this is along the lines of my thinking. people organize and summarize data before throwing it into spreadsheets, where deep learning models do their thing by ge

this is along the lines of my thinking. people organize and summarize data before throwing it into spreadsheets, where deep learning models do their thing by generating new representations from raw data. in a sense, most data in spreadsheets is compressed and deep learning models prefer to find their own compression that best suits the task at hand. or in human terms: "these spreadsheets are garbage. i can't work with this. can you bring me the raw data please?" :)