Post on 06-Apr-2018
transcript
8/3/2019 Economics of the Singularity
http://slidepdf.com/reader/full/economics-of-the-singularity 1/4
O
ur global economy would stupey a Roman
merchant as much as the Roman economy would
have conounded a caveman. But we would be sim-
ilarly amazed to see the economy that awaits our
randchildren, or I expect it to ollow a societal
discontinuity more dramatic than those brouht on by the ar icultural and industrial revolutions. The key, o course, is
technolog. A revolutionary speedup in economic rowth requires
an unprecedented and remarkable enablin tool. Machine intelli-
ence on a human level, i not hiher, would do nicely. Its arrival
could produce a sinularity—an overwhelmin departure rom
prior trends, with uneven and dizzyinly rapid chane thereater.
A uture shock to end uture shocks.
Stuffed into SkyScraperS by the billion,brainy bugbotS will be the knowledge workerSof the future by robin hanSon
Eoos OTe St
36 int • ieee Spectrum • ju ne 2008 www.SpEcTrum.iEEE.Org june 2008 • ieee Spectrum • in twww.SpEcTrum.iEEE.Org
Yes, this theory o mine is a social and
economic one, and thereore not as unail-inly accurate or testable as one in the
physical sciences. Nevertheless, social sci-
entists routinely make short-term orecasts
that hit the mark, and economists oten
oer insihtul orecasts about unprece-dented situations.
So indule me as I outline how we econ-
omists view technoloical chane. In sodoin, I hope to explain why it’s reasonable
to view past history as a series o abrupt,
seeminly unheralded transitions rom
one economic era to another, transitionsmarked by the sudden and drastic increase
in the rate o economic rowth. I will then
show why another sinularity is perhaps
just around the corner. Finally, I will out-
line its possible consequences.A complex device, like a tractor or a build-
in, can have thousands o parts, and each
part can rely on dozens o technoloies. Yet
in most cases even a spectacular ain in
the quality o one part bestows at best onlya small improvement on the whole. Keep
improvin a part in successive increments
o equal deree and you’ll et ever smallerains to the whole. This is the law o dimin-
ishin returns, and it applies not only to
devices and oranizations but to entire in-
dustries. Consider your personal computer:every couple o years its power-to-cost ratio
has doubled, and yet as you o rom one en-
eration to the next, you probably notice only
b r y a n c hr i s t i e d e s i gn
Proof #4 5/12/08 @ 11:16 am INT COLOR
8/3/2019 Economics of the Singularity
http://slidepdf.com/reader/full/economics-of-the-singularity 2/4
a small improvement as you plu away on
your word processors and spreadsheets.
It turns out that most o these small,innovative ains come not rom research
labs but rom hands-on builders and
users. So the more a thin ets used,
the more it tends to improve. It doesn’tmatter whether that thin is a physical
device, such as a car, or a social orani-
zation, such as a corporation.
I any lare system o interactinparts tends to improve by smooth ra-
dations, then we should expect systems
o systems, with their larer number o
components and interactions, to improve
even more smoothly. By this reason-in, the world economy should improve
most smoothly o all. The world econ-
omy consists o the larest number o interactin parts o any man-made sys-
tem, and everyone not stranded on an
uncharted island contributes to the
improvements in all those parts by usinthem. Finally, in each economic era the
question o whether rowth speeds up
or slows down depends on two compet-
in actors. Deceleration typically ensues
as innovators exhaust the easy ideas—
the low-hanin ruit. But accelerationalso ensues as the economy, by ettin
larer, enables its members to explore an
ever-increasin number o innovations.
We have the tools to measure the
world’s economic product not only ortoday—it’s about US $50 trillion per year—
but also or times lon past. A ew years
ao Anus Maddison, an economic his-torian at the University o Groninen, in
the Netherlands, plotted a raph o world
economic product—basically everythin
o value produced lobally: bananas, sub-marines, maazine articles, you name it.
