The Future Role of Computation in Science and Society

The Future Role o~ Computation
in Science and SocIety
Patrick Suppes
1 Some Examples of Large-Scale Computation
Let me begin with three simple examples, well, perhaps they are not so simple. The
first consists of the extreme demands for large-scale computation of data coming
from the Large-Scale Hadron Collider (LHC) at the laboratory of Conseil Europeen
pour La Recherche NucLeaire (CERN).
The LHC produces at design parameters over 600 millions collisions (~10 9 )
proton-proton collisions per second in ATLAS or CMS detectors. The amount of
data collected for each event is around 1 MB (1 Megabyte).
10
9
collisions/s x 1 Mbyte/collision
=
=
1015 bytes/s
1 PB/s (1 Petabyte/second)
Since 1 DVD ~5GB: 200,000 DVDs per second would be filled, or about 6,000
IPods (ones with 160 GB of storage) per second!
A trigger is designed to reject the uninteresting events and keep the interesting
ones. For example, the ATLAS trigger system is designed to collect about 200 events
per second.
200 events/s x 1 Mbyte
= 200 MB/s
Taking two shifts of 10 h per day, and about 300 days per year:
200 MB/s x 2 x 10 x 3,600 x 300 ~ 4 . 1015
bytes/year
= 4 PB/year
p. Suppes (181)
Center for the Study of Language and Information, Stanford University,
Ventura Hall, Stanford, CA 94305-4115, USA
e-mail: [email protected]
M.e. Galavotti et al. (eds.), New Directions in the Philosophy of Science, The Philosophy
of Science in a European Perspective 5, DOl 10.1007/978-3-319-04382-1_3,
© Springer International Publishing Switzerland 2014
35
P. Suppes
36
Collectively, the LHC experiments produce about 15 petabytes of raw data each
year that must be stored, processed, and analyzed.
The second example is the astronomers' square kilometer array (SKA).
Astronomers will need a top ranking supercomputer to combat the data deluge
from SKA. The amount of computer data generated by the entire world in 2012
will need to be stored in a single day for the world's most powerful telescope - the
Square Kilometre Array (SKA) - and the International Centre for Radio Astronomy
Research (ICRAR) is gearing up to meet that unprecedented need.
ICRAR scientists say the $2 billion SKA will generate one exabyte of data - a
million terabytes (or one quintillion bytes) - every day while it searches the sky with
the power to detect airport radars in other solar systems 50 light years away.
There is a potentially important use of the computing power of SKA. This kind
of computational focus is essential to make accurate predictions of the paths of
meteorites or other bodies entering the solar system, some with a possible trajectory
close to that of Earth. The better the forecast of the past of the body headed toward
the general trajectory of Earth, the better the chance of avoiding a disaster of major
proportions.
On February 15, 2013 there was a massive splintering of a meteorite over
Siberia which would have caused enormous damage if it had been over a major
metropolitan center. This kind of future event would focus the entire popUlation
of our planet, if it were to have a magnitude considerably greater than that of the
recent Siberian meteorite. It is somewhat surprising that until recently astronomers
paid little attention to this kind of possibility. But computation limitations played a
role in the neglect.
It was understood how difficult it is to make observations sufficiently early to
forecast with any real accuracy the trajectory of a meteorite, comet or other foreign
body approaching Earth. With SKA, we now have that ability more than at any
other time in the past. No doubt, in the future better efforts will concentrate even
more on this kind of problem. Another aspect of this example of the need for largescale computation is the sheer number of potential candidates for the cause of such
a disastrous event.
This example exhibits well a feature of what we may expect from ever better
methods of computation. On the one hand, we will learn much more about the world
we live in and understand it better. Many positive things will follow as a result. But
we will also have a deeper and a more troubling view of the disasters that may lie
ahead in the not-too-distant future.
Here is a third example from meteorology.
In .196 Edward N. Lorenz (a meteorol gi t and a mathematician) wrote II
remarkable paper. Lorenz eareh for a three-dimensional set of ordinary differential equations whkh would mod I orne of the unpredictable behavior which we
normally a oeiate with the weather. The equation. which he eventually hit LIpan,
carne from a model of fluid convection. They are
dx
dt
- = a (y -
x)
37
.
tation in Science and Society
The: FUlure Rol e 01 Compu
I' I ,\ lIulII 'ricall),
I 1~;p\lle I wllll il111 lo.' hc
I lenl " Illill iOIl~ I l'oJcctcd
nllI the x. : pl llnc (I == 10,
"'"
40
~. r =28.()
30
r
z
20
10
o
-10
10
20
x
dy
- = rx - y - Xz
dt
(*)
dz
- = xy -bz
dt
Where (J, rand b are three real positive parameters.
