The Downfall Meme: YouTube and Content ID

If you’re a YouTube user, chances are, at one point or another, you’ll have seen a notice that goes something like this:

(via Public Knowledge)

I’ve seen it enough times myself – searching for clips ad nauseum, digging up content for the blog, following long-dead links from friends – yet I’d never really paid attention to the mechanics behind it. How does YouTube, with its infinite number of daily uploads, regulate such a flood of data? How does it balance the needs of its audience against the interests of bigger film and music corporations? And, as part of copyright ‘rebel’ Google, how does it negotiate the difficult task of shaping the digital landscape of the future, preferably according to its own interests? What follows, I hope, will be a basic foray into each of these questions, even if the answers remain somewhat grey.

Since its inception in 2005, YouTube has experienced its fair share of legal action, with Viacom’s $1 billion claim being the latest example. Similar pressure over the years led to the development of the Content ID system, which I think began life around 2007. Basically, it works as an automated filter. Companies must sign up and choose from a variety of usage policies. This includes for example ‘monetising’, where copyright holders allow videos to be published in return for additional exposure, advertising on the same page or links to points of sale such as iTunes. On the other end of the scale, they can simply ban matched content altogether.

Margaret Gould Stewart, YouTube’s User Experience Team Manager (fancy that for a title), summed it up pretty well with her TED talk:

We compare each upload against all of the reference files in our database … The system compares every moment of one to the other to see if there’s a match. Now this means that we can identify a match even if the copy used is just a portion of the original file, plays it in slow motion and has degraded audio and video quality. And we do this every time that a video us uploaded to YouTube. And that’s over 20 hours of video every minute. When we find a match, we apply the policy that the rights owner has set down.

… And the scale and speed of this system is truly breathtaking. We’re not just talking about a few videos. We’re talking about over 100 years of video everyday… [compared] against millions of reference files in our database. It would be like 36 000 people staring at 36 000 monitors each and every day, without so much as a coffee break.

(For a look at how these notifications work from a user’s perspective, have a look at this post on Off on a Tangent).

So is this system a progressive coup the copyleft or just another flawed and arbitrary beast? Well, probably a bit of both, although YouTube clearly hope it’s the former. And granted, it’s been saddled with the extremely difficult task of balancing the interests of multiple stakeholders. Content ID is the end result of attempting to strike a delicate balance between everyday users and big media firms, as safeguard against becoming the Napster of the decade. BPM made this observation:

I realized that although YouTube tells everyone to get permission from copyright holders before uploading material, they have a system in place to deal with it after the fact. This, in my mind, quite as step forward in the world of copyright. YouTube must follow the law, but it has a created a system which gives incentives to rights holders to allow copyrighted material to remain in place even if permission wasn’t granted in advance.

… YouTube has actually created a system, though flawed, which is working and pushing the envelope without going so far as to get shut down.

However, as with any automated system (the key word here being ‘automated’), there are glitches. A large part of YouTube relies on DIY culture – cover songs, fan tributes, mash-ups, parodies, amateur movies and borrowed files. Arguably, this is what has made YouTube the success it is today – the bread and butter that it trades on. One problem with Content ID is that sometimes even a snippet is enough to have videos removed and, as often as not, it’s the everyday users, fans and experimenters that suffer. The issue becomes all the more apparent when copyright holders choose to take the automatic blocking approach.

One example is the recent controversy over the Hitler Downfall parodies. These videos, which number in the hundreds, take the pivotal bunker scene in Der Untergang and replace the subtitles with fake ones. They have been used to mock everything from politicians to iPads and Twitter, as well as the Downfall Memes themselves.

Earlier this year, Constantin Films sought to have all of them removed from YouTube due to copyright infringement. Many popular Memes are still blocked, yet rather than stopping the parodies altogether, the move ensured that even more of them were viewed, blogged about and created, including this one, Hitler reacts to the Hitler parodies being removed from YouTube.

These repeated attempts and reactions probably reveal more about the subversive nature of the internet than they do the regulation of copyright. The instinct of online users has always been to do exactly what they are told not to – to hack, pirate and redistribute. Forcing users to adhere to old rules can often be inflammatory in itself – something well worth keeping in mind for future models. Typically, the moment the Content ID system got up, the web was rife with examples of how to cheat it. This via Boing Boing:

[Scott] Smitelli has poked around at the system, uploading copies of the copyrighted song ‘I Know What Boys Like,’ sonically altered in various ways … to roughly summarize: Altering parameters more than 5% often seems to fool the Identifier, and using less than 30 seconds also seems to let the clip slip through the rule-bound robot’s shiny little nets. Playing clips in reverse confused the Identifier, but stripping out everything except the vocals did not.

Finally, there’s also this site, aptly titled YouTome, which acts as a digital graveyard for all videos taken down due to copyright infringement.

This is a Spike cross-post.