252
12 Digital Democracy
discrimination.
16 They may also be sensitive
17 to the data set, to the algorithm,
or even the hardware used. In many cases, there will be classification errors
18
and spurious correlations.
19 So, Chris Anderson’s dream,
20 according to which
Big Data would just tell us what is true or false, and what had to be done, has
NOT come true. Big Data has NOT made science obsolete. In the data deluge,
where there is ever more “dark” data
21 than ever before,
22 science is needed to
decide what data to look at and in what ways. Note that these problems do not
go away with more data.
5. When there is no solution for the world’s existential problems within the current
system (which, therefore, cannot be found by optimization), creativity and innovation is the right approach. That is, we need to think out of the box in order to
expand the solution space and find solutions outside the current system.
23 Note,
however, that innovation always challenges established ways of doing things—it
challenges the system. This may not happen, if people are punished for deviations from the grand plan of a superintelligent system. Hence, controlling people
and manipulating their behavior may create an even bigger problem. Solutions
derived from data of the past may reinforce a system that is not serving us
well. This can cause disasters, where there would otherwise be solutions (not
contained in the data of the past).
6. Tragically, if we do not have one centralized system optimizing the world, but
different companies trying to optimize it in parallel, with different goal functions,
this is not making things better. The problem is that each optimization implies
constraints—it reduces freedoms in order to reach the optimal solution. So, when
many optimization processes are happening in parallel, a lot of freedoms will be
sacrificed. Your car insurance, your health insurance, your doctor, your dentist,
your employer, your electricity provider, your sustainability guide…—all of them
will demand something, and some of them will probably ask you for contradictory
actions. You will become “a slave of many masters”, so to say. Compared to this,
it is easier to follow the demands of one master, as in the Chinese credit score
system, but this is totalitarian in nature. It always claims to be right. It cannot
be questioned (while in a system with competitive demands, one can at least say
“but X demanded something else—please sort it out with X”).
Given the above problems, I would say “the (idea of the) benevolent dictator
is dead”. Therefore, in the following, I will propose systems that are based on
16 Helbing [4].
17 See https://en.wikipedia.org/wiki/Sensitivity_analysis.
18 See https://en.wikipedia.org/wiki/False_positives_and_false_negatives.
19 Vigen [5].
20 The End of Theory: The Data Deluge Makes the Scientific Method Obsolete, Wired (June 23,
2008) https://www.wired.com/2008/06/pb-theory/.
21 Helbing [6].
22 The data volume generated doubles in less than a year, i.e. in one year we are producing more
data than in all the years before in human history.
23 Helbing [7].
12 Digital Democracy
discrimination.
16 They may also be sensitive
17 to the data set, to the algorithm,
or even the hardware used. In many cases, there will be classification errors
18
and spurious correlations.
19 So, Chris Anderson’s dream,
20 according to which
Big Data would just tell us what is true or false, and what had to be done, has
NOT come true. Big Data has NOT made science obsolete. In the data deluge,
where there is ever more “dark” data
21 than ever before,
22 science is needed to
decide what data to look at and in what ways. Note that these problems do not
go away with more data.
5. When there is no solution for the world’s existential problems within the current
system (which, therefore, cannot be found by optimization), creativity and innovation is the right approach. That is, we need to think out of the box in order to
expand the solution space and find solutions outside the current system.
23 Note,
however, that innovation always challenges established ways of doing things—it
challenges the system. This may not happen, if people are punished for deviations from the grand plan of a superintelligent system. Hence, controlling people
and manipulating their behavior may create an even bigger problem. Solutions
derived from data of the past may reinforce a system that is not serving us
well. This can cause disasters, where there would otherwise be solutions (not
contained in the data of the past).
6. Tragically, if we do not have one centralized system optimizing the world, but
different companies trying to optimize it in parallel, with different goal functions,
this is not making things better. The problem is that each optimization implies
constraints—it reduces freedoms in order to reach the optimal solution. So, when
many optimization processes are happening in parallel, a lot of freedoms will be
sacrificed. Your car insurance, your health insurance, your doctor, your dentist,
your employer, your electricity provider, your sustainability guide…—all of them
will demand something, and some of them will probably ask you for contradictory
actions. You will become “a slave of many masters”, so to say. Compared to this,
it is easier to follow the demands of one master, as in the Chinese credit score
system, but this is totalitarian in nature. It always claims to be right. It cannot
be questioned (while in a system with competitive demands, one can at least say
“but X demanded something else—please sort it out with X”).
Given the above problems, I would say “the (idea of the) benevolent dictator
is dead”. Therefore, in the following, I will propose systems that are based on
16 Helbing [4].
17 See https://en.wikipedia.org/wiki/Sensitivity_analysis.
18 See https://en.wikipedia.org/wiki/False_positives_and_false_negatives.
19 Vigen [5].
20 The End of Theory: The Data Deluge Makes the Scientific Method Obsolete, Wired (June 23,
2008) https://www.wired.com/2008/06/pb-theory/.
21 Helbing [6].
22 The data volume generated doubles in less than a year, i.e. in one year we are producing more
data than in all the years before in human history.
23 Helbing [7].