5 ms·

Isn't that typically done to show thumbnail previews under the pasted links?

by 0x0 14y ago

choult 14y ago
Facebook generally will present the OpenGraph data from pasted links; for sites lacking it they might present the first image on the page. It would be vast overkill to download a whole video just to present a thumbnail, and the user experience in the delay while the whole file is downloaded and processed would render the feature fairly useless. Which leaves the question of why Facebook would do this open... It might just be really bad practice and an under-/non-optimized crawler.
- Cthulhu_ 14y ago
  Or they download the whole video to pull it through an analyzer to look for stuff prohibited to be republished via FB, such as child pornography. If I were FB, I'd screen everything published via my site.
  - pyre 14y ago
    Right, because we have automated software child pornography recognizers.
    - ceejayoz 14y ago
      It wouldn't be surprising. YouTube looks for video fingerprints of copyrighted data. Repurposing such technology to find copies of known CP videos would be pretty straightforward. Wouldn't help you with new material, of course.
      - pyre 14y ago
        If you start filtering for child pornography, and people know about it, they will come at your hard if you miss anything. Even for a technology that gets all known child porn videos, people would be beating down the door to sue/bad mouth Facebook for missing the new ones. Mostly due to people being people, but also because the technology is 'magical' to far too many people in society so they feel like there are no limitations (or it's hard for them to grasp them).
    - runamok 14y ago
      They already have image algorithms to detect nudity with a good degree of accuracy so that's probably enough to forbid an image or video.
      - pyre 14y ago
        Like that won't cause a shitstorm. A social networking site that won't let people trade possibly nude images privately?
        kamjam 14y ago
        I don't think he's talking about necessarily private, but they detect nudity on the "publicly" available images available in your albums which your friends an see... http://petapixel.com/2012/11/25/facebook-removes-risque-photograph-of-woman-showing-an-elbow/ http://petapixel.com/2012/11/25/facebook-removes-risque-phot...
      - fnordfnordfnord 14y ago
        That sort of thing might also flag a benign "Baby's first bath" video as child porn.
    - runamok 14y ago
      They already have image algorithms to detect nudity with a good degree of accuracy so that's probably enough to forbid an image or video.
    - octo_t 14y ago
      We do: Automatic detection of child pornography using color visual words - http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=6011977&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D6011977 http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=60... Automatic detection of child pornography - http://ro.ecu.edu.au/adf/60/ http://ro.ecu.edu.au/adf/60/
      - pyre 14y ago
        I wouldn't call it a dectector: our approach can achieve a prioritization of content (equal error 11 – 24%) to improve the efficiency of forensic investigations of child sexual abuse at least not in the context of this discussion of fully automated solutions.
    - deepblueocean 14y ago
      Not only do these exist, there are significant institutional practices that exist to facilitate the sharing of detection signatures between companies and between companies and various governments. It's good to prevent the re-sharing of child pornography, but a lot of companies don't like to talk about the systems because they're very much the same systems that the same companies are busy telling the content industry that they can't build to enforce copyrights. So they're generally kept pretty hush-hush by a kind of gentleman's agreement. I think general copyright is actually a thornier problem and this secrecy is unnecessary. CP is just something you can block if you see it. Tracking down exactly which uses of a copyrighted work are intended to be allowed and which are not is really much harder. And I think that's a legitimate argument, especially because it recasts the copyright debate in terms of compromises that make sense: if clearing a bunch of rights is confusing, maybe certain types of compulsory licenses are more attractive. But that debate is for another thread...
      - pyre 14y ago
        | I think general copyright is actually a thornier | problem and this secrecy is unnecessary 1. When we talk about 'detecting copyright infringement' there is an implicit assumption that we're talking about the IP of major corporations. No one is going to bend over backwards to prevent copyright infringement of 'the small guy'[1]. 2. Even if you detect that a video contains copyrighted material, there is no central registry for you to determine if the person (or entity) has permission to distribute/use that material. Even if there was such a registry, how do you link it to an online id on some random website? 3. There is no middle-ground for fair use. Fair Use is something that is usually only definitively determined after a court battle. If restrictions are too strict, then the only way for one to exercise their Fair Use rights is a court battle. This effectively means that Fair Use would be dead for all but some rich, motivated people (at least on large content networks, e.g. YouTube). [1] Unless it wins brownie points with the public for getting a system in place to protect the 'big guys,' who are obviously the only ones that matter.
  - artursapek 14y ago
    Child porn detection bot: add that to the list of projects I never want to work on.
    - jlgreco 14y ago
      I have to imagine they contract that work out. Unless doing that was my business, I wouldn't want to touch that training set with a thousand foot pole. Well I wouldn't regardless..., but you see what I'm saying about encapsulating and isolating that.
      - rmc 14y ago
        I've always wondered about that. Surely possesing the 'training data' (basically child porn) is illegal. Only police can posses it for investigations etc. So how can you make something that can recongnise it?
        steveklabnik 14y ago
        It's humans. There was a story about the 'dark side of Google' within the last few months about how Google contracts this out, and they burn out, but I'm not searching for that story directly myself.
        Spooky23 14y ago
        There are companies who have special licenses, etc to do that type of work. I was supposed to meet with a Scandinavian vendor a few years ago that provides a service where you can match SHA hashes of files against known illegal content. I recall the pitch deck said that they had millions of records to match against, which I found incredibly depressing. I ended up bailing on the meeting... it's a subject I don't want to learn more about.
- pbhjpbhj 14y ago
  >It might just be really bad practice and an under-/non-optimized crawler. // I know they've got lots of money but really, are they wasting bandwidth and memory so badly as to download terrabytes of video they don't need?
uncoder0 14y ago
That is certainly one possible explanation. For fun if you share a Google Drive video you can watch the FB user popup on the active user list and then disappear after a certain amount of time that varies based on video length and quality.

