Online video service Guba, fresh from distribution deals with several major movie studios, has launched a new technology that it claims will automatically detect and flag copyrighted video footage, even if that video has been altered. The company says it intends to license the technology, nicknamed “Johnny” (after Johnny Mnemonic), to other video sharing sites. “Johnny” was developed in partnership with the Motion Picture Association of America (MPAA).
This could really change the online video landscape, if it works. Goodbye Animal Planet clips on YouTube, for one thing.
Guba says that more than one million TV shows and movies are already included in its filter. That filter was an essential part of the deals the company has made recently with movie studios. The technology itself is a proprietary implementation of artificial intelligence four years in the making. It could be applied in any context including crawling the web, Guba co-founder Tom McInerney told me. Johnny finds and flags copyrighted video content, the business rules used to respond to that content can be determined by whoever licences it.
Guba says they are already talking with other video sharing sites about Johnny. McInerney says the technology also makes Guba much more viable as acquisition bait, especially since the 8 year old company hasn’t taken any outside funding it would need to multiply.
If other online video services have an effective, automatic way to monitor copyright made available to them – will they choose to use it? Or are many of these sites in reality made possible by the availability of copyrighted materials? Guba’s McInerney told me that in a perfect world Guba would be full of South Park clips, but given the lawsuits he’s sure are coming it only made sense to dedicate resources to creating the technology behind Johnny.
It may be time to start working on making better home movies and video blogs. If Guba works then a lot of theoretical questions about copyright could be faced with a technological answer.
Neil Kjeldsen helped with this story.
It will be interesting to see how they actually do it…Is it the same as Riya technology?
I’m starting a pool on how long it will take for a hack to emerge that renders Johnny irrelevant. I’m saying 3 months, tops.
I’m sure the MPAA execs were all creaming their jeans as they were reeled in hook, line and sinker by the Guba spin doctors, who undoubtedly pumped this amazing technological solution to the ‘copyright problem’.
Mat
“If Guba works then a lot of theoretical questions about copyright could be faced with a technological answer”
OK I’ll bite: name five “theoretical questions about copyright” (as opposed, I assume, to practical questions about copyright) that could faced with an answer if Guba works.
1.why do people like YouTube?
2. is the arguement “we’re just a delivery mechanism…don’t hold us responsible” just a convenient excuse?
3. are complaints about copyright enforcement just coming from a tiny niche of geeks, or if widespread removal of copyrighted media online occurred would a lot of people actually freak out?
there’s 3 questions off the top of my head that we’d find more out about if this thing moves forward.
“If other online video services have an effective, automatic way to monitor copyright made available to them – will they choose to use it?”
I think the real customer for this software would be copyright holders, not video hosts. (Assuming it actually works, as Mat pointed out). Guba could host or license the software to TimeWarner, Disney, etc. as an easy way for their lawyers to pursue copyright violators.
Obviously a huge portion of YouTube’s videos are copyright violations and YouTube currently turns the other cheeck becuase they don’t want to loose the traffic they generate. However, if copyright holders started actively harassing YouTube about the content (which would be possible via this software) then YouTube would have an incentive to purchase the software themselves and actively remove videos owned by the more harassing companies.
Jack’s mention of Riya is also interesting. Could one easily adapt Riya’s software to find the faces of celebrities in videos? For example, NBC could find violating videos by searching for the face of Conan O’Brien in videos? Maybe this is a direction Riya will take (or maybe they already have, I don’t follow them in detail).
There, I just gave two companies a million dollars worth of advice. I will be expecting my checks in the mail soon.
Personally I am surprised the technology hasn’t surfaced yet. It’s absolutly absurd that media copywrite is an issue. If its out there its out there and the public owns it. Hell one day the only people that won’t have to worry about it are the Chinese – lucky bastards… Do the get youtube in china?
http://www.cafe...om/urbangrenade
All this technology is pretty useless.
Fancy screening based on some expert system or AI ?? No matter how powerful, it will not work 100%. The most it do is to flag to administrator the potential of a video that violate copyright.
