I don't know about everyone else, but personally I hate that there are so many fraudulent videos being created and shared on the internet, with far too few willing or able to identify them. Sometimes they're benign, simply a "Look at this unbelievable thing that happened", which of course didn't happen but it is on video so it is believed. At worst, these are a time-waster, or hurt your pride when you realize you've been taken. Sometimes they're not so benign, such as videos edited in a manner to enrage people against an innocent man or group. Most such videos(at least the ones I've seen), will use fancy editing of real events(hiding context that explains perceived misconduct), or uses a combination of clips from different events to completely fabricate a scenario(for example, video of police misconduct from years past, and long forgotten by most, mixed with news clips from modern riots leaving viewers with the impression of new crimes by government). And it isn't just end-users on social media who are fooled by such clips, they regularly are spread by popular news and political websites as well.
With the massive volume of such fraudulent creations, we need a way to more easily track video origin. What would be nice is if there is a way to see links to the original uses of individual segments as you are watching them. This way, when you see that amazing or infuriating clip you can immediately see that it is bunk, or at least see that it has been highly edited from the original. It may seem that this is a huge problem to tackle(as how can we build such an index to search against), but I think there are two very good possibilities.
First, since most of these videos are either initially or eventually posted to Youtube, perhaps there is a way to utilize the anti-copyright-infringment tech to help trace parts of videos. I'm sure they're already breaking all videos into bits and creating a hash or other log in order to find infringing parts. Depending on how this is done, it may be trivial to use this to find the "first sight" of each clip, or perhaps a list of all uses of the clip.
In many ways, I can't believe I am suggesting this. When Google first started automatically monitoring videos for copyright infringement, I opposed it due to the risk of using that tech for censorship of fair-use content. Since then, there have been many instances of take-down requests resulting in censorship of fair-use, but I can't blame the monitoring tech for that as I believe most are the result of human-error(or human-corruption, depending on the case).
If we could use the anti-infringement tech for fraud-detection we can possibly improve the quality of amateur reporting by making frauds harder(and therefore the remaining video more trustworthy). This could also be used for searching. If you want to find video of a current event(say a recent riot), search by date excluding clips that first appeared prior to the date(not just by posting date of the individual video). Only a google/youtube engineer can tell use if such a thing is possible, but it would be wonderful if they were to do so.
Another possibility is to build that tech into the video decoder. While you are watching video, it would automatically be analysing and fingerprinting what you watch and when you watch it. These logs would be sent(anonymously), to a public database which merely tracks the first few appearances of that video. Then, when that video gets remixed into the latest fraud, a user who wants to check whether it is true would simply press a button, and their browser/app would query the public database and present the list of earliest known uses of each segment within it.
I'm sure there are a number of difficulties to overcome, but I don't think any of them are too difficult for this to be a reasonable goal. Obviously, there would have to be coordination between several parties for the decoder-based version to work(utilizing the Youtube anti-infringement tech would be more straightforward, but dependent on the dedication of a single company and assuming it could be re-purposed for this use).