Isn't that typically done to show thumbnail previews under the pasted links?

by 0x0 14y ago

Isn't that typically done to show thumbnail previews under the pasted links?

Facebook generally will present the OpenGraph data from pasted links; for sites lacking it they might present the first image on the page. It would be vast overkill to download a whole video just to present a thumbnail, and the user experience in the delay while the whole file is downloaded and processed would render the feature fairly useless. Which leaves the question of why Facebook would do this open... It might just be really bad practice and an under-/non-optimized crawler.

Cthulhu_ 14y ago
Or they download the whole video to pull it through an analyzer to look for stuff prohibited to be republished via FB, such as child pornography. If I were FB, I'd screen everything published via my site.
- pyre 14y ago
  Right, because we have automated software child pornography recognizers.
  - ceejayoz 14y ago
    It wouldn't be surprising. YouTube looks for video fingerprints of copyrighted data. Repurposing such technology to find copies of known CP videos would be pretty straightforward. Wouldn't help you with new material, of course.
    - pyre 14y ago
      If you start filtering for child pornography, and people know about it, they will come at your hard if you miss anything. Even for a technology that gets all known child porn videos, people would be beating down the door to sue/bad mouth Facebook for missing the new ones. Mostly due to people being people, but also because the technology is 'magical' to far too many people in society so they feel like there are no limitations (or it's hard for them to grasp them).
  - runamok 14y ago
    They already have image algorithms to detect nudity with a good degree of accuracy so that's probably enough to forbid an image or video.
    - pyre 14y ago
      Like that won't cause a shitstorm. A social networking site that won't let people trade possibly nude images privately?
      - kamjam 14y ago
        I don't think he's talking about necessarily private, but they detect nudity on the "publicly" available images available in your albums which your friends an see... http://petapixel.com/2012/11/25/facebook-removes-risque-photograph-of-woman-showing-an-elbow/ http://petapixel.com/2012/11/25/facebook-removes-risque-phot...
    - fnordfnordfnord 14y ago
      That sort of thing might also flag a benign "Baby's first bath" video as child porn.
  - runamok 14y ago
    They already have image algorithms to detect nudity with a good degree of accuracy so that's probably enough to forbid an image or video.
  - octo_t 14y ago
    We do: Automatic detection of child pornography using color visual words - http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=6011977&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D6011977 http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=60... Automatic detection of child pornography - http://ro.ecu.edu.au/adf/60/ http://ro.ecu.edu.au/adf/60/
    - pyre 14y ago
      I wouldn't call it a dectector: our approach can achieve a prioritization of content (equal error 11 – 24%) to improve the efficiency of forensic investigations of child sexual abuse at least not in the context of this discussion of fully automated solutions.
  - deepblueocean 14y ago
