4 ms·

The issue is less about detection and more about the sheer volume of content. Even if your false positives and negative rates are low the numbers are huge. Whic

The issue is less about detection and more about the sheer volume of content. Even if your false positives and negative rates are low the numbers are huge. Which is why everyone ends up hiring more and more people as content volume increases. What has changed over the years is - the cost to spam the entire world with zillion messages a second is tending to ZERO. While this is an achievement it's also a disaster cause it's akin to everyone having an almost free high power broadcast transmitter sitting on their roof. We regulate and auction radio spectrum and don't allow everyone on the planet to broadcast simultaneously on what ever frequency they want. Cause the noise would make signal drop to 0. But on the internet it's allowed and it's almost free.

j16sdiz 3y ago
Agree. At this volume, nobody could possibly win. Social networks need to break down quite a bit to get something manageable.
phkahler 3y ago
>> What has changed over the years is - the cost to spam the entire world with zillion messages a second is tending to ZERO. One solution to that is whitelisting. But we don't want 3rd party censorship, so a federated system where we rate sources and content ourselves seems like a good idea. Unfortunately kids won't filter anything - they are drawn to the unusual and forbidden.
- chongli 3y ago
  How is whitelisting going to work? You can’t whitelist Instagram photos because that would prevent people from posting new ones. You also can’t whitelist users because bad actors just hack their accounts. All of the disallowed content I’ve come across (pornography on Facebook or Instagram) was posted by an obviously hacked account. If you add an extra checkmark people have to get to be able to post then hackers will just target those people.
  - shadowgovt 3y ago
    It turns out this is actually a very hard problem which is why centralized social media companies get it wrong so often (and why that might still be preferable to the alternatives).
  - rightbyte 3y ago
    > How is whitelisting going to work? Facebook and Instagram used to work like that. It only showed you stuff you choose to want to see from those you followed. But there is more money in forcing stuff on people I guess. Hence the algorithms.
    - saurik 3y ago
      The person you are responding to is trying to obtain a 100% success rate on filtering, and explicitly discusses the Facebook model as insufficient as the people you follow might suddenly become porn (due to their account being hacked).
  - saurik 3y ago
    I think the real problem here is that you seem to have a zero tolerance policy on bad content: yes, peoples' accounts get hacked... but that sucks for them so much more than for you, and yet you seem concerned that your eyes were forced to see some porn? And look, I get that we are talking about children as the viewers, not you... but children simply shouldn't be on the Internet browsing random stuff in the first place!! The whitelist can then be all encompassing, as it should be their parents showing them stuff.
    - chongli 3y ago
      Please don't project your feelings onto me. I was stating an observation I made in a completely neutral way. I'm not concerned about it at all.
    - mikrotikker 3y ago
      How hard is it to classify images with AI and put a NSFW blue for underaged accounts. We had that stuff back in 2005 scanning corporate emails.
- mikrotikker 3y ago
  I was looking at rotten.com when I was like 14. I could almost handle the person gore but the animal gore affected me and still does if I see it today.
seer 3y ago
Hmm the now old post by Zed Shaw from the dawn of ruby on rails era - “saving the internet with hate” comes to mind. https://weblog.masukomi.org/2018/03/25/zed-shaws-utu-saving-the-internet-with-hate/ https://weblog.masukomi.org/2018/03/25/zed-shaws-utu-saving-... Basically baking in a reputation system into network protocols, to deter bad actors… not sure how viable a solution it is, just fun to think how far back people were thinking about solutions to these problems.
- layer8 3y ago
  Any reputation system can and will be gamed. There are parallels to SEO.

j16sdiz 3y ago

Agree. At this volume, nobody could possibly win. Social networks need to break down quite a bit to get something manageable.

phkahler 3y ago

>> What has changed over the years is - the cost to spam the entire world with zillion messages a second is tending to ZERO. One solution to that is whitelisting. But we don't want 3rd party censorship, so a federated system where we rate sources and content ourselves seems like a good idea. Unfortunately kids won't filter anything - they are drawn to the unusual and forbidden.

chongli 3y ago
How is whitelisting going to work? You can’t whitelist Instagram photos because that would prevent people from posting new ones. You also can’t whitelist users because bad actors just hack their accounts. All of the disallowed content I’ve come across (pornography on Facebook or Instagram) was posted by an obviously hacked account. If you add an extra checkmark people have to get to be able to post then hackers will just target those people.
- shadowgovt 3y ago
  It turns out this is actually a very hard problem which is why centralized social media companies get it wrong so often (and why that might still be preferable to the alternatives).
- rightbyte 3y ago
  > How is whitelisting going to work? Facebook and Instagram used to work like that. It only showed you stuff you choose to want to see from those you followed. But there is more money in forcing stuff on people I guess. Hence the algorithms.
  - saurik 3y ago
    The person you are responding to is trying to obtain a 100% success rate on filtering, and explicitly discusses the Facebook model as insufficient as the people you follow might suddenly become porn (due to their account being hacked).
- saurik 3y ago
  I think the real problem here is that you seem to have a zero tolerance policy on bad content: yes, peoples' accounts get hacked... but that sucks for them so much more than for you, and yet you seem concerned that your eyes were forced to see some porn? And look, I get that we are talking about children as the viewers, not you... but children simply shouldn't be on the Internet browsing random stuff in the first place!! The whitelist can then be all encompassing, as it should be their parents showing them stuff.
  - chongli 3y ago
    Please don't project your feelings onto me. I was stating an observation I made in a completely neutral way. I'm not concerned about it at all.
  - mikrotikker 3y ago
    How hard is it to classify images with AI and put a NSFW blue for underaged accounts. We had that stuff back in 2005 scanning corporate emails.
mikrotikker 3y ago
I was looking at rotten.com when I was like 14. I could almost handle the person gore but the animal gore affected me and still does if I see it today.

seer 3y ago

Hmm the now old post by Zed Shaw from the dawn of ruby on rails era - “saving the internet with hate” comes to mind. https://weblog.masukomi.org/2018/03/25/zed-shaws-utu-saving-the-internet-with-hate/ https://weblog.masukomi.org/2018/03/25/zed-shaws-utu-saving-... Basically baking in a reputation system into network protocols, to deter bad actors… not sure how viable a solution it is, just fun to think how far back people were thinking about solutions to these problems.

layer8 3y ago
Any reputation system can and will be gamed. There are parallels to SEO.