As the author says in a Reddit summary, "the research community will gradually move from a model-centered research to data-centered research." However, even though we have access to big data, the size of the data sets has remained relatively unchanged, revealing a trend toward data scarcity. And data scarcity becomes more acute when looking at the distribution of data - for example, in research on human faces, we may seem to have abundant data, but that's only because some faces (those of celebrities and stars) appear far more frequently. The author argues that data scarcity is best addressed by focusing on making better use of data, for example, by researching data labeling technology.
Today: 1 Total: 137 [Direct link] [Share]
