Not long ago, I read an interesting solicitation from DARPA (Defense Advanced Research Projects Agency) and they were looking for a mathematical formula which would allow an artificial intelligent search through hundreds of thousands of hours worth of military UAV video surveillance.
They're looking for the mathematical formula or algorithm which would allow a computer to search through the abundance of video footage based on queries that were asked.
After all, if you are looking for certain surveillance to prove that someone was somewhere at a certain time, or that a certain trip by a certain type of vehicle in a certain area or region was an anomaly and out of place, then you'd have to search through all the videos from that particular region going back years.
Nevertheless, going back in time to look at such video could very well lead you to the proper answer to your question.
And being able to think through time, or to pull up all the surveillance videos that had to do with a certain anomaly in a certain place would be of the utmost advantage - wouldn't it? Therefore, there needs to be a better way to code each and every video with certain data criteria to make it easy to search it.
Well, that is the old way of thinking, and it appears the DARPA is looking for something totally new, still, is the old-way a viable component to all of this? It could be, unfortunately the most important thing you can do is know which questions in the future that you are going to ask of the past so that you can properly code each and every single video by sequence.
Since we cannot know in the future which questions we will ask of the past, our only solution is to capture all of the data and then take all of the videos and swoosh them into one gigantic barrel, and you can search from there.
However I would submit to you that you need to ask yourself first what you are going to be looking for in the future, prior to capturing the data in the present, if you are in the future to know the true past.
The dilemma of course is the abundance of data that is required to capture everything, and at what point is it okay to dump the data when no anomalies exist, that you know of? And who's to say that you know what an anomaly is if you don't know which spectrum of light you are looking for the first place.
Don't for a minute think that a video is not capturing fragments and scraps of the non-visual light spectrum that humans cannot see - because it surely is.
Much of what we may not consider today as an anomaly is encoded in those old videos, but we don't know enough right now to know what we should be asking for, or what we should be collecting.
And even if that is not the case, if the non-visual light spectrum is nothing that are to care about in the future, this philosophical example does illustrate the challenges and problems that we face when trying to design an algorithm which will be able to search past video data for questions we do not know yet exist.
Now then, there is a lot of talk about search engines seeing the future based on what people are searching, or the wisdom of the crowd.
In other words, search engines are beginning to predict often into the future based on the will and intent of the populace moving forward, all by their questions and queries of past events stored within the abundance of human data on the Internet.
They are asking the questions in the present, and revealing the future based on their queries of the past.
And it seems to work, so I ask the question; what questions are we going to be asking in the future present, of our past data? It matters, and don't you dare tell me it doesn't, in fact it is the most important question in solving this problem.
Indeed I hope you will please consider all this and if you have any comments, concerns, or questions then you know you can send me an e-mail at any time.