Briefly, the original derivation (Lorenz 1963, 1979) can be described as follows.
A two-dimensional fluid cell is warmed from below and cooled from above and
the resulting convective motion is modelled by a partial differential equation. The
variables in this partial differential equation are expanded into an infinite number of
modes, all but three of which are then set identically to zero. The three remaining
modes give the equations (*). Roughly speaking, the variable x measures the rate of
convective overturning, the variable y measures the horizontal temperature variation,
and the variable z measures the vertical temperature variation. The three parameters
(J, r and b are respectively proportional to the Prandtl number, the Rayleigh number,
and Some physical proportions of the region under consideration; consequently, all
three are taken to be positive.
For wide ranges of values of the parameters, approximate solutions to the
equations (*), calculated on a computer, look extremely complicated.
Figure 1 shows the projection onto the x, z plane (y = constant) of one such
solution, when
.
(J
= 10, b = ~
3
and r = 28. Note that the trajectory shown does not
llltersect itself if we consider the full three-dimensional picture. The crossings in
Fig. 1 are the result of projection onto two dimensions.
38
P. Suppes
These first examples arc exotic. They show the new directions in science al their
most extreme. [n fact. the Lorenz-Iypcexample is nlC3nliO be a negative one, which
shows thm in principle Ihere is no hope for predicting the wealher in detail at
any long range. This conclusion is now pretty generally accepted for our present
regime of scientific meteorology without a scientific revoluti on thm we cannot now
foresee. It is reasonable to say thai it has been shown decisively thaI the limitations
of science arc reflected in everyday life in no better way than they arc in limitations
on prediction of Ihe wemher, for even 2 or 3 weeks in advance. The source of this
negative finding is easy to locate. In a general conceptual way. it is the enormous
complexity o f the weather system itself. It is nc t scme simple Newtcnian system
'Of a few particles, say twc, the mest which we can study in the greatest possible
detail . (It is imponanttc remember that the general thecry 'O f even three Newton ian
particles is unmanageable.)
It is hepeless \0 predict the behavior. with any precision, of any aClUal weather
system fcr an extended period 'O f time. The cutccme 'Of thi s clear example is Ihe
conclusi'O n Ihal we are optim istic, indeed aSlonishinglycverly 'Optimistic. if we think
Ihat all the major problems of modern science can, without doubt, in due time be
solved by suffiCiently powerful ccmputaticn methods.
This extended example 'Of a simplified model 'Of the convection of the atmcsphere
illustrates an importnnl strategy in using complicated ccmputation models to prove
that a si mplified model of something like the atmospherecannol be fully understood
by the methods of analysis mathematically currently available. If such simplified
examples can be shown to be impossible to find 11 sclution for, then it surely follows
that the full system will exhibit a si milar impossibility. The virtue of such si mplified
models is that we can study very thoroughly the unpredictable behavior. Usually
the sources are parameters enormously sensitive to initial 'Or boundary conditions,
but not entirely. Turbulence, which contributes t'O the unpredictable behavior of the
weather. is an example that in its unpredictability is not entirely dependent on initial
and boundary conditions, but on the motion 'O f the !luid itself.
What such negative examples as those of Lorenz shcw is that the over optimism
of popu lar predictions about the continucd success of science are often exaggerated.
Not enough IIttention is paid to how difficu lt it is to solve any real problem
o f any complexity. My favorite cxample is the NewlOnian mechanics of point
particles. A simpler physical model of any interest can scarcely be thought of.
But. as mentio nccl earlier, consider the situation as it is today, we have a superb
understanding of the behavior of one particle by itself with specified forces. and
a good understanding 'O f twc particles forming an isolated system with Newtonian
forces of mutual interaction. But even this isolated kind of system is in genel'lll
unmanageable, from a ccmputntional standpoint as soon as we reach the problem
of predicti ng the behavior of three such particles. In terms o f the ordi nary affairs of
the world, this seems absurd. nnd yet this is an important and now well recognizetl
limitntion of that stronghold of de term inistic science, classical particle mechanics.