It shows that rom 1950 to 2003, rowth
was relatively steady. Durin that time,
despite enormous technical chane, noparticular technolog let much o a n-
erprint on the data; no short-term accel-
erations in rowth could be attributed to
this or that technoloical development.Also, Maddison’s data oer little support
or the idea that innovation and rowth
have been acceleratin recently.Now look at the data or world product
over the past 7,000 years, estimated by
Bradord DeLon, an economic historian at
the University o Caliornia, Berkeley. The
data here tell a somewhat dierent story.For most o that time, rowth proceeded at
a relatively steady exponential rate, with a
doublin o output about every 900 years.
But within the past ew centuries, some-thin dramatic happened: output bean
doublin aster and aster, approachin a
new steady doublin time o about 15 years.
That’s about 60 times as ast as it had beenin the previous seven millennia.
W
e call this transition
the Industrial Revolu-
tion, but that does notmean we understand it
well or even know pre-
cisely how and why it arose. But what-ever the Industrial Revolution was,
clearly it was an event worthy o the
name “sinularity.”
I we look urther back, we see whatappears to be at least one previous
sinularity—the transition to an econ-
omy based on ariculture. And slow as
economic rowth durin the aricul-
tural era may seem in the atermath o
the Industrial Revolution, it was actu-ally lihtnin ast compared with that o
the economic era that came beore, which
was based on huntin and atherin.
In the rouhly 2 million years our
ancestors lived as hunters and ath-erers, the population rose rom about
10 000 protohumans to about 4 million
modern humans. I, as we believe, therowth pattern durin this era was
airly steady, then the population must
have doubled about every quarter mil-
lion years, on averae. Then, beinninabout 10 000 years ao, a ew o those
4 million humans bean to settle down
and live as armers. The resultin com-
munities rew so ast that they quicklyaccounted or most o the world popula-
tion. From that time on, the armin pop-
ulation doubled about every 900 years—
some 250 times as ast as beore.Our understandin o the existence,
nature, and relevance o these transi-
tions clearly becomes more specula-tive the urther back we look in time[see sidebar, “How Many Sinularities
Have There Been?”]. There may well
have been two earlier sinularities that
started eras o this sort, althouh ourability to identiy them and weih their
relevance is very speculative. I suest
an era dened by the rowth o the brain
rom the emerence o animal li e to theirst protohumans and perhaps an ear-
lier era dened by the rowth o the u ni-
verse rom a time shortly ater the bi
ban to the rst animals.So we have perhaps ve eras durin
which the thin whose rowth is at issue—
the universe, brains, the huntin economy,
the armin economy, and the industrial
economy—doubled in size at xed inter-vals. Each era o rowth beore now, how-
ever, has eventually switched suddenly to
a new era havin a rowth rate that was between 60 and 250 times as ast. Each
switch was completed in much less time
than it had taken the previous reime to
double—rom a ew millennia or the ari-cultural revolution to a ew centuries or
the industrial one. These switches consti-
tuted sinularities.
Whatever may have been the key
innovations behind these transitions, it is
clear that they were ar more potent thansuch amiliar textbook examples o reat
innovations as re, writin, computers, or
plastics. Most innovations happen within
a iven rowth era and do not chane its basic nature, includin its basic rowth
rate. A ew exceedinly rare innovations,
however, do suddenly chane everythin.
One such innovation led to ariculture;another led to industry.
Thereore, we must admit that another
sinularity—at least the third one, and
perhaps the th, dependin on how you
count—could lie ahead. Furthermore,data on these previous apparently simi-
lar sinularities are some o the ew con-
crete uides available to what such a tran-sition miht look like. We would be ools
i we condently expected all patterns to
continue. But it strikes me as pretty ool-
ish to inore the patterns we see.I a new transition were to show the
same pattern as the past two, then rowth
would quickly speed up by between
60- and 250 -old. The world economy,which now doubles in 15 years or so, would
soon double in somewhere rom a week
to a month. I the new transition were
as radual (in power-law terms) as theIndustrial Revolution was, then within
three years o a noticeable departure
rom typical uctuations, it would beinto double annually, and within two moreyears, it miht row a mil lion-old. I the
new transition were as rapid as the ari-
cultural revolution seems to have been,
chane would be even more sudden.Thouh such rowth may seem pre-
posterous, consider that in the era o
huntin and atherin, the economy dou-
bled nine times; in the era o armin, itdoubled seven times; and in the cur-
rent era o industry, it has so ar doubled
10 times. I, or some as yet unknown
reason, the number o doublins is sim-ilar across these three eras, then we
seem already overdue or another tran-
sition. I we instead compare our era
with the era o brain rowth, which dou-
bled 16 times beore humans appeared,we would expect the next transition by
around 2075.
