If I was them, I would not back up YouTube, but I might carefully scrape and discard.
Hell, if I was a _benevolent_ surveillance program, I'd probably run routine searches for illegal stuff on YouTube, both to find it myself, and to make sure YouTube's tripwires are working.
There is so much low-hanging fruit in terms of "interesting secrets per byte"
Like, I could believe all SMS messages are stored for a year or so.
Some random source says, "Over 6 billion texts are sent every day".
If a text is about 140 characters, and you use a dumb image classifier to transcribe photos as "Nude woman", "nude man", "dick pic", "image macro", "guns", etc., that's only about 1 TB per day, right?
365 TB to keep all US text messages for a year? Maybe my source is wrong. That sounds low. But, it's just text. Maybe it's right.
In fact, the upper bound for all US keyboard input for a year must be below 4.6 petabytes.
(350 million people typing 365 days a year, 16 hours a day, 40 words per minute, 1 bit of entropy per character after compression, 8 bits per word)
Hell, if I was a _benevolent_ surveillance program, I'd probably run routine searches for illegal stuff on YouTube, both to find it myself, and to make sure YouTube's tripwires are working.
There is so much low-hanging fruit in terms of "interesting secrets per byte"
Like, I could believe all SMS messages are stored for a year or so.
Some random source says, "Over 6 billion texts are sent every day".
If a text is about 140 characters, and you use a dumb image classifier to transcribe photos as "Nude woman", "nude man", "dick pic", "image macro", "guns", etc., that's only about 1 TB per day, right?
365 TB to keep all US text messages for a year? Maybe my source is wrong. That sounds low. But, it's just text. Maybe it's right.
In fact, the upper bound for all US keyboard input for a year must be below 4.6 petabytes.
(350 million people typing 365 days a year, 16 hours a day, 40 words per minute, 1 bit of entropy per character after compression, 8 bits per word)