I will be very interested to see how this actually works. The technology must be very clever to automatically identify video that is subject to copyright conditions.

It's an advanced version of their in-the-works image recognition software that would look at the keyframes of the uploaded videos, extract and average common subjects along the way of user provided data, and match them up to the most popular, just aired TV shows' stars, hosts, sports teams and artists. If they are from a publisher that they don't have a deal with ( ie. the image recognition software identifies Jon Stewart on either of the keyframes ), the video is marked as copyrighted material.

There's also a secret signal within TV shows, ( just like VCR Plus+ or ShowView codes, or perhaps it uses these ) that when get encoded into video files ring an alarm at YouTube, with the exact time, title, and channel the show was aired on. If it's not a channel they have a deal with, the video gets filtered out. Also, this new system will let them provide much better targeted ads next to the videos.

Yes.

...

Either these or it's a text based filter.

-J0n St3wart

*snip*
Idea on how it could be done removed. They probably know anyway, but if not...

[edited by: Miamacs at 12:30 pm (utc) on April 17, 2007]

Sounds too complicated. Maybe youre giving them too much credit, and they're just banning videos with tags like: John + Stewart + Daily Show. At any rate, they can control the problem but not solve it.

Doesn't seem too complicated to me. Makes perfect sense. A strictly tag/text based filter would be completely useless and too easy to game.

Actually, I believe that the computer is reading the closed captioning of the videos and filtering on this.

I was joking.

...

But just by using the terminology of their other sci-fi press releases, laics may think it's actual reality ( not you of course ). This news was propaganda. A nod to the public, the investigators, lawyers, judges of present and upcoming cases of YouTube.

You can encode any file in a way to exclude information that they'd track. With that said, they could try to filter certain subjects without looking at a single frame of VIDEO data, ( let's not give them ideas, but it's not the captioning ). But even if they knew how to do it efficiently, it'd eat away an enormous amount of CPU power.

[edited by: Miamacs at 2:20 pm (utc) on April 17, 2007]

What about combining the user database (where you can actually recognize the users that continuously upload copyrighted material) + IPS + cookies? + some other variable your browser or OS has, to create a blacklist?

sounds clever.

Cookies are not the best thing, neither ips, neither the multiuser accounts some open, but combining a lot of this can actually help.

sci-fi press releases

Yep. I call b/s.
Sure, they could throw together a "filter" that looked for text alerts ("John Stewart") plus a banned list of users/IP's, plus maybe a user reporting system. Nothing special, just standard monitoring techniques.

They're not going to unveil the eighth wonder of the world.

very near enacting a filtering service

Picture a room full of terminals and human intelligence, along with a pamphlet of Do's and Dont's.

I was joking.

The way that software and cameras are able to recognize faces maybe it wouldn't be too difficult to analyze keyframes and compare to copyrighted programs. I'm sure Google will be going after the low hanging fruit but a high tech solution may be possible.

[edited by: MrSpeed at 4:55 pm (utc) on April 17, 2007]

No one said it wasn't possible in theory to develop an AI that would recognize faces, objects, subjects on a video. Actually it was me who said that they *could*, in theory. Some universities are on to this, although they work on still images, and neither has anything that works efficiently.

It's just that it's not Google who'll do it, not for YouTube, just to calm a couple of copyright claims. If they did it, they could SELL AdWords and Adsense, and wouldn't accounce it through the defensive commentary of Erik in between remarks on Viacom and Microsoft.

A research lab *somewhere*, founded by the US and the EU, the FBI and the CIA, -- with dozens of Japanese and German technicians aiding them -- is probably working on such a system right now though.

But Google is a search engine / advertising network.

Not NASA.

Though they do show some satellite imagery as if it was their own.

Lets not give the government too much credit either. This technology is most likely to come from where most technologies come from, competition in the market place. Sure, everyone can name a few government invented technologies, but the vast amount of innovation in the world comes from entrepreneurs working together in a freeish market.

Google almost does something news at 11

Sure, everyone can name a few government invented technologies

yeah, here are a couple. The internet, space flight, atomic weapons, nuclear power, radar, computers.
Your turn.

Btw pagerank, google's big breakthrough technology - I don't want to detract from it, because it really was genius, - but it is basically the application of neural network activation flow to the internet.
Now, what sector invented neural networks in the first place? Um...