What innovation could possiblyinduce so abulous a speedup in economic
rowth? It is easier to say what could not.
Because o diminishin returns, no chane
that improved just one small sector o theeconomy could do the trick. In advanced
countries today, armin, minin, energ,
communications, transportation, and
construction each account or only a small
percentae o economic activity. Even so
extraordinary an innovation as radicalnanotechnolog would do no more than
dramatically lower the cost o capital or
manuacturin, which now makes up less
than 10 percent o U.S. GDP.No, the next radical jump in economic
rowth seems more likely to come rom
somethin that has a proound eect
on everythin, because it addresses theone permanent shortae in our entire
economy: human time and attention.
They are by ar the most productive
components o today’s economy. About
two-thirds o all income in the rich coun-tries is paid directly or waes, and much
o the remainin third represents indi-
rect costs o labor. (For example, corpo-rate income larely reects earlier eorts
by entrepreneurs.) So any innovation that
could replace or dramatically improve
human labor would be a very bi deal.
One of the pillars o
the modern sinularity
hypothesis in its manyorms is that intellience
is a eneral elixir, able to
cure many i not all economic ailments.
Typically, this belie is expressed in theorm o an arument that the arrival o
very intellient machines will produce
the next sinularity. Some people hopethis arrival will ollow a new Einstein,who will discover a powerul eneral
theory o intellience applicable to those
machines. Others envision an “intelli-
ence explosion” via a series o poweruldesin innovations, beinnin with one
that would make machines smart enouh
to help us quickly nd a second innova-
tion, allowin even smarter machines,and so on. A ew even imaine innova-
tions so unprecedentedly potent that a
sinle machine embodyin the rst inno-
vation could o throuh the entire inno-vation series by itsel, unnoticed, within a
week, and then take over the world.
There are many views on how intel-
lience miht arise in a machine. One
arument holds that hardware is thecritical limitin actor and predicts that
human-level machine intellience will
come soon ater we have computer hard-ware whose perormance is comparable
with that o the human brain.
Another arument ocuses on knowl-
ede as the true limitin actor. Thisview is behind several hue artiicial-
intellience database projects, includ-
in Cyc, under construction or 23 years
38 int • ieee Spectrum • june 2008 www.SpEcTrum.iEEE.Org june 2008 • ieee Spectrum • intwww.SpEcTrum.iEEE.Org
Te o eoo, oobes 15 es, o sooobe eek to ot
ExpErT ViEw:
J CsWHO HE ISSenior Research Scholar,the International Instituteor Applied Systems Analysis,in Laxenburg, Austria, andcoounder o the Kenos Circle,a Vienna-based society orexploration o the uture.Builds computer simulationso complex human systems,like the stock market,highway trac, and theinsurance industry. Author opopular books about science,both ction and nonction,including The CambridgeQuintet, a ctional account oa dinner-party conversationabout the creation o athinking machine.
SINGULARITYWILL OCCURWithin 70 years
MACHINECONSCIOUSNESSWILL OCCURQuestionable
MOORE’S LAWWILL CONTINUE FOR20 more years withcurrent technology
THOUGHTS“I think it’s scientically andphilosophically on soundooting. The only real issueor me is the time rameover which the singularitywill unold. [The singularityrepresents] the end o thesupremacy o Homo sapiens as the dominant species onplanet Earth. At that pointa new species appears, andhumans and machines willgo their separate ways, notmerge one with the other. I donot believe this necessarilyimplies a malevolentmachine takeover; rather,machines will become
increasingly uninterested inhuman afairs just as we areuninterested in the afairso ants or bees. But in myview it’s more likely thannot that the two species willcomortably and more or lesspeaceully coexist—unlesshuman interests startto interere with those othe machines.”