No doubt in the dccades ahead, we shall learn more and more about the beha viOf
of such systems, and yet the results o f Lorenz and others show Ihat, within the
'c
Th~ !'ulUre Rolc ot
am
putation in Science and Society
39
'k of mathemati al Lhollght, a full and complete olution will not
,
f
..
one mIght formulate. The problem 0 ' predlctmg
~ r S<; II" ~hC Irajectorie of foreign bodie comi ngcJose to Earth i a vivid example
curSI ) ' ccled and strong I'1I1l1tauon
. ,
0 f t hc pre d'Icta b'l
I 'Lty 0 f even wIlat arc
of III' lIlle. I,
.
, ,
. '
011 " plUally very simple sy lem ' T~e p01l1t of mentlonmg ,stich sy. tcm here I
t make clear that even the mo t rna. stve large-scale.cbmplitatlOual method we can
ni \ I1visage will not solve mrulY ImpJc problem 111 any complete form.
L ' \ me conclude mi. di cussion of cientific example with ome more po itive
on' . . ('or which we cxpect to make con iderable progres in the rea. onably near
ru tu r', The purpo e of the e example i to give a balrulced picture of the futw'c
r computation. All of these examples, ru weU as the earlier one, depend upon
mputfttion . but in the e last set of in lances I predict the future is positive, because
th dCllll1nds ~ r computation are not too difficult to meet.
Perhap the ea ie t way to find a large number of po itive examples i to look
at the revolution Lhat i occurringiu all part of medicine and a ociated healtb
. ciences, One computational a pect of direct importance in the modern digital world
i. thnt the sample ize of testing new drug can be managed andincrea ed by
s vera I orders of magnitude. At relatively little co t we have ju I begun to learn
how to use digital data on a given medical problem by looking ar many millions
of medical records in a short period of time. What can be done now would have
been unthinkable even 20 years ago. In fa l, one of the mo t noticeable feature. of
this work is its widespread international character, We can read daily abou l di. ease
spreading in Africa, China, Russia, Norway or any other part of the world abou t a
easily as we can read about diseases among ou r current neighbors.
A second matter that is less in the press but of great importance is the improvement in surgical proc~dures, many of which are ba. ed on cru'eful computation , or
often on the interpretive use of digitally based imaging devices which, either for
their construction or their use themselves depend upon ma ive · omputations.
A third medical example is the small computer chip of a pacemaker tJlat can
be inserted in 15 min of surgery and that can play such a radical role in the
cardiovascular health of the patient. Again we are deaEng with technology that
Would have surprised everyone 40 or 50 year ago in its power and implicity
of implementation, From all indications, tbis i ouly a imple example of the
Complicated process underway of ino'oducing evemlOre artificial computing power
lUto Our bodies. With nanotechnology, we can now think in a practical way of
computers in the bloodstream monitoring clo ely and contjnually many subtle
chemical and physical properties, which information can be easi ly ob en/ed and
automatically tran mined to the appropriate medical workstation.
WiLhin tl1e range of computation thaI have been discussed, perhaps the most
exciting and igoificant development is a current well-supported effort by governmenr and many private foundalions to llJ1der tand how the brain computes. Because
of my own intere t I will try to ay something more detailed about this area of
research,
,
'0
• 'nl I ra III I,; \
pre
I
, . ' II ~ for most problem
40
P. Suppes
1. The system ignaling is ele tromagnetic.
2. A reasonable hypothe j i that collection of neurons synchronize to approximate weakly coupled elecu'omagnelic 0 cillators to do system computation and
signaling.
3. I focus on one major problem: the physical account of brain computations in
talking, listening, writing and reading, particularly on verbal memory storage
and retrieved.
4. There are literally thousands of psychological papers on this problem, but no
detailed physical models.
5. This is what I call abstract psychology.
6. Here is a sketch of one physical model on using phase to recognize English
phonem in the brain.
7. EEG experimelll with thou ands of trial and about 32 GB of data.
8. Signal eem compo edofwaves between 2 and 9 Hz.
9. Each electromagnetic ine wave ha. a fj'equency pha e and amplitude.
10. Amplitude i of lillIe u e. So pha e of frequencies have a pattern for a given
phoneme such as p, b, t, etc.
11. Here are the mean phase brain patterns for four fi;equencies of six initial
phonemes (Wang et a1. 2012).
12. This is only a beginning, but promising.
13. There is also progress on the semantic side in terms of semantic associative
networks, but no time for any details here.
I perhaps have not stressed enough how very far we are from thoroughly
understanding the computations of the brain that are essential to our daily activities.