It just like anti-spam, it reduce the amount of screening by admin, but still up to them for deletion. If YOUtube used this system to auto-delete video file, youtube will soon be renamed as yousuke
or even worse…
http://www.yousuck.com for worse video censorship.
Let’s face it. The success of any video site is due to copyright material. Why do you think most of the time, ppl want to search and explore the video site to see other’s dickies ?? I believe most want to see material that bring nostagic memory of the past. And that means copyright material from the past. Only a few user-made video is worth seeing at all.
It’s useless. They scan a file and match it to a known database of copyrighted data. The comparison may use AI, but if the file is encrypyted, no-one can match it. The technology to bypass such methods is just and available.
The only similarity it may hold with Riya is in pattern matching images. Riya is not facial recognition per se, it just knows that one thing is the same (or similar) to another thing.
It sounds like Johnny is doing something similar. It has a bank of patterns (being some sort of representation of what copyright material is) and then it compares each video (probably frame by frame, or random frames) to the bank of ‘known copyrights’.
A few thing that I would like to know:
* How fast is it? Matching every video either being uploaded or stored on the web to a known database of 1,000,000+ doesn’t sound scalable. Then again generic matching probably would never work for copyright (while it may for porn detection)
* Can it be adopted to find porn? If it can, and if it is fast enough to be wrapped in a web service, then Guba is going to be a huge company
There is a lot of research going into this sector, there were bound to be breakthroughs sooner or later, but the source of this breakthrough for me is a bit surprising.
“Riya is not facial recognition per se, it just knows that one thing is the same (or similar) to another thing.”
their alogrithms are specifically designed for recognizing faces and text.
and even with porn this will become difficult.
just consider the movies:
bad boys II (many women with very little clothing dancing, looks similar to many women withut clothing dancing)
the human stain by phillip roth (intellgient conversation while kind of having sex might look very similar to dumb porn style conversation while really having sex)
swimming pool by the director francois ozon (french art / independent thriller with quiet clear images would be considered even by most american conservatives as porn, but most europeans will consider it for what it is)
(obviously i know, that these three examples should not be on a video sharing site at all, but they might serve to underline my argument.)
while copy right inflicting is quiet clearly defined (warner owns a movie and if you publish it on youtube this is copyright inflicting) porn is not. there is many images and videos we would all agree to be porn. on the other hand there is many movies and images in a gray in-between area. if you ever lived in a country, where porn is censored out of movies, you’d know what i mean. a number of central scenes of “american beauty” has been cut out in egypt for example.
so what is your machine going to do. in american beauty you can see a young woman / girl with bare breasts. is that already porn or is it only a scene of a well done hollywood movie?
considering all these problems and short comings i am not sure in which way they should be able to capitalize on a porn detector. if the pron detector is supposed to find porn (for people looking for porn), being imperfect might be no problem. the other way around, it gets more dificult.
moritz
can’t work won’t work.
too huge to examine, to prone to fixing e.g. rotation to get around detection. MPAA should stop wasting time looking for this holy grail, and adapt business models to suit the youtube generation. etc. I don’t like to be told, when I go to the cinema, not to pirate the movie – I mean, I just paid for the ticket, I don’t need someone telling me that. It makes me angry… Although I am sure they will make money their product won’t work really.
same shit with online music and the RIAA. Their industry is adjusting, slowly.
I’d be more interested in the technology, but I do sense that it’s likely more hype than substance at this point in time. Having a bit of a background in image processing, I know that image recognition and classification is a very, very hard problem to solve. There are lots of Computer Science people at MIT, Standford, and CalTech that do this stuff with very, very iffy results.
Seriously, people are giving YouTube the golden Internet award waaay too soon.
Lest we forget that pratically no one had even heard of YouTube prior to ‘Lazy Sunday’.
Guba is cutting its problems off at the pass. Filtering illegal content will save them millions in legal fees, which will, if nothing else, make them more profitable (percentage wise) than YouTube.
