see its potential clearly."

DeepMind's office has always been in London, UK.

Looking out from Hassabis' office now, you can see Google's new office under construction, a beige building. DeepMind will move to the new building next year. The tech giant has invested so much in the UK because Hassabis insisted on staying in London.

"Our first supporters said, 'You have to move to San Francisco,' but I wanted to prove it could work here," he said. "I knew there was untapped talent here. And I knew how important AI would be to the world if we succeeded, so I thought it was important to adopt a global strategy, not just be confined to 100 square miles (about 100 square kilometers) in Silicon Valley. I still believe this is important."

In 2016, DeepMind's AI defeated one of the world's top Go players. Go is a board game much more complex than chess, which once again caught the attention of the tech world.

The breakthrough of AlphaFold in protein structure was another leap: DeepMind has now analyzed the structures of over 200 million proteins and made the relevant resources public.

But with the release of OpenAI's ChatGPT in 2022, the AI landscape changed dramatically. ChatGPT caught the public's imagination with its incredible abilities, being able to solve a wide range of problems from strategic planning to writing poems. ChatGPT caught large tech companies off - guard, especially Google.

"They really went all - out, almost 'betting the farm', which is very impressive. Maybe that's what a startup has to do.

We leading labs actually have very similar systems, but we can also see their flaws, such as occasional 'hallucinations'.

I don't think anyone, including OpenAI, really expected it to generate such amazing use - cases and bring so much value to users. This incident reminds us that when you're too close to your own technology, you might not see its potential clearly."

When will AGI arrive?

"We should live in a world that I sometimes call 'extreme abundance'."

The race has begun. As Hassabis said, DeepMind has now become the "engine room of Google", with AI embedded in every corner of its business.

Meanwhile, competitors are accelerating their pursuit.

Companies like Mark Zuckerberg's Meta, Amazon, Apple, and Microsoft are making large - scale investments and poaching talent from rivals. Zuckerberg even offers an annual salary of $100 million to top researchers.

Mustafa Suleyman, who left DeepMind in 2019, now heads Microsoft AI and has just poached more than 20 engineers from DeepMind.

Hassabis isn't eager to call his former friend a competitor: "Our directions are completely different. I think he's more focused on commercial applications, while we still focus on cutting - edge research."

The upcoming frontier is undoubtedly AGI (Artificial General Intelligence) - the key point when AI reaches the level of human intelligence.

"I don't know if it will be a sudden moment or just a gradual process. But within the next five to ten years, perhaps at the lower end of this time frame, we will eventually have something that can reasonably be called AGI, which exhibits all the cognitive abilities of humans."

In other words, we're in the last few years of the pre - AGI civilization, after which everything may be different. Some people see it as the end of the world, but Hassabis sees it as a utopia.

"Assuming we safely and responsibly deliver resources to the world, and obviously we're also trying to play our part, then we should live in a world that I sometimes call 'extreme abundance'."

Hassabis' vision of the future includes medical progress, room - temperature superconductors, nuclear fusion, and progress in materials and mathematics.

"This should bring amazing productivity and promote social prosperity. Of course, we must ensure that resources are fairly distributed, but that's more of a political issue. If so, we might enter an amazing world of abundance for the first time in human history, where everything is no longer a zero - sum game. If all this really happens, we should really start interstellar travel."

What about the dark side of AI? Is the future really all rosy?

"As long as we have time, I believe in human wisdom."

There are many dark sides to AI, or rather, there are still many challenges to overcome in the future.

For example, there's the problem of resource consumption. The amount of water and electricity required by future AI data centers is predicted to be astronomical, especially when the world is facing droughts and a climate crisis. Before the day when AI cracks nuclear fusion arrives, we might not have a planet to live on.

Regarding this, Hassabis believes that the rewards AI brings to humanity will far exceed its consumption:

"There are many solutions to this. Yes, AI systems do require a lot of energy. But even if they're only used narrowly for climate - related solutions, the rewards from these models will far exceed the energy they consume."

There's also a concern that "radical abundance" is just another way of saying "mass unemployment": AI has started to replace human jobs. When we "no longer need to work" - as many tech companies promise - does it really mean we're handing over economic power to those who control AI?

"That will be one of the biggest problems we have to solve," he admitted. "Assuming we really achieve radical abundance and distribute it well, then what will happen next?"

Hassabis has two teenage sons, and he also wonders about their future: "It's a bit like when I was growing up and home computers were just emerging. Obviously, this time the scale is much larger, but you have to embrace new technologies... If you can become an expert in these tools and use them as skillfully as a ninja, it will really empower those who are good at using them."

Of course, those who aren't "ninjas" also have a place: "We need good philosophers and economists to think about what the world should be like when such things happen. What's the purpose? What's the meaning?"

He pointed out that many of our actions aren't for practical purposes, such as sports, meditation, and art.

"As a society, we'll be more deeply involved in these fields because we'll have the time and resources to do so."

So, is he completely without concerns about the future?

"I'm a cautious optimist.

Generally speaking, as long as we have time, I believe in human wisdom. I think we'll do things right. I also think humans have infinite adaptability.

I mean, look at where we are today. Our brains evolved for a hunter - gatherer lifestyle, and now we're in modern civilization.

The difference this time is that this change will be ten times greater and perhaps ten times faster than the Industrial Revolution."

He admitted that the Industrial Revolution wasn't smooth for everyone, but it was necessary and irresistible:

"We wouldn't wish it hadn't happened just because it brought some shocks. Obviously, we should try our best to reduce that chaos, but change will come - hopefully for the better."

This article is from the WeChat official account "Facing AI". The author is Bi Andi, and the editor is Hu Run. It is published by 36Kr with permission.