Uh, no thanks. You can pry my JSON and YAML out of my cold, dead hands.
For tabular data? JSON and YAML are great for serialization, but they both have a lot of extra punctuation, and require keys for each field unless you’re using arrays which makes the files slightly larger.
Seems like a perfectly acceptable tradeoff to me, and it’s pretty much negligible unless you have a large number of empty string columns. And yes, you’d want to use arrays to make a fair comparison with CSV.