    Not only do these exist, there are significant institutional practices that exist to facilitate the sharing of detection signatures between companies and between companies and various governments. It's good to prevent the re-sharing of child pornography, but a lot of companies don't like to talk about the systems because they're very much the same systems that the same companies are busy telling the content industry that they can't build to enforce copyrights. So they're generally kept pretty hush-hush by a kind of gentleman's agreement. I think general copyright is actually a thornier problem and this secrecy is unnecessary. CP is just something you can block if you see it. Tracking down exactly which uses of a copyrighted work are intended to be allowed and which are not is really much harder. And I think that's a legitimate argument, especially because it recasts the copyright debate in terms of compromises that make sense: if clearing a bunch of rights is confusing, maybe certain types of compulsory licenses are more attractive. But that debate is for another thread...
    - pyre 14y ago
      | I think general copyright is actually a thornier | problem and this secrecy is unnecessary 1. When we talk about 'detecting copyright infringement' there is an implicit assumption that we're talking about the IP of major corporations. No one is going to bend over backwards to prevent copyright infringement of 'the small guy'[1]. 2. Even if you detect that a video contains copyrighted material, there is no central registry for you to determine if the person (or entity) has permission to distribute/use that material. Even if there was such a registry, how do you link it to an online id on some random website? 3. There is no middle-ground for fair use. Fair Use is something that is usually only definitively determined after a court battle. If restrictions are too strict, then the only way for one to exercise their Fair Use rights is a court battle. This effectively means that Fair Use would be dead for all but some rich, motivated people (at least on large content networks, e.g. YouTube). [1] Unless it wins brownie points with the public for getting a system in place to protect the 'big guys,' who are obviously the only ones that matter.
- artursapek 14y ago
  Child porn detection bot: add that to the list of projects I never want to work on.
  - jlgreco 14y ago
    I have to imagine they contract that work out. Unless doing that was my business, I wouldn't want to touch that training set with a thousand foot pole. Well I wouldn't regardless..., but you see what I'm saying about encapsulating and isolating that.
    - rmc 14y ago
      I've always wondered about that. Surely possesing the 'training data' (basically child porn) is illegal. Only police can posses it for investigations etc. So how can you make something that can recongnise it?
      - steveklabnik 14y ago
        It's humans. There was a story about the 'dark side of Google' within the last few months about how Google contracts this out, and they burn out, but I'm not searching for that story directly myself.
      - Spooky23 14y ago
        There are companies who have special licenses, etc to do that type of work. I was supposed to meet with a Scandinavian vendor a few years ago that provides a service where you can match SHA hashes of files against known illegal content. I recall the pitch deck said that they had millions of records to match against, which I found incredibly depressing. I ended up bailing on the meeting... it's a subject I don't want to learn more about.
pbhjpbhj 14y ago
>It might just be really bad practice and an under-/non-optimized crawler. // I know they've got lots of money but really, are they wasting bandwidth and memory so badly as to download terrabytes of video they don't need?

uncoder0 14y ago

That is certainly one possible explanation. For fun if you share a Google Drive video you can watch the FB user popup on the active user list and then disappear after a certain amount of time that varies based on video length and quality.