Google probably refer to encoded info - this would (of course) not be applicable to the back catalogue, but as the vast majority of copyright thieves have the imagination of a hamster, that should net 98% of illegal vids by next tuesday pm.

If they are more ambitious, then it'll be soundtrack based.

No way is it image recognition; that's pretty unreliable for still images, let alone these new fangled 'moving pictures'; reditied, recolored and pan and scanned. They'd have to compare every frame with a frame library the size of Mars, on current technology.

But sound has a wave form. Nice, simple, linear job. Easy, and would take 0.00000000000000001% of the memory that image recognition would require.

I'm surprised no-one has yet remarked on the name "Claim Your Copyright". This sounds suspiciously to me as if it will be some copyright-owner-instigated action.

Because it's bound to be; Google won't pro-actively search for copyright theft; it'll always be down to the owners to file claim - then google will 'investigate'.

Hopefully, if you click on "Claim Your Copyright" once, google will search for similar clips ;)

But unless your middle name is Viacom, I'd not hold your breath :)

If they are more ambitious, then it'll be soundtrack based.

... there. You gave it away.
This is what I edited out from my first post, just in case Google hadn't thought of it yet.

Almost sure they did though.

[edited by: Miamacs at 12:21 pm (utc) on April 18, 2007]

Ha!

Your self censorship would have been too late; there's a previous thread where similar issues came up:

[webmasterworld.com...]

That one was called Google Gearing Up Antipiracy Tools for YouTube "very soon"

I guess we'll need to call the next one "Antipiracy Tools for YouTube Tuesday Week If It's Not Raining, Says Google" :)

I still don't know how youtube remains magically porn-free, when it is supposedly impossible to police.

atomic weapons were invented by the government, but I would not consider that an advancement of civilization.
Atomic weapons are anti-civilization, invented to kill people who are not naturally against each other.

The following were not invented by the government, and in some cases despite it:
Google,
webmaster world,
The Works of Jesus, Shakespeare, Socrates
Language,
Rubber Tires (Firestone), Flight (Wright),
Ipods
ebay
standardized parts
holly wood movies,
John Deere
Jeans
Furby Dolls

Have a limited government to defend your liberty, but I wouldn't look to government as a provider or inventor.

The following were not invented by the government

My point was that scientific breakthroughs are usually the result of government backed public sector research. Technology on the other hand is the domain of the market. If that violates your worldview, adjust the worldview.
There are exceptions to both, sure, but the list you gave mostly contains clever technological applications, which just reinforces the point. Webmasterworld, with all respect to Brett, is not equivalent to the moon landings. (And btw "language" is not an invention per se, any more than sex, friendship, and nostalgia are inventions.)
Getting back to the thread, the kind of technology required for identifying copyrighted segments of video would require something of a scientific breakthrough, and is more likely to emerge from a major MIT or CMU undertaking, than a room at the back of the 'plex. Google are great at technology. They don't do science.

"My point was that scientific breakthroughs are usually the result of government backed public sector research. "

I do not concur. A few scientific breakthroughs came from government, but most did not.

Vaccines, for example, are more important than landing on the moon in terms of helping people and they came from markets. The government has had many discoveries and they are important, but market competition has founded many more important discoveries than governments.

In fact, it is the very nature of the market through competition to discover new products. Markets are naturally geared toward making a better mousetrap, i.e invention and discovery.

So Google is more likely to create this technology, than some government scientist.

If they are more ambitious, then it'll be soundtrack based.

No way is it image recognition; that's pretty unreliable for still images, let alone these new fangled 'moving pictures'; reditied, recolored and pan and scanned. They'd have to compare every frame with a frame library the size of Mars, on current technology.

But sound has a wave form. Nice, simple, linear job. Easy, and would take 0.00000000000000001% of the memory that image recognition would require.

quadrille:

a couple of months ago it was suggested by techcrunch that google was licensing audiblemagic for the youtube copyright filter.
i haven't seen this mentioned in the latest spate of news, but i would agree that sonic analysis is the solution.

I don't know that I've ever seen anything on YouTube that was entertaining or interesting which was not a copyright violation.

Is Google shutting down YouTube? Or is it a Napster-Headed Ho?

p/g