Notice that they are not actually providing the raw data; they are providing access to a dataset that has been anonymized using differential privacy. This is a way of adding noise to the data that makes it impossible to tell (much better than random guessing) whether a person is in the dataset, even by combining the dataset with external data.
Is anyone else a bit surprised at these wages, on the high side? I mean certainly it's not a lot hourly and there must be a lot of uncertainty (writing pieces on spec etc) but Buzzfeed giving $.50 a word for a 6000 word article is $3000... Which is better than I had assumed. But maybe that's just me.
Something about this syndrome reminds me of tinnitus. Tinnitus often occurs due to destruction of the cilia in the ear canal, causing the brain to overcompensate by constantly sensing the frequencies the destroyed cilia would have responded to. (At least, this is one commonly hypothesized theory.) While this may be a bit of reasoning by analogy, I wouldn't be surprised if something similar were at work (particularly to the extent that nerves or other sensitive apparatuses are involved).
Certainly it's still far from being able to deceive a human into thinking synthesized speech of any speaker saying anything is real, but it has definitely and clearly capture a certain quality to each of those voices. Really cool project and I'm sure it portends even more awesome work in the area.