Autonomy has criticized Silicon Valley's Web 2.0 obsession with tagging to classify data, while launching its own software to help media companies scan web sites for pirated content. The enterprise search specialist has launched its Automatic Copyright Infringement Detection (ACID) software, which it claimed is capable of …
I look forward to the day when posting a photo of Times Square that contains a Coke advertisement (or any other copyrighted image) will result in an automatic DMCA takedown notice.
Additionally, if the software really *does* recognize any subset of protected material in an audio stream regardless of edits (which would be a fantastic feat of engineering given that this is quite difficult for trained humans), I can't wait for it to single out every rap album ever created for eggregious infringement of biblical proportions.
Please note that this isn't a dig on rap (except rap produced by Timbaland, who's well known for ripping entire songs from indie musicians and taking full personal credit, because he can): I create music using samples myself, and despite the fact that the samples I use are heavily obfuscated and melted into the mix such that they're basically reconstituted, I still run a risk just posting my work for free. And I'll face even more of a threat with software like this.
Rap artists, however, being commercially successful, are just a more likely, more juicy and ironic target for automatic copyright infringement detection - when you build an automated closet looker, better do a skeleton check in your own house first!
Oh, wonderful. So now the DRM cretins have hijacked the ACID acronym. When I saw the headline, I wondered how the MPAA had managed to compromise the ideals of atomicity, consistency, isolation and durabilty.
So they filter based on what?
So they look for noise. So if I play the music for my podcast and record it on my mike (stereo mike), they won't pick it up since the noise has changed. If they filter based on things like tempo, then MC Hammer is the most egregious violator, stealing not just the licks or the melody, but the actual recording. For the entire song. (Can't Touch This vs Super Freak).
ACID will, in the end, turn out to be a failure that becomes and industry standard. I'm surprised Microsoft hasn't submitted it.
dillon in Tejas
well puffy you must like lil john and puffy too :)
one is lazy and they other takes songs he hopes the younger crowd wont reconize as music from thier parents gen.
Taming the Wilderness
"The industry obsession with structure and tagging is fundamentally wrong," Lynch said. "It's like the idea we can tame the wilderness by building car parks over it."
No. That analogy only works from the viewpoint of a content producer who's frightened his content will be stolen (which does fit nicely in with their product, surprise surprise.)
From the viewpoint of content consumers, metadata (tags are just a specific form of metadata so I won't perpetuate the redundancy) is more like putting labels on boxes so you can find out what's in them.
The real problem with metadata as it stands is that it's generated by the content publishers, and so only reflects what they consider relevant about the content. A system which allows content consumers to produce metadata would allow for much richer description of the content. For example, a botanist might publish a picture of a particular plant, and tag it as such, whereas an entomologist might notice the insect walking up the stem, and want to tag that.
Mark my words - they have outsourced the "engine" to East Timor or Ruanda, where full sweatshops of kids have to listen to amateur tracks and watch (oh, the pain!) freak home videos from YouTube! in a quest for copyrighted content.
However, until This Ultimate Truth is brought to light by TheReg's finest, let us assume that Lynch's claim is correct and the technology in question runs actual software.
I wonder how usable such a software would be and to whom it will be useful. Imagine a copyright giant such as Viacom trying to unleash this ACID IDOL on the abovementioned YouTube. How large the envisioned database of copyrighted material should be in order to cover all tunes, videos and images in Viacom possession? The complexity of comparing a single video clip with all or most of protected videos in search for several seconds or even a single frame of infringement renders Autonomy's bold claims somewhat untrustworthy.
Unless they're using algorithms as complex as human mind, i.e. real people from Borogravia or Moldovastan. Which happens to prove my conspiracy theory.