What is needed most, and because of its importance I emphasize it again, is
an understanding of the phy ical mechanism thal do the ontinual computation
required for walking talking, listening, and in general perceiving what i going
on around u , not as a talie picture, but as a continually changing environment
affecting not only hat we ce, but what we hear and toucll a well (Fig. 2).
No doubt given the complexity of the y rem, we will 'oon be seeing theorems
on the impo ibility of having a omplete theory of dle computation . A in the
ca e of the weather but of a still more urgent nature, we will pre on determined
to tlnd approximations tllat are ever more refined a research continue. But at no
time in the near future are we goi ng to have anything like a omplete under tanding
of how we are producing or listening to the sU'eam of natural language in ordinary
conver alion at about tJlree words per econd. At first glance, the computational
proee e required to upport the ea 'y and natural activi ties of talking and Ii tening
eern out of reach. But ju t a in the case of tJle weather WitJl0Ut bei.ng fooli h
in our prediction and too p simi tic about our findings we will discover much
that i fundamental and that can be of great u e in perfecting. if nothing el e, Lhe
conversations between us and our devices.
The
41
' on in Science and Society
' lfC011lPU ta tl
F\llUr~ R(l IC
p
(
t
b
9
v
z
o
f
s
o
o
I-4HZ -5Hz -6Hz -7HZ I
Fig. 2 Brain phase pattern of six phonemes
2 From Science to Society
Here I discuss two problems that may not be at the very center of scientific research,
but are of the greatest importance to society. The first is the possibility of the
indefinite extension of human life, and the second is the end of work for most people.
Let us begin with some data on the world's oldest humans. The record setter,
Jeanne Louise Calment, at death was 122 years old. She was born February 21,
1875 and died August 4, 1997.
Eight of the last nine world's oldest persons at death were all under 120 years.
The most prominent group were 6 at 114 years.
Let us look at Japan, a keeper of excellent records. In 1990, there were 3,000
Japanese more than 100 years old. The oldest was 114. In 2010, 44,000 Japanese
were more than 100 years old, and again the oldest were 114.
Some different data come from experiments and studies on calorie reduction
(CR), which was first studied in 1934. CR alone led to at least 50 % life extension
of short-lived animals, such as mice (in 1934) and later many kinds of insects. Of
almost as great importance; the animals and humans had better health as well.
In 1989 University of Wisconsin scientists started a study of 20 adult male rhesus
monkeys with 9-CR, 11-normal subjects. In 2009 it was too early for life extension
results, but the CR monkeys were much healthier.
P. Suppes
42
These (wo kinds of arguments plus others that arc easily given. but I have om itted ,
lead to the following conjectures or proposals for extendi ng human life.
I.
2.
3.
4.
5.
Use of stem cells to replace failing organs.
Careful monitoring of input of various chemicals. such as calcium.
More fundamental changes in gene and protein structures.
The big question: Replacing old brai ns with new ones.
Totally exaggerated hope: full body replacement.
I now turn 10 new technology and the end of jobs.
The role of humnns as [he most imponant factor of producliOfl is bound to diminish in the
same way thai lhe role of horses in agricultural production was first diminished and then
eliminated by lhe introduction o rlruclOfS.
Wassily Leonticf and Faye Duchin,
Tht Future Impact of .... U/omn/iOtI QlI lVorktr:J. 1986
I support the view of this paragraph wi th a variety o f considerations and data.
1. TIle tcchnology revol ution is re placing human beings with computers, robots and
other intelligent devices in every sector of employment in the g lOth11 economy.
As is well-known, mi llions o f workers have already been eliminated from many
kinds of labor, ranging from agricultural workers 10 bank clerks. The handwriting
is on the wall. as computers improve and become more intelligent, the nu mber
of jobs for which humans are needed will be negligible compared to the numbers
employed at present.
2. Here is an amusing. but significant kind of example. In order to avoid various
kinds of terrorism of human workers, Israelis have developed a melon pickcr
(ROM PER). It uses special sensors to determine whether a melon or othcr lype
o f fruil is ripe to pick.
Similar robots have been developed. and arc gelting smarter by the day, to
p low and seed fie lds, feed dairy cows, pigs and other stock. It is predicted thut by
the second half of this century we wi ll see a wide-spread development o f fully
automated and computer-driven fac tory farms.
Note that thedeclineofthc number of agricultural workers in the United States
and other advanced economics, in the twentieth century. was the most significall!
change in agriculture yet seen anywhere in the world.
