Slashdot Mirror
YouTube's Content Identification Failure Raises Eyebrows
MSNBC is carrying a story looking at YouTube's failure to follow through with a promised 'content identification system' by the end of the year. The article goes on to discuss the possible impact this failure will have on the site's (so far) good relations with television, music, and movie studios. From the article: "If the delay lasts for more than a week or two into the new year, suggesting more than just a slight technical hitch, 'this is certainly going to be a serious issue', [Mike McGuire, a digital media analyst at Gartner] added. Leading music companies have already made clear they see completion of YouTube's anti-piracy technology as an important step in any closer co-operation. Failure to build adequate systems to protect copyright owners could also add to the risk of legal action against the site."
_WE_ don't know if they did or not. This kind of negotiations are usually behind closed doors, and on this level this means vault doors.
Let's wait for some time and we will know. Any lawsuit - they haven't. Simple.
"an experienced, industrious, ambitious, and often, quite often, picturesque liar" - Mark Twain
Because once they show that they can identify bad content within video files won't the MPAA/RIAA/* start to bug them about soing the same with normal search results?
:(
:(
Instead of Perfect 10 having to search and list the illegal boobies on display, google will have to automatically remove them from view
Won't somebody think of the boobies
liqbase
This may sound a little OT - sorry for that - but this story raised an old question here: is it really possible to do an automated content identifier/filter solution? Personally I've always found these kind of solutions full of flaws. Take web surfing filtering for an instance: it's pretty common that the filtering software makes a mistake and end up identifying a "false positive bad content site". After all - google or not - both things follow the same basic principles, right?
Er Galvão Abbott - IT Consultant and Developer
Isn't all they need to comply with DMCA a link to allow reporting of DMCA violation/copyrighted protected content and removing of the content once verified?
The solution would be to perform some sort of hash check against previously taken down material. So actually posting copyrighted material once and having it spotted, would stop it from recurring on the system. It just needs to still match submittions with bits cut out and varying watermarks and source qualities with some kind of identification algorythm. (similar to fingerprinting)
I don't know -- IANAM (I Am Not A Mathematician), but it sounds like an exceedingly difficult problem. To fingerprint a video, you're going to have to use specific information from it, and I don't know what information will remain constant between different encoding qualities and even encoding applications using the same theoretical quality. I assume you have to fuzz it up (the mathematical equivalent of "this area of the image from time index X to time index X+2 is reddish-orange, and this other area during this other time range is pinky-russet"), and that will result in false positives. And, if the algorithm is publicly known, people will mess with the areas that are known to be used in the fingerprinting to cause false negatives. Or just mess slightly with the saturation or intensity of the entire video, if they don't know the precise locations but do know that this is the kind of thing the algorithm checks. And with the number of people hammering at it, I don't expect the rough workings of the algorithm to remain entirely secret for long. And then it's a footrace like with Google's ranking algorithms, with lots of folks working to figure out how to beat it, Google improving it, and hackers having at it again.
In any case, I shall watch developments with much interest.
You seem to be assuming that the only "fingerprint" algorithm that exists is something like MD5.
You have something better? MD5 is the easiest computationally and produces the smallest result to store, using other techniques will increase the size of your database and computational expense. They could do FFT on single frame images, but you would need one for each scene of interest. The result could be made independent of size but not encoding quality. It would also be large and could create hundreds of fingerprints from each "protected" film. Trivial modifications, cropping and luck would circumvent the routine. Any way you look at the problem, what's required is intelligence well beyond the state of the art for machines and impractically expensive though human labor. It's not going to work.
A perfect result, as I argued, would unravel the whole game. It would prove that the industry "pirates" it's own content. That would lead to an impossible leagal quagmire that would show the folly of existing copyright law. If individuals could be sued years after the fact for fair use based on the results of a computer program, so could the big companies. If they did not pounce on each other, they would be in violation of anti-trust laws. It would be easier to skip all the heartache of DRM and recognize fair use in the first place. No one lost any money when Star Wars used themes, images and dialog from previous movies. Most of all, it would be a tremendous waste of intelligence. There are much better uses for a machine that could do the kind of matching the MPAA and RIAA dream of.
Finally, the MPAA and other copyright holders should bear the cost of policing their own works. YouTube is more of a common carrier than it is an entertainment company. Forcing Google to police it's users is kind of like holding the phone company or the post office responsible for pranks and crimes committed by their users.
The project is a pipe dream by companies that already enjoy tremendous compensation for the most minor of infractions. It's impossible, impractical, immoral and stupid, but they have gotten away with so much they might as well try this too. The sooner the broadcast monopoly companies die off, the better off all of us will be.
Friends don't help friends install M$ junk.