Research advances from the Brain and DeepMind teams, from Deep Reinforcement Learning to Transformers, laid much of the foundations of the current AI industry.

True if the evidence shows that Google Brain and/or DeepMind produced foundational work in both areas named: deep reinforcement learning (for example DQN and AlphaGo by DeepMind) and the Transformer (Vaswani et al. 2017, mainly by Google Brain/Google Research authors), and that these are widely regarded as foundations of the AI industry as of April 2023. Mostly true if both attributions hold but the Transformer paper's authorship is mainly Google Research rather than Brain alone, or 'much of the foundations' is a modest overstatement. Mixed if only one of the two named advances can fairly be credited to these teams. Mostly false if the teams' contributions were real but minor next to others, so that the claim substantially overstates them. False if neither advance came from these teams.

Set before the research began.

1. E1 Human-level control through deep reinforcement learning - Nature Nature · 25 February 2015

   “Mnih, V., Kavukcuoglu, K., Silver, D. et al. Human-level control through deep reinforcement learning. Nature 518, 529–533 (2015).”

2. E2 Human-level control through deep reinforcement learning - Nature Nature · 25 February 2015

   “and the rest of the DeepMind team for their support, ideas and encouragement.”

3. E3 Mastering the game of Go with deep neural networks and tree search - Nature Nature · 27 January 2016

   “Silver, D., Huang, A., Maddison, C. et al. Mastering the game of Go with deep neural networks and tree search. Nature 529, 484–489 (2016).”

4. E4 Attention Is All You Need ar5iv (arXiv mirror)

   “We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely.”

5. E5 Attention Is All You Need ar5iv (arXiv HTML mirror)

   “Provided proper attribution is provided, Google hereby grants permission to reproduce the tables and figures in this paper solely for use in journalistic or scholarly works.”

6. E6 Attention is All you Need NeurIPS Proceedings (31st Conference on Neural Information Processing Systems, NIPS 2017) · 2017

   “Ashish, with Illia, designed and implemented the first Transformer models and has been crucially involved in every aspect of this work.”

7. E7 Mastering the game of Go with deep neural networks and tree search - Nature Nature (Springer Nature) · 27 January 2016 Own record

   “Using this search algorithm, our program AlphaGo achieved a 99.8% winning rate against other Go programs, and defeated the human European Go champion by 5 games to 0.”

8. E8 About Google DeepMind Google DeepMind Own record

   “In 2017, Brain invented the Transformer architecture, an elegant system of neural networks that underpin almost all large language models and revolutionized the field of AI.”

9. E9 About Google DeepMind Google DeepMind Own record

   “One of its early breakthroughs was a program called DQN, which learned to play 49 different Atari games from scratch just by observing the raw pixels on the screen and being told to maximize the score.”

The 2015 Nature paper on deep reinforcement learning (DQN) thanks "the rest of the DeepMind team" (E1, E2), and the AlphaGo paper (E3, E7) reports a program that defeated the European Go champion 5 games to 0. Together they credit DeepMind with foundational deep reinforcement learning work. The Transformer paper (E4, E6) carries a Google copyright notice (E5), and its author footnote describes the first Transformer models being built (E6). However, the quoted text does not show that Google Brain alone wrote it. The pointer note for E6 says the authors were split between Google Brain and Google Research, with one from the University of Toronto, but the quote itself does not contain those affiliations. The statements that Brain invented the Transformer, which underpins almost all large language models, and that DQN was an early DeepMind breakthrough come from the subject's own organisation (E8, E9). So both attributions hold broadly, but the Brain-specific credit and the "much of the foundations" framing are only partly backed by independent evidence. That fits "mostly true" under the criteria.