3. No more factory workers. In the last part of the twentieth century, Japan was
famou s for its nine auto makers produc ing more than 12 million cars per year
with fewer than 600,000 workers. In eomrasl. Detro it auto makers employed
more than 2.5 million workers to produce about the same number.
1l is a common prediction lhat, by the cnd of this century. there will be [10
workers in the automobile factories, but on ly a small number managing the ever
larger num ber o f robots. Moreover, it has been widely noted thm these roboi S
work 24n. without benefits. coffee breaks. medical illnesses or demands for p~ y
increases.
4. By the end o f this century. blue-collar workers will be gone in the developed
cou ntries, and fading away in the rest o f the world.
The F1Jlurc Role 01
.1' ('
• Wilh
, C 1putation in Science and Society
43
011
"lli /l of cour e the e same remarks apply to the service industries,
,
.
.
llrUllce and all kind of retad and clencal work,
n1()ul I . ·
•
' 1
bn ll\.:1I1 1!. 11 •
J J ' r ' i, nl 1:1 ( topic, back to the ~a t: A future without work with a popUlation
I r Ari~[ crnr havjng II furure without work.
n III Jane Au ten' " Pride Gilt! Prejudice (1959), to mark her social status Elizabeth
13 'n n I shm'l Iy re ponded to Mr. Darcy that her fa ther was just as much a gentleman
be h . Will we return to the life of ari tocrats in the eighteenth and nineteenth
'l'nlUric. , or ladie and gentlemen, who would not dirty their hands in commerce,
liS the phra 'c went, or in any other kind of productive labor?
Will human become a breed of amateur managers telling their smart robots and
olher devices what to do but only in the mo t general term thaI reflect ignorance
r the technical detai l that mu st be ma, tered by the e new 'devices", not by human
workers'? If thi s i. what really comes aboLlt, then thc nalureofthe change will indeed
bl; dramotic. The new ma. tel" of Earth wi II be smart robots telling all of their devices
IlIld human. , a well, what to do, or rather, as in the case of the Aristocrats of Jane
Austin's days, what will be accepted as proper behavior of a non-working human
Aristocrat.
At the same time, these smart robots will be controlling their own revolution and
moving ever further away from satisfying any natural concept of humanity. In this
process, humans will become as extinct as dinosaurs are now.
I am sorry to say this seems the most likely future, but there remains a small
probability that we, as human can tay ahead of the games and be Aristocrats
in the intellectual tyle of Euclid, Apollonius, Claudius Ptolemy, the unknown
Mayan astronomer, Sir I aae Newton, Henry Cavendish, Pierre-Simon Laplace,
Jame Clerk Maxwell , and the many other who belong on this distinguished list.
Like competitors today from China to California and Cambridge, the mart
robots of tomorrow will engage in a dynamic competition to ee which can evolve
and develop the deepest new scientific cOllcep . My bel i till on the odd. favoring
the robots. The only saving thought is that what might still be the actual olltcome
will be a wonderful hybrid species of thinking creature . half human and half
robotic, with powerful computation potenlial on board , and truly amazing further
remote computational power easily and quickly available.
To end, here are some new philosophical questions generated by this discussion
of computation in the future.
1. ShOUld we accept an economically supported life of leisure for all those who do
not want to work?
2. Are there serious normative negative arguments against life extension if possible?
3. Should we normatively argue for the desirability of less people, given life
exten, ion and end of routine work?
4. But in I.ronger terms i there a meanjngfuJ optimum size of population, as work
For humans decrea es and alm ost cea e ,yet life extension continues?
5. Do our fundamenta l concerns for freedom of action and thought, as well as the
rule of law, apply with some possible changes, to robots as well?
44
P. Suppes
References
Austen, J. 1959. Pride and prejudice. New York: Dell.
Leontief, W.W., and F. Duchin. 1986. The future impact of automation on workers. New York:
Oxford University Press.
Lorenz, E.N. 1963. The predictability of hydrodynamic flow. Transactions of the New York
Academy of Sciences Serial II 25(4): 409-432.
Lorenz, E.N. 1979. On the prevalence of aperiodicity in simple systems. In Global analysis, ed.
M . Grmela and J.E. Marsden, 53-75. New York: Springer.
Wang, R., M. Perreau-Guimaraes, C. Carvalhaes, and P. Suppes. 2012. Using phase to recognize
English phonemes and their distinctive features in the brain. Proceedings of the National
Academy of Sciences 94(5): 20685-20690.