[MUSIC] You're on vacation at the beach. [MUSIC] And you take a photo. Your point of the photo is to capture this beautiful bay, the water, the beach, the sand, and surf. Of course there are some people on the beach and they're in your photo, but they're incidentally in your photo. That wasn't what you were trying to capture, you were capturing the scenery. You come back home, you post this on the web as part of your travel log. This face recognition software, the people on the beach get auto-tagged with their names. There's search software that indexes photographs based on tags on the photographs. And so now we have the individuals who were in your photograph, indexed as being on a particular beach, at a particular time, and in that particular photograph. Now it turns out that in your photograph was this couple who were having an affair and their spouses didn't know. Once this photograph was on the web and indexed, the next time one of their spouses just Googled her husband there she finds this photo. So there's significant consequences to their marriage. And the question is who's responsible in this chain of events? So let's analyze the steps in this process. Surely, somebody taking a vacation photo at the beach is just fine. And surely, somebody posting a photo that they took on the beach is just fine. The next step is the tagging of names on the photo. Well, if these are names of people that the photographer didn't know, I think that there is some question. And one should wonder what the auto-tagging software is being used for and what value is being generated. Since there are dangers of misidentification or dangers of undesired identification, that the potential costs might outweigh the benefits of auto-recognition. Once these photos have been actually tagged, then it's a question of whether search engines should index photos based on the tags available to them. Here again, there are cases I suppose where one would be able to argue that there are things they shouldn't index. But as a matter of general principle, it's probably not easy for a search engine to make a value determination of good tag, bad tag. And so, just in terms of increasing their coverage and allowing their users to find more things, I would expect them to use all tags available. And then the final step of course, is somebody searching, which everybody does there isn't an issue there. So we have a complex chain of events, and the only step in this chain where it seems as if one needs to be careful is the step of auto-tagging images. Which if one were to consider in isolation doesn't seem like such a terrible thing to do. But when you think about it as a step in a chain of events, such as the one that we just described, that is potentially a significant issue that we have to deal with.