4 ms·

I think I see a hole in this logic. A very well thought article and I agree with the main premise, but allow me to nitpick a crucial point... [Warning: specul

I think I see a hole in this logic. A very well thought article and I agree with the main premise, but allow me to nitpick a crucial point... [Warning: speculation, I do not work for Google...] Google has developed a very effective algorithm for detecting porn, not child porn. They can use flesh tone analysis for detecting nudity but algorithmic analysis of a single image to determine the age of the person in the picture seems way harder. File names don't work either: a picture of a handbag with a file name of "16yr_old_porn.jpg" is not illegal. Battling child pornography requires a human to look at stuff. Search results get flagged by normal users (and by paid teams I'd bet). These flags are analyzed and used as patterns to detect other similar images (similar colors, sizes, names, patterns in the image, source urls, watermarks) and this secondary analysis can be done with fancy algorithms but it still requires groups of people to flag things and create these patterns. If I'm right, then a similar method could be used to help combat infringing material. But as danwin points out, there is a new factor: to figure out if a particular piece of content is licensed or infringing, which is hard, even for a human. But there are some people who can look at a video, picture, (listen to a) song, etc, and be pretty sure that it is infringing. Clips or full videos of some popular movies on youtube are obviously infringing. Most mp3s on the web are infringing. There would be false positives and false negatives (just like with child porn), but you could build a similar type of pattern database that could be used to find similar files. This same type of flagging/reporting could be used to identify infringing sites. I know youtube has an existing algorithm for identifying patterns in audio tracks to some database, but I don't think these patterns are based on flags, I think they are hard-coded. Users could still browse directly to a URL (or IP) hosting questionable content but it would definitely cut down on the easy access and spread of files. NOTE: I don't like the idea of Google (or any organization) being mandated to implement such a process but it may be better than interfering with DNS mechanisms and lawsuits and court orders. I also don't like how reliant we are on a private organization holding such power over what the world can find in their results. Nearly everyone agrees on child porn but I think one of the main reasons SOPA is controversial is because we don't all agree on the importance of fighting infringement. (Counterfeit goods maybe, but not license violations) By no means do I claim that SOPA is a good bill. I think doing nothing is better than implementing what is in SOPA. EDIT: An assumption made here is that we consider the way Google is fighting child pornography is effective, sufficient, and ethical. I think that is a reasonable assumption. EDIT2: One factor I didn't consider is the quantity and diversity of child pornography versus infringing material. I'm not sure how the algorithm would be affected but it'd probably be slightly less effective.

danso 15y ago
I think the issue is that not only is this burdensome, but it practically invites the enforcer to err on the side of caution, leading to a chilling effect. For those who haven't taken journalism law classes, a newspaper can print falsehoods without fear of being sued, especially about public figures. The threshold is that they cannot do it with "actual malice" or "gross negligence."...this is not the case in many other countries, including Britain. Why not make it a suable offense to print any kind of falsehood, which are most definitely detectable by human (and probably machine) readers? Because if a newspaper -- or any media source -- had to worry about being sued because of simple mistakes of fact or, more problematically, because a whistleblower misled them -- then most newspaper businesses would err so far on the side of caution that they would not print anything useful about any public figure. [insert ridicule for how the corporate-controlled media doesn't do that anyway] To go back to the SOPA case...a company like Google has to deal with something much more difficult to ascertain than errors of fact, and so it worries that it will have to adopt a "it's better to apologize than to have to ask" policy, which would be detrimental to its search property.
- thebigshane 15y ago
  I agree with your premise if this algorithm was enforced by threat of lawsuit. An alternative would be that, instead of lobbying&lawsuits, the RIAA/MPAA pay Google to implement such a thing and in return also grant them immunity/amnesty. [insert ridicule for how Big Content would never do such a thing]
burgerbrain 15y ago
"Clips... ...of some popular movies on youtube are obviously infringing" Or you know, fair use.
- thebigshane 15y ago
  "Fair use" is pretty specific. Most interpretations allow to use a clip in your work. But if we are taking about a clip of a movie on its own, like the first 10 minutes, without any transformative addition/change, I don't see a "fair use" defense.

danso 15y ago

I think the issue is that not only is this burdensome, but it practically invites the enforcer to err on the side of caution, leading to a chilling effect. For those who haven't taken journalism law classes, a newspaper can print falsehoods without fear of being sued, especially about public figures. The threshold is that they cannot do it with "actual malice" or "gross negligence."...this is not the case in many other countries, including Britain. Why not make it a suable offense to print any kind of falsehood, which are most definitely detectable by human (and probably machine) readers? Because if a newspaper -- or any media source -- had to worry about being sued because of simple mistakes of fact or, more problematically, because a whistleblower misled them -- then most newspaper businesses would err so far on the side of caution that they would not print anything useful about any public figure. [insert ridicule for how the corporate-controlled media doesn't do that anyway] To go back to the SOPA case...a company like Google has to deal with something much more difficult to ascertain than errors of fact, and so it worries that it will have to adopt a "it's better to apologize than to have to ask" policy, which would be detrimental to its search property.

thebigshane 15y ago
I agree with your premise if this algorithm was enforced by threat of lawsuit. An alternative would be that, instead of lobbying&lawsuits, the RIAA/MPAA pay Google to implement such a thing and in return also grant them immunity/amnesty. [insert ridicule for how Big Content would never do such a thing]

burgerbrain 15y ago

"Clips... ...of some popular movies on youtube are obviously infringing" Or you know, fair use.

thebigshane 15y ago
"Fair use" is pretty specific. Most interpretations allow to use a clip in your work. But if we are taking about a clip of a movie on its own, like the first 10 minutes, without any transformative addition/change, I don't see a "fair use" defense.