If we’re looking at this objectively:
-YouTube is ugly, illegal, yet very fun to peel through
-Metacafe is pretty, slow, illegal, full of filth
-Guba is by far the prettiest, legal, but not known… at all.
Give Guba a couple of years… they’ll be on top. The Internet has a short memory.
Whichever television network or movie studios use something like “Johnny” to prevent clips of their property from showing up on YouTube will have to contend with the upcoming streaming bittorrent sites which will be about as easy to shut down as ThePirateBay has been.
The value of sites like YouTube and the others is the wealth of information concerning their audiences. YouTube needs to tap into that and make it available to advertisers. Other problems will work themselves out.
I believe this tech is totally possible, it is just going to take someone figuring out how to index attributes of a video the way words are indexed from pages for search engines. Then comparisons are easy. There are a lot of different things that can be indexed that would make a video unique: color patterns, scene changes, motion, perhaps some words in frames that can be pulled out at certain times of the movie. I bet just scene changes alone is good enough so that if you say there are scene changes at a,b,c,d,e,f,g,etc times in the movie then that is a pretty damn good unique identifier for that movie. This is just one way to do it, I am sure there are dozens of others.
Wow, this will be an amazing technology if it can work as well as they say it can. Will be interesting to see the effective it could have on Youtube.
Why on Earth would you report on this? You are going to ruin a good thing by giving it any publicity at all! I like that I can go on Youtube and see copyrighted material. I appreciate the technology but use it for something else like SPAM or something.
Hi,
i think it’s not that difficult to artificialy detect videos, and it would be hard to bypass this detection method. Every Mpeg codec reduces the file size by detecting the motion of the objects in the movie, i think you can use that to indetify the content very accurate.
hey eph, that is a cool site. did you hear that some kids got thrown out of school for wearing their shirt in philidelphia?
http://www.cafe...om/urbangrenade
Not only these type of technologies are doable, they have already track record with companies such as PicScout [ http://www.picscout.com ] which is being used by many content owners to monitor their media.
If this “Johnny” technology is actually effective (which I doubt), then youtube will ultimately be cutting off its nose to spite it’s face. It will be a copyright violation free zone but nobody will be interested in watching it. Now that’s a shrewd strategy if I’ve ever seen one. It’s like the old joke where someone is told: “The operation was a success…but the patient died.”
Some new website will just emerge to take its place. Good.
The RIAA and MPAA STILL don’t get it. All those copyrighted videos and copyrighted songs on youtube function as FREE marketing for record and movie companies. I have bought numerous CD’s because I saw a video or heard a song that I liked in the internet. And the marketing value multiplies when you consider the cross-posting on myspace pages. Not to mention folks who will then buy concert tickets because they were exposed to a performer.
The corporate monkeys are stuck in mindset where they have to control and own every little thing. IT CAN”T BE DONE ON THE INTERNET. Instead of attacking some kid who added a song to their home video or some kid who wanted to add a “captured” video clip to their myspace page, they need to figure out how to work WITH the internet — use it to their advantage. Regretably, this new youtube deal is just the same old RIAA & MPAA approach with a new coat of paint. Who do they think that they are fooling?
The record and movie industry’s biggest problem is that they need to seriously improve their content. Album sales aren’t down IMHO because of copyright violations. Sales are down because most of the artists on major labels suck. And most of the movies being made suck too.
It’s the problem of a cookie cutter mentality. The entertainment industry seems nearly incapable of any real creativity. So they just keep doing the same thing over and over again and blame the internet when they don’t make as much money as they expected.
So good luck to the RIAA and the MPAA and their clients who are trying to hang onto the old business model with minor variations around the margins. By all means, keep beating that dead horse. It will be entertaining to watch.
I second James completely.
Its like there are bad apples packed in an opaque bag.
The seller says pay and then do whatever you want to.
Technology has enabled you to taste apples before buying.
If the apples are really good, the consumer will surely go for them.
And anyways, online advertising allows even the bad apples to earn some money.
Check out my cool clips