J ohn c a s t i
How ManySingularities HaveThere Been?
ThE TwO SOlidly dEmOnSTraTEd
stes, te t st
evotos, e t tte .
wee tee stes beoe
st? i e ook bk
te te, e eve soe
oes o ot tt e s e
tstos to ste oes. Fo ee,
te-tees vst ee
te e se toot te o
boo sot eo o te. Tt
tsto et s e ossbe b
se ovtos te s eoto b. Beoe tt tsto,
te te eeee o s
soe 500 o es ee, te est
bs obe sze o eve
30 o es—ess t 1 eet o te
ot te o bs.
look te bk, t s t to
o-te tes tt ve ve te
o te eeee o s. St,
t s teest to ote tt te voe o
o e 14 bo-e-o vese s
bee e eoet e to
steos “k ee,” t ob
te o 3 bo es—bot 1 eet te
ot te o b s ze.
O ose, bese e ve o ete
teo s vos ot oes
tstos so be ete,
stes betee te be e
oee. Bt te o osttte
eeets, o te so tt vst es
e see ovet. —r.h.
t h e
S i n g u l a r i t y |
S p e c i a l
r e p o r t
8/3/2019 Economics of the Singularity
http://slidepdf.com/reader/full/economics-of-the-singularity 3/4
and now at Cycorp in Austin, Texas. Cyc
now possesses millions o pieces o com-
monsense knowlede, added mostly by
hand. Eventually, Cyc may know enouhto bein to read and assimilate all writ-
ten knowlede, and the more it knows,
the aster it should be able to learn. Soit is possible, thouh hardly inevitable,that Cyc will eventually undero a rapid
knowlede explosion.
I nd those scenarios interestin but
unlikely to come to pass anytime soon.Reardin advanced machine intelli-
ence, my uess is that our best chance
o achievin it within a century is to
put aside the attempt to understand themind, at least or now, and instead sim-
ply ocus on copyin the brain.
This approach, known as whole
brain emulation, starts with a realhuman brain, scanned in enouh detail
to see the exact location and type o each
part o each neuron, such as dendrites,
axons, and synapses. Then, usin mod-
els o how each o these neuronal com-ponents turns input sinals into output
sinals, you would construct a computer
model o this specic brain. With accu-rate enouh models and scans, the nal
simulation should have the same input-
output behavior as the oriinal brain.
It would, in a sense, be the “uploadedmind” o whoever served as the template.
Whether the emulation indeed consti-
tutes a person and whether that person
in achievement, but it is not imme-
diately obvious that it would launch a
new era o much aster rowth, withdoublin times measured in months
or less. Ater all, more and more capa-
ble machines have been replacin and
aidin humans or centuries with-out sparkin such an explosion. To
answer that objection, we’ve ot to
start with the undamentals: what eco-
nomic theory says about rowth rates.
T
o keep a modern economy
thrivin, we must accomplish
many mental tasks. Some peo-
ple (we call them enineers)have to desin new products,
systems, and services. Other people have
to build, market, transport, distribute, andmaintain them, and so on. These myriad
tasks are mostly complements, so that
doin one task better increases the value
o doin other tasks well. But or each task,humans and machines may also be substi-
tutes; it can be a wasted eort to have them
both do the same task.
The relative advantaes o humansand machines vary rom one task to the
next. Imaine a chart resemblin a top-
oraphic cross section, with the tasks
that are “most human” ormin a humanadvantae curve on the hiher round.
Here you nd chores best done by humans,
like ourmet cookin or elite hairdressin.Then there is a “shore” consistin o tasksthat humans and machines are equally
able to perorm and, beyond them an
“ocean” o tasks best done by machines.
When machines et cheaper or smarteror both, the water level rises, as it were,
and the shore moves inland.
