This is cross-posted to my Speculative Nonfiction blog.
It's been estimated that humans experience about 20x10^12 bytes of data per day (20 terabytes). Our episodic memory - our long-term storage - is at most 10 terabytes. You experience in a single day twice what you can possibly remember in your entire life. Multiplied by 7 x 10^9 humans, that's 1.4 x 10^23 bytes per day, almost all of it being forgotten immediately. The vast majority of that is sense data, especially what we're seeing; almost all of that is forgotten less than 2 seconds after being experienced.
Compare that 1.4 x 10^23 of data humans are experiencing per day with IBM's estimate that the world is storing 2.5 x 10^18 bytes per day. Assuming that's near capacity, if we're performing total capture - storing everything that people are seeing and hearing and thinking and feeling - we're still only able to store the experiences of 125,000 people. That's Topeka, Kansas. (I'm ignoring information theoretic considerations here - just what we have machines to store and retrieve from right now.) This is likely to increase, but for now we're not capable of anything like total capture. Experience is indeed ephemeral.
There's a bit of panic in realizing that to a first approximation, everything you experience disappears immediately forever. I can't help but wonder to what degree that's part of the motivation of the total capture crowd. Those folks still aren't getting the full 20 Tb of their days, but even so , the question for them is, what do they intend to do with this information? That is, how will they behave differently because of it? More important, how can they sort it for anything actionable. Our working memory is our bandwidth which limits our ability to make decisions and process information, and unless you think you're going to sit in quiet contemplation for a thousand years (after you retire presumably and are not generating any new data), then it's hard to see why total capture can be a worthy goal. Increasing working memory and improving search and decision algorithms seems obviously a better strategy. Interestingly this comports well with studies showing that working memory is an even better predictor of life outcomes than IQ or episodic memory - it's thinking critically on your feet and controlling your attention, not being able to remember the date of the Battle of Waterloo, that determines your decision-making ability and consequently, your happiness.