The Thinking Game (2024), and Google DeepMind story

From the time Google’s DeepMind cracked one of humanity’s toughest scientific problems, protein folding, in 2020, global attention shifted firmly towards the company and its leadership. That attention only grew when Demis Hassabis and his colleague John M Jumper won the Nobel Prize in Chemistry in 2024. Even before all this, DeepMind had stunned the world by defeating the reigning world champion at Go, one of the most complex board games ever created. Given these milestones, it is natural that there is curiosity about DeepMind and about Demis Hassabis. This documentary, co-produced by the company itself, sets out to address that curiosity. The Thinking Game (2024) is available for free on YouTube. I would recommend watching it. Mangoidiots rates it Ripe.

The film traces the journey of DeepMind, with a brief look into Demis Hassabis’s childhood. Much of what is shown here has already been shared by Demis in earlier interviews and by his co-founders when speaking about the company. Still, it is useful to see his long-term clarity of thought, especially during his time at Cambridge, where he speaks about the need for a General Intelligence. As a child, he was a candidate master in chess. During a gap year from Cambridge, he worked at Bullfrog, a video game company, where he explored simulating real human behaviour inside the Theme Park game environment.

The documentary then moves into how DeepMind was funded and how the company got started. It touches on the early days of the team and briefly covers the acquisition by Google. It shows how their early AI systems were trained to play Atari video games, first using human input, and later through fully autonomous learning with no human intervention. When DeepMind achieved this, I remember reading an article in IEEE Spectrum and being completely taken in by the elegance of the design they described. The film then moves on to DeepMind’s decision to participate in the CASP competition to tackle the protein folding problem. It covers the steady progress, leading up to the major breakthrough in 2020, when AlphaFold made a dramatic leap in prediction accuracy. The documentary closes with a title card referencing the Nobel Prize.

At about 84 minutes, this is all a single documentary can realistically cover. DeepMind is a company with an unusual story, and compressing fifteen years of such a meteoric journey is never going to be easy. I did wish the film had spent more time on the challenges and setbacks, and on how the team navigated critical moments. Without those, the story feels too smooth, like a straight road with no bends, bumps, or near misses. In reality, we know that could not have been the case.