This sea chane has two eects. First,
machines will substitute or humans bytakin over newly “ooded” tasks. Second,
doin machine tasks better complements
human tasks, raisin the value o doin
them well. Human waes may rise or all,dependin on which eect is stroner.
For example, in the 1920s, when the
mass-produced automobile came alon,
it was produced larely by machines,
with human help. So machines domi-nated that unction—the assembly o cars.
The resultin prolieration o machine-
assembled cars raised the value o relatedhuman tasks, such as desinin those
cars, because the inancial stakes were
now much hiher. Sure enouh, auto-
mobiles raised the waes o machinistsand desiners—in these cases, the com-
plementary eect dominated. At the same
time, the automobile industry lowered the
pay o saddle makers and stable hands, an
example o the substitution eect.
So ar, machines have displaced rel-atively ew human workers, and when
they have done so, they have in most
cases reatly raised the incomes o other
workers. That is, the complementaryeect has outweihed the substitution
eect—but this trend need not continue.
In our raph o machines and humans,
imaine that the ocean o machine tasksreached a wide plateau. This would
happen i, or instance, machines were
almost capable enouh to take on a vast
array o human jobs. For example, it
miht occur i machines were on the verycusp o human-level conition. In this
situation, a small additional rise in sea
level would ood that plateau and pushthe shoreline so ar inland that a hue
number o important tasks ormerly in
the human realm were now achievable
with machines. We’d expect such a wideplateau i the cheapest smart machines
were whole-brain emulations whose rel-
ative abilities on most tasks should be
close to those o human beins.In such a scenario, the economy would
start rowin much aster, or three rea-
sons. First, we could create capable
machines in much less time than it takesto breed, rear, and educate new human
workers. Bein able to make and retire
machine workers as ast as needed couldeasily double or quadruple rowth rates.
Second, the cost o computin has lon
been allin much aster than the econ-
omy has been rowin. When the work-
orce is larely composed o computers,the cost o makin workers will there-
ore all at that aster rate, with all that
this entails or economic rowth.
Third, as the economy beins rowinaster, computer usae and the resources
devoted to developin computers will
also row aster. And because innova-
tion is aster when more people use andstudy somethin, we should expect com-
puter perormance to improve even aster
than in the past.
Toether these eects seem quite
capable o producin economic dou- blin times much shorter than anythin
the world has ever seen. And note that
this orecast does not depend on therate at which we achieve machine intel-
lience capabilities or the rate at which
the intellience o machines increases.
Merely havin computer-like machinesable to do most important mental tasks
as well as humans do seems sufcient to
produce very rapid rowth.
An emulation o a brain could merely
do what that brain can already do,
althouh i done in suiciently power-
ul hardware, the conition miht occuraster. Still, even i all we were able to
achieve was a computer with the men-
tal powers o a particular human, thatwould be more than just interestin—it would also be incredibly useul.
Thouh it miht cost many billions o
dollars to build one such machine, the
rst copy miht cost only millions andthe millionth copy perhaps thousands
or less. Mass production could then sup-
ply what has so ar been the one actor o
production that has remained criticallyscarce throuhout human history: intel-
lient, hihly trained labor.
Okay, so miht these machines beconscious, with wills o their own, and
i so, could they be selsh, even malevo-
lent? Yes, yes, yes, and yes. More on that
later; or now, let’s et back to the eco-nomic arument.
Creatin human-level intellect
in a machine would be an astound-
5000 B.C.
W o r l d p r o d u c t ( b i l l i o n s o f 1 9 9 0 U S d o l l a r s )
Year4000 B.C.
1
10
100
1,000
10 000
100 000
2000 B.C.3000 B.C. 1000 B.C. 0 A.D. 1000 A.D. 2000
Agricultural era
Industrial eraSource: Bradford DeLong,economics professor,UniversityofCalifornia,Berkeley
Spot the
tranSition:
ate e
o so ot
te to s
oto, teo eoo
took of. Fo te
st te eve,
ott e t
ose bove ee
sbsstee
eves. it’s bee
s eve se.
