The claim is poorly phrased in the headline. From the article (https://royalsocietypublishing.org/doi/10.1098/rsos.181393 https://royalsocietypublishing.org/doi

The claim is poorly phrased in the headline. From the article (https://royalsocietypublishing.org/doi/10.1098/rsos.181393 https://royalsocietypublishing.org/doi/10.1098/rsos.181393), you are exposed to around 2000 bits of language-related information each day, which apparently adds up to about 1.5 Mb by the time you turn 18 (an entirely arbitrary cutoff in the world of studying language acquisition). This has some interesting interplay with the idea that your brain has a lot of the necessary machinery for learning language built-in from birth. We aren't exposed to that much data before we get reasonably good at language (https://en.wikipedia.org/wiki/Poverty_of_the_stimulus https://en.wikipedia.org/wiki/Poverty_of_the_stimulus). That helps account for the fact that you are exposed to so little (around a flop-disk's worth, according to the authors) data, but learn so quickly. Most linguists (but not all) subscribe to this idea, known as UG (https://en.wikipedia.org/wiki/Universal_grammar https://en.wikipedia.org/wiki/Universal_grammar). What's interesting is that most UG people argue that your brain encodes morphosyntactic stuff via UG, and the authors of this paper state that most of the data they're counting in their 1.5 Mb figure is lexical semantic info.