Slashdot Mirror
Schmidt Says YouTube 'Very Close' to Filtering System
cnetfeed writes "Google CEO says an automated system will soon be available to track pirated content and prevent it from being uploaded to video sharing site. The system was supposed to be rolled out as early as last October, and the long delay in brining the technology online has resulted in ill will from companies like NBC and Viacom. 'Network executives accused Google of stalling so YouTube could reap the big traffic that professionally-created shows generate. Viacom filed a $1 billion lawsuit against Google last month and accused Google of massive intentional copyright infringement. "Ah Viacom," [CEO Eric Schmidt] Schmidt said. "You're either doing business with them or being sued by them...we chose the former, but ended up the latter." Schmidt took the opportunity to poke fun at Microsoft's assertion that Google's pending acquisition of DoubleClick may be a threat to fair competition. Other companies, including Yahoo and AT&T have also asked regulators to review the transaction closely.'"
Maybe Google can turn Doubleclick into a better company--who knows? I am not drawing any conclusions yet, I want to see what they do with it.
This is NOT a hard AI problem. This is a problem of measuring various features and statistics of videos that are independent of video format and frame rate, and using those measurements in a hash function. Once you have a good function, the biggest problem is making it independent of video start and end time. The solution is probably done by hashing small time-windows. With small enough time-window segments, one could look for videos with sequential time-window hashes that matched stored copyright material.