40 int • ieee Spectrum • ju ne 2008 www.SpEcTrum.iEEE.Org june 2008 • ieee Spectrum • in twww.SpEcTrum.iEEE.Org
has rihts is another story, to which I
will return later.
I current trends continue, we should
have computer hardware and brainscans ast and cheap enouh to support
this scenario in a ew decades. What may
well take loner are input-output modelsin sufcient detail or every relevant typeo human neuron part. But I think those
details will accrue in time. We already
have suicient models or some types
o neuronal components, athered ateronly a modest eort. And we have no rea-
son to expect the other types to be harder.
Project Blue Brain, a joint eort by IBM
and the Ecole Polytechnique Fédéralede Lausanne, in Switzerland, has made
some impressive proress: in December
2006, the project nished mappin andmodelin the 10 000-odd neurons and
30 million synapses in a rat’s neocortical
column. Similarly impressive, in 2004 a
Stockholm University team observedrealistic behavior in a simulation o
8 million neurons and 4 billion synapses.
But we still have ar to o.
ExpErT ViEw:
t.J. rdgsWHO HE ISFounder and CEO o CypressSemiconductor Corp., in SanJose, Cali., known or hisbrash opinions about the
business world and politics.Owner o the Clos de la Techwinery and vineyards, inCaliornia, where he’s tryingto make the best Americanpinot noir.
SINGULARITYWILL OCCURNever
THOUGHTS“I don’t believe intechnological singularities.It’s like extraterrestriallie—i it were there,we would have seen itby now. However, I dobelieve in somethingthat is more powerulbecause it is real—namely,exponential learning. Anexponential unction hasthe property that its slopeis proportional to its value.
The more we know, theaster we can learn.
“Technological transitionsare required to maintainan exponential rate olearning. The rst airplaneswere certainly not asgood as well-appointedtrains in moving massescomortably, but thetransition later provedessential to maintainingour progress in humanmobility. Gene splicing is abreakthrough technologybut has not yet done (orbeen allowed to do) a lot ormankind. That will change.
“I don’t believe in the goodold days. We will be reer,more well-educated andeven smarter in the uture—but exponentially so, not asa result o some singularity.”
c y p r e s s s e mi c ond u c t or
B eto o stete “oe ” o oeveseve s te tete
t h e
S i n g u l a r i t y |
S p e c i a l
r e p o r t
8/3/2019 Economics of the Singularity
http://slidepdf.com/reader/full/economics-of-the-singularity 4/4
42 int • ieee Spectrum • june 2008 www.SpEcTrum.iEEE.Org june 2008 • ieee Spectrum • intwww.SpEcTrum.iEEE.Org
Life in a robot economy would
not be merely a sped-up ver-
sion o our lives today. WhenI apply basic economic theory
and some common sense to this
scenario, I conclude that humans would
probably be neither the immortal, all-powerul ods that some hope or nor the
hated and hunted prey that some ear.
Yes, robot-human wars would be
possible, but it is important to remem- ber that ew dierences between
humans ever lead to war. We do not
ear that the short will conspire to mur-
der the tall in their sleep, nor that the
riht-handed will exterminate the let-handed. Short, tall, let-handed, and
riht-handed people all trade with,
beriend, and marry one another withabandon, makin such wars almost
unthinkable. Instead, wars today hap-
pen between larely separate nations
and ethnic roups. Similarly, robotswell-interated into our economy would
be unlikely to exterminate us.
Would robots be slaves? Laws could
conceivably ban robots or only allow
robots “born” with enouh wealth to
aord a lie o leisure. But without lobaland draconian enorcement o such laws,
the vast wealth that cheap robots oer
would quickly induce a sprawlin, unruly
black market. Realistically, since modestenorcement could maintain only modest
restrictions, hue numbers o cheap (and
thus poor) robots would probably exist;
only their leal status would be in question.
Dependin on local politics, cheap robotscould be “undocumented” illeals, leal
slaves o their creators or owners, “ree”
minds rentin their bodies and servicesand subject to “eviction” or nonpayment,
or ree minds saddled with debts and sub-
ject to “repossession” or nonpayment.
