3 ms·

I'm not arguing with you any more. I'm a software engineer who works on web applications and I know for a fact what you are saying is bullshit. You are arguing

I'm not arguing with you any more. I'm a software engineer who works on web applications and I know for a fact what you are saying is bullshit. You are arguing in bad faith. These are the most profitable companies of all time. Your argument is unreasonable. Later days

manfredo 8y ago
I'm a software engineer working on web applications as well. My company handles data in the exabyte scale. I've spoken in depths with coworkers tasked with filtering illegal or prohibited content, and have done limited amount of work on this system (to be fair, mostly refactoring logging and other orthogonal tasks - I've never introduced new heuristics or changed the filtering logic itself). It's easy to do when it's the same files or data trying to be uploaded, or when there's a way to consistently identify it as prohibited. It's hard when people are actively trying to evade the filtering algorithms. Here is one easy example: a lot of copyrighted work like to shows can be found on YouTube where the copyrighted video is displayed on a smaller section of the screen (e.g. a rectangle in a corner with some animated stuff around the rest of the screen). This is because such editing makes it harder to detect. Maybe YouTube has beaten this countermeasure by now - it's a cat and mouse dynamic by nature. To be fair, this is just my perspective. If someone with greater authority on this subject can give me technical explanation as to why Facebook's 80% preemptively blocked rate is not good I'm happy to hear it. But from what I know, an 80% preemptive block rate on an hours-old video with a significant number of people actively trying to avoid the filtering logic is an impressive feat.
- sdenton4 8y ago
  I have a similar background (worked about a year on some anti-abuse systems), and completely agree. This is an incredibly hard problem space. Machine learning is already hard before you start introducing adversaries...
dang 8y ago
Would you please stop breaking the site guidelines so we don't have to ban you? https://news.ycombinator.com/newsguidelines.html https://news.ycombinator.com/newsguidelines.html

manfredo 8y ago

I'm a software engineer working on web applications as well. My company handles data in the exabyte scale. I've spoken in depths with coworkers tasked with filtering illegal or prohibited content, and have done limited amount of work on this system (to be fair, mostly refactoring logging and other orthogonal tasks - I've never introduced new heuristics or changed the filtering logic itself). It's easy to do when it's the same files or data trying to be uploaded, or when there's a way to consistently identify it as prohibited. It's hard when people are actively trying to evade the filtering algorithms. Here is one easy example: a lot of copyrighted work like to shows can be found on YouTube where the copyrighted video is displayed on a smaller section of the screen (e.g. a rectangle in a corner with some animated stuff around the rest of the screen). This is because such editing makes it harder to detect. Maybe YouTube has beaten this countermeasure by now - it's a cat and mouse dynamic by nature. To be fair, this is just my perspective. If someone with greater authority on this subject can give me technical explanation as to why Facebook's 80% preemptively blocked rate is not good I'm happy to hear it. But from what I know, an 80% preemptive block rate on an hours-old video with a significant number of people actively trying to avoid the filtering logic is an impressive feat.

sdenton4 8y ago
I have a similar background (worked about a year on some anti-abuse systems), and completely agree. This is an incredibly hard problem space. Machine learning is already hard before you start introducing adversaries...

dang 8y ago

Would you please stop breaking the site guidelines so we don't have to ban you? https://news.ycombinator.com/newsguidelines.html https://news.ycombinator.com/newsguidelines.html