#NeurIPS2021 invited talks round-up: part one – Duolingo, the banality of scale and estimating the mean