Whoever says they have to use it just for medical care? They can have a separate secret one for “enemies of the regime” or “politically incorrect”
and a (ahem) special tag which feeds those names into the rest of the process (for the “real” no moar care flags) midstream.
Who’s gonna reverse engineer, or DNN the whole encapsulating process?
(Because, like you said, “science!”)
My assumption would be simulating markets and maximizing profits...more money in that than medicine, for the moment. Your idea of targeting of political enemies, I assume that is already in the works. But the work would be incredibly sensitive, garbage in garbage out. If you have biased input, the output will be biased.
As far as have a DNN look into itself...or another DNN for reverse engineering, I believe it would be about as complicated as unraveling the hidden meaning of DNA. We have small bits, that indicate certain problems. We cannot cut and paste, yet. We cannot replicate the sequence and put a simulation of the entity it came from, yet. There are protective codes we don’t have a clue about, even though the Nobel Prize went to someone who discounted the protective coding as junk DNA...he did not have to give the prize back.
In the new movie Ex Machina, and I have a mixed review of it, the Billionaire Wunderkind turned on all the devices he wanted to, to get the information to train his android human-like entities. We process information at a chemical speed mostly and the electronics is 100 to 1000 times (or more) faster. If we were to train a DNN with 1000 nodes for a year on a “problem”... how would we evaluate or deconstruct the way it arrives at conclusions? If we complicate that with what IBM has been working on for four years...an electronic synapse low power chip...
http://www.engadget.com/2014/08/07/ibm-synapse-supercomputing-chip-mimics-human-brain/
using Moore’s Law
>>That single-core prototype has now been significantly scaled up, leading to a new, production-ready SyNAPSE chip that blows past its predecessor with 1 million neurons, 256 million synapses and 4,096 neurosynaptic cores, all the while only requiring 70mW of power.<<
In five years, it should be 16 times more powerful. It is getting $53 million from DARPA. 4 billion synapses, hundreds of times faster than a brain synapse, the software is the limiting factor. It will drive your car, better than you will. Better than NASCAR drivers, a litter after that. And opening the black box, will lead to mostly scratching of the head wondering if studying phrenology would lead to more understandable conclusions.
DK