Lately I read this two piece story on Wait But Why which I really would recommend to anyone wanting to get a better overview of the topic:
- Part 1: http://waitbutwhy.com/2015/01/artificial-intelligence-revolu...
- Part 2: http://waitbutwhy.com/2015/01/artificial-intelligence-revolu...
He then outlines the mechanism for all of these progress-(non?-)deaths[1] to occur, which he calls superintelligence. I find that explanation as either the cause or solution to all these DPUs very unsatisfying compared to the historically supported causes of death on that scale: disease, natural disaster, famine, war. I may be the "stubborn old man" called out in the article, but I don't on the face of it believe superintelligence will eliminate most deaths in all four of those categories at the same time. It's positing a mechanism seemingly immune to the selection pressure that got us here.
But even if I put aside my doubts about superintelligence, I would find it significantly more helpful to see a hypothesis as to why increasing computational power is a general mechanism whereby there will be fewer deaths from disease, natural disasters, famine, and war. I suspect it is more fruitful to focus on how computational power will help solve problems with critical infrastructure (shelter, supply, safety, communication, transport, resource control, &c) rather that puzzling over how it may create a new cause of death exceeding the magnitude of disease.
1: The author seems to switch signs here and imply these will be non-deaths, which is supported by actual population growth, so let's proceed with that assumption.
The interpretation I believe you've developed is connected to the number of people who die prematurely at a given point in history, and that the author's point then is something about how superintelligence will impact the number of people dying of disease, famine, etc, either raising or lowering that number. I can't find any spot in the article where the author raises the concept of the rate of people dying prematurely at a given point in history, and definitely not any place where he connects that concept to DPUs.
My understanding of the DPU is as the amount of change required in daily life for a single time travelling individual to "die of shock" upon experiencing another moment of time. In the examples provided, 100,000 BC to 12,000 BC was enough change in day-to-day life experience to cause a person from 100,000 BC to "die of shock" if they were transported to 12,000 BC instantaneously. The same assertion was made for 12,000 BC to 1750 AD, and 1750 AD to 2015 AD, with the author's conclusion being that there has been an exponential shortening in the timescale required for enough change to occur in the daily experience of human life to cause a time traveler to "die of shock", and that such shortening will continue into our future -- possibly to the point of allowing such a level of change to occur multiple times within our own lifespans.
I took the entire discussion of DPUs solely as an exercise in generating an evocative image for illustrating the increasing rapidity of change in our qualitative experience of life. I don't think the "die of shock" idea is meant to be taken literally; it's just a convenient stand-in for "extremely shocking, to the point that the experiencer may be incapable of processing the instantaneous change rationally", not a measure of people actually dying.
(Sorry for going to such length, I just wanted to be precise. This is a perfect illustration of https://xkcd.com/386/)