4 ms·

I see it as well. I even see the "subject" of the ban at the top of the page. I agree it's probably something automated that was overturned (although I feel tha

I see it as well. I even see the "subject" of the ban at the top of the page. I agree it's probably something automated that was overturned (although I feel that these AI's are being too heavy-handed, there should have definitely been some human review before taking offline the entirety of the Cornell YT library).

p_l 4y ago
The big issue is that if it wasn't Cornell Library, which can wield considerable weight into requesting a manual review, the reinstatement would have low chance of happening - even if you did hit a human being with your request, they would be overworked and spend barely few seconds before selecting automated negative response, and that's assuming they wouldn't deliberately decide that by whatever personal or cultural mores the content is obscene and warrants destruction.
- vanderZwan 4y ago
  She even makes an indirect point about this when mentioning her earlier experiences: > As one Google rep emailed me in the 1990s, “If you were a 10-figure net worth institution, we woudn’t bother. But you aren’t.”
AzzieElbab 4y ago
there is no way anything "automated" would not ban gayle rubin
shadowgovt 4y ago
It's complicated. Google is (for Cloud at least; couldn't say for certain regarding YouTube) already in the business of identifying certain "high-touch" customers where automatic downing of their service would be embarrassing to Google and so manual review happens. If YouTube doesn't already support that, adding it would be controversial ("Oh, they're playing favorites with content? Come and see the censorship inherent in the system! Help help I'm being repressed!"). If YouTube does support it, Cornell may simply not have made the list yet because building and maintaining that list across millions of accounts is a hell of a cat-and-mouse game. Google is rather heavily incentivized to try and automate as much as possible (and deal with the political blow-back when they stumble) instead of playing the cat-and-mouse game. (As a fun thought experiment: ask yourself "If I were to design the algorithm that wouldn't have kicked Cornell off but would still auto-enforce the TOS, how would I do it?" And then ask yourself "If I were a malicious actor and the algorithm I designed were the one implemented, what kind of content could I create that the algorithm would fail to take down?").

p_l 4y ago

The big issue is that if it wasn't Cornell Library, which can wield considerable weight into requesting a manual review, the reinstatement would have low chance of happening - even if you did hit a human being with your request, they would be overworked and spend barely few seconds before selecting automated negative response, and that's assuming they wouldn't deliberately decide that by whatever personal or cultural mores the content is obscene and warrants destruction.

vanderZwan 4y ago
She even makes an indirect point about this when mentioning her earlier experiences: > As one Google rep emailed me in the 1990s, “If you were a 10-figure net worth institution, we woudn’t bother. But you aren’t.”

AzzieElbab 4y ago

there is no way anything "automated" would not ban gayle rubin

shadowgovt 4y ago

It's complicated. Google is (for Cloud at least; couldn't say for certain regarding YouTube) already in the business of identifying certain "high-touch" customers where automatic downing of their service would be embarrassing to Google and so manual review happens. If YouTube doesn't already support that, adding it would be controversial ("Oh, they're playing favorites with content? Come and see the censorship inherent in the system! Help help I'm being repressed!"). If YouTube does support it, Cornell may simply not have made the list yet because building and maintaining that list across millions of accounts is a hell of a cat-and-mouse game. Google is rather heavily incentivized to try and automate as much as possible (and deal with the political blow-back when they stumble) instead of playing the cat-and-mouse game. (As a fun thought experiment: ask yourself "If I were to design the algorithm that wouldn't have kicked Cornell off but would still auto-enforce the TOS, how would I do it?" And then ask yourself "If I were a malicious actor and the algorithm I designed were the one implemented, what kind of content could I create that the algorithm would fail to take down?").