The ollowin conclusions do not muchdepend on which o these cases is more
common. For example, in any o these
cases human waes would rise or all rap-
idly, dependin on the shape o the human
advantae landscape. Ater the ood o the
plateau, there miht still be some moun-tain peaks o human tasks let. Some rich
people miht still want to be served and
entertained by real human beins. So or
those jobs, human waes could rise. But i in the end the machine ocean completely
inundated all o Task-Land, then waes
would all so ar that most humans would
not, throuh their labor alone, be able tolive on them, thouh they miht work or
other reasons.
In either case, human labor would no
loner earn most income. Owners o real
estate or o businesses that build, main-tain, or supply machines would see their
wealth row at a abulous rate—about as
ast as the economy rows. Interest rateswould be similarly reat. Any small part
o this wealth should allow humans to
live comortably somewhere, even i not
as all-powerul ods.Because copyin a machine mind
would be cheap, trainin and education
would cost no more than a sotware update.
Instead o lon years to train each worker,a ew machines would be trained i ntensely,
and then many copies would be made o
the very best trainees. Presumably, stron
security would prevent bootle copies.Oranizational decision cycles
would shorten, avorin streamlined,
decentralized processes run by ast
machine minds in key positions o
authority. Fast minds could be whole- brain emulations sped up relative to
human brains. This scenario would
marinalize slow bureaucratic human
committees, reulators, and the like.Fast rowth rates would likely discour-
ae slow lon-distance transport and
encourae local production.
Some robots responsible or admin-
istration, research, law, and other coni-tive work miht live and work entirely in
virtual environments. For others, crude
calculations suest that tiny bodies aew millimeters tall, with sped-up minds
to match their aster body motions, miht
allow insectlike urban densities, with
many billions livin in the volume o acurrent skyscraper, payin astronomical
rents that would exclude most humans.
As emulations o humans, these crea-
tures would do the same sorts o thins
in their virtual realities and skyscrapers
that humans have done or hundreds o thousands o years: orm communities
and coalitions, all in love, ossip, arue,
make art, commit crimes, et work done,
innovate, and have un. Just as arminwas more alien to our human nature than
huntin and atherin, and industry was
more alien still, their world would be even
more distant rom human oriins. Buthuman nature seems exible enouh to
accommodate such chanes.
The population o smart machines
would explode even aster than the
economy. So even thouh total wealthwould increase very rapidly, wealth per
machine would all rapidly. I these smart
machines are considered “people,” thenmost people would be machines, and per-
person wealth and waes would quickly
all to machine-subsistence levels, which
would be ar below human-subsistencelevels. Salaries would probably be just
hih enouh to cover the rent on a tiny
body, a ew cubic centimeters o space, the
odd spare part, a ew watts o energ andheat dumpin, and a Net connection.
While copyin would make robot
immortality easible in principle, ew
robots would be able to aord it. Andwhen reproduction via copyin domi-
nates, ew robots would be able to aord
robot versions o human children.While whole-brain-emulation robotswould be copies o particular humans,
we should expect vast inequality in copy
rates. Investors who paid the hih costs
or scannin a human brain would care-ully select the ew humans most likely
to be lexible, cooperative, and produc-
tive workers, even while livin a short,
hardscrabble, childless, and alien lie inrobotic bodies or virtual ofces. Investors
who paid or copyin existin machine
minds would select robots with a track
record o achievin this ideal. As a result,there would be lare rst-mover advan-
taes and wi nner-take-all eects. For
example, i docile minds turned out to be
the most productive, then the robot world
miht consist mainly o trillions o copieseach o a ew very docile human minds.
In this case, the meek would indeed
inherit the Earth. o
TO PROBE FURTHER For additional
resources on reconstructing the deep economic
past, speculations on a rapid intelligenceexplosion, and the likely effects of machine
intelligence on economic growth, see http://
spectrum.ieee.org/jun08/singularityprobe.
Wes o so tt ost so ot ve o te
t h e
S i n g u l a r i t y
|
S p e c i a l
r e p o r t