Google DeepMind

On ToTheMoon, Google DeepMind rarely appears as a product story. The hosts bring it up as one of the laboratories they use to gauge how fast the race toward AGI is moving, and as the home of Demis Hassabis, whose caution about timelines they set against other lab leaders. The recurring question is what is already happening inside such labs and how much their current models help build the next ones.

Earlier mentions are tied to results. In episode 068 (2025) the hosts recalled that DeepMind had entered the math olympiad a year before and come away with silver, while labs were now taking gold. In episode 073, DeepMind together with Google Cloud was cited as the reason Apple could buy a Google model but not an organization that ships strong AI products regularly. By 2026 the focus had shifted to definitions: episode 102 covered DeepMind's attempt to formalize AGI levels, with the hosts hoping for automated checks against that scale while noting that for ordinary users the race is decided on context and daily usefulness.

The most detailed treatment comes in episodes 120 and 132. Episode 120 described DeepMind as a lab where a large share of code is written with AI and algorithms are proposed, tested and improved in a loop. Episode 132 walked through the Anthropic and Google DeepMind discussion on the future of AGI: Dario Amodei looks at practical productivity, whereas Hassabis separates high task performance from general intelligence. Episode 133 returned to that stance, noting that the head of DeepMind says AGI is in some sense already here and at the same time not yet.