Around 15 years ago, when DeepMind was founded, almost nobody was working on artificial general intelligence/AI of this kind.

Judged as of around 2010, when DeepMind was founded. True: historical accounts show that AGI research around 2010 was a fringe pursuit with very few dedicated researchers or organisations, and was outside mainstream AI. Mostly true: AGI work was clearly marginal, but a small identifiable community did exist (e.g. AGI conferences, a few labs), so 'nobody' is a modest overstatement within the claim's hedging. Mixed: there was a modest but noticeable body of AGI-focused work. Mostly false: a substantial number of groups were openly working on AGI. False: AGI was a mainstream, well-staffed field in 2010.

Set before the research began.

1. E1 Artificial General Intelligence Scholarpedia · 30 November 2015

   “During the early 2000s interest in the grand goals began to rise in various research centers around the world, including IDSIA in Switzerland, RPI and Carnegie Mellon in the US, and many others.”

2. E2 Artificial General Intelligence Scholarpedia · 30 November 2015

   “in a survey of researchers at the AGI-2010 conference, the majority of respondents felt that human-level AGI was likely to arise before 2050”

3. E3 Conference on Artificial General Intelligence - Wikipedia Wikipedia · 24 February 2013 Own record

   “AGI-2010 University of Lugano, Lugano, Switzerland (In Memoriam Ray Solomonoff and sponsored by AAAI and Ray Kurzweil)”

4. E4 Concept to Challenge: The History of Artificial General Intelligence Brewminate · 26 May 2025

   “While the broader AI community was preoccupied with specialized applications, a smaller group of thinkers revived the foundational dream of building machines with human-level cognitive flexibility.”

The evidence shows that an AGI community existed around 2010, including an AGI-2010 conference [E2, E3] and rising interest at several research centres from the early 2000s [E1]. Only an undated history describes AGI as the work of a smaller group outside mainstream AI [E4], so the evidence does not show how marginal AGI work was around 2010.