The Thinking Game

"How do you keep power forever over something that's much more powerful than you?" - Prof. Stuart Russell

This very important question from the trailer is unfortunately not what The Thinking Game is about.

While Demis Hassabis acknowledges loss of control and extinction risk from AGI elsewhere, in The Thinking Game he calls the technology itself neutral.

DeepMind co-founder Shane Legg does somewhat better, saying that he worries the opposite of what investors and colleagues worry: Legg thinks that AGI may come too soon before society can prepare rather than too late such that working on it is a waste of time.

Multiple people interviewed including Legg and VP of Research Raia Hadsell state that DeepMind is caught up in an AGI race. It is said that AlphaGo was the Sputnik moment for China. PauseAI protesters are briefly shown in London with signs to the effect of "If you can't steer it, don't build it."

The engaging documentary is reminiscent of the documentary AlphaGo (2017), except it starts earlier covering Hassabis' background as chess prodigy realizing in 1988 at age ~12 after resigning in a drawn position 10 hours into a game of chess that spending his life working on mastering chess would be a waste of talent, and finishes later with the story of DeepMind solving protein folding and Hassabis winning the Nobel prize.

There is an emotional scene at the end where someone in an Alpha Fold meeting says that they could fold all the proteins in a month and Hassabis says that it's a great idea, why didn't anyone suggest it before, why don't they just fold all the proteins and publish it. The good this does for the medicine and the world is immense and in my opinion Hassabis's Nobel Prize was well deserved.

There was not much new that followers of Hassabis and DeepMind in the media did not already know, but for me the film strengthened my conviction that this deeply likeable man in this incredible position of power and influence over the entire future, for all the good he has done and hopefully will continue to do, is not being as vocal and forthright about the existential risk that AGI poses.

I'm just a layperson and everything I say about the film is just from memory after having just watched it, but I'll leave you with this Hassabis quote from October 1, 2024:

"And so it would be immensely powerful, but it has to be handled with care and I think that we just don't know. So the reason I signed that letter ("Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war") is I just wanted to give some weight against the sort of Pollyannaism of like there's nothing here to see, that actually there's some unknown risk, and we need to define. We've got, I think we've got time, not, but 10 years is not a lot of time for something that monumental coming down the road. So we need to do a lot more research on things like controllability, understanding what these systems do on a theoretical level, and you know very important things like how do we define goals for the systems and values for the systems, and how do we make sure they stick to them. These are all kind of unknowns with the current nascent technology.

"So I would say I'm a cautious optimist. So I think we'll solve it if we get our act together, we do this internationally, put all the best minds on it, we get going now-I'm very happy to see the AI Safety Institutes that have been set up in in UK and US and we were big advocates of that happening-and testing the latest models, but we need a lot more of that. So I was just sort of encouraging that really to happen. And I think given sufficient time with sufficient brain power-I believe in human ingenuity-I think we'll get this right. But, there are, you know, there are risks and we can't cut corners and we need to treat it, I think, with the respect and almost the reverence, I think, that this technology deserves, that we're on the cusp of."

Hassabis gives a lot of qualifications after "I think we'll solve it." That worries me a lot given that if we don't solve it, if we don't get AI alignment right on the first critical try, then everyone dies.

I would really like Demis Hassabis and everyone else to pre-order the book If Anyone Builds It, Everyone Dies: Why Superhuman AI Would Kill Us All by Eliezer Yudkowsky and Nate Soares, coming September 2025.

I am not confident they are right, but it seems like they very well might be, and I'd feel a lot better about the future if Hassabis had acknowledged that an AI extinction event in our lifetimes is a very real possibility in the documentary. The world still needs to wake up and confront the reality of the situation we find ourselves in. Thanks for reading, and check out the blurbs for "If Anyone Builds It, Everyone Dies."

Heading into the DeepMind documentary, I admittedly carried a degree of skepticism, half-expecting a polished corporate showcase replete with Google branding and perhaps even a cameo from CEO Sundar Pichai. It was, therefore, a pleasant and commendable surprise to find the focus remained steadfastly on the science and the people driving it. By largely eschewing overt corporate promotion, the film gains significant credibility, allowing the gravity of DeepMind's achievements to resonate more authentically, feeling less like an advertisement and more like a genuine exploration of groundbreaking research.

The true highlight, and arguably the core substance of the narrative, revolves around the revolutionary work on protein folding with AlphaFold. Witnessing the explanation of this breakthrough is genuinely awe-inspiring. What DeepMind has accomplished here isn't merely an incremental advancement; it's a monumental leap that addresses one of biology's grand challenges - predicting the three-dimensional structure of proteins from their amino acid sequences, a puzzle that has perplexed scientists for decades. The potential implications are staggering, promising to accelerate drug discovery, deepen our understanding of diseases, and potentially revolutionize fields across medicine and biology.

This focus on fundamental scientific discovery powerfully recontextualizes the current frenzy surrounding large language models (LLMs). While technologies like ChatGPT and their ilk are undeniably impressive feats of engineering, demonstrating remarkable capabilities in text generation and human-like interaction, the documentary subtly underscores a difference in the magnitude and nature of the problems being solved. Compared to unraveling the intricate machinery of life itself, the (admittedly sophisticated) pattern matching and predictive text capabilities of many contemporary LLMs begin to feel, relatively speaking, less profound. It's not to dismiss the utility or ingenuity of LLMs, but AlphaFold's success serves as a potent reminder that AI's greatest promise might lie in tackling foundational scientific questions with tangible, potentially life-altering consequences, rather than solely mimicking human conversation. The film certainly recalibrated my perspective on the current AI landscape and its most impactful applications.

The engaging story of how one of the most profound scientific discoveries where made by a group of people not being humiliated by trying to achieve the impossible.

I recommend this documentary to anyone interested in science and our future. I will watch it again together with my wife on her birthday, as this is the kind of story that gives her hope and warms her heart. The story and the achievement is a huge inspiration. It is a good thing that we have organizations that choose to fund progress like this.

And, had all scientist and leaders in the world been as considerate and reflected as these people, the world would probably be a better place for everyone.

The film chronicles Hassabis's evolution from a child chess prodigy to a visionary scientist. His early fascination with games and the human mind laid the foundation for DeepMind's mission: to create AI that can think and learn like humans. Through archival footage and candid interviews, viewers witness the challenges and triumphs that marked DeepMind's path, including the development of AlphaGo and the groundbreaking AlphaFold, which solved a 50-year-old biological puzzle .

The Thinking Game is a thought-provoking documentary that offers an intimate look into the world of AI research. While it could have explored the ethical dimensions more thoroughly, its strengths lie in its engaging storytelling and ability to make complex topics accessible. For anyone interested in the future of technology and the minds shaping it, this film is a must-watch.