Вы находитесь на странице: 1из 34

The Singularity is Near

A Book Prcis
by
Raymond Kurzweil
Raymond Kurzweil's upcoming book, to be entitled
"The Singularity Is Near"
You will get $40 trillion just by reading this prcis and
understanding what it says. For complete details, see
below. (Its true that authors will do just about anything
to !eep your attention, but Im serious about this
statement. "ntil I return to a #urther e$planation,
howe%er, do read the #irst sentence o# this paragraph
care#ully.&
'ow bac! to the #uture( its widely misunderstood.
)ur #orebears e$pected the #uture to be pretty much
li!e their present, which had been pretty much li!e
their past. *lthough e$ponential trends did e$ist a
thousand years ago, they were at that %ery early stage
where an e$ponential trend is so #lat that it loo!s li!e
no trend at all. +o their lac! o# e$pectations was
largely #ul#illed. ,oday, in accordance with the
common wisdom, e%eryone e$pects continuous
technological progress and the social repercussions
that #ollow. -ut the #uture will be #ar more surprising
than most obser%ers reali.e( #ew ha%e truly internali.ed
the implications o# the #act that the rate o# change itsel#
is accelerating.
The Intuiti!e "inear #iew !ersus the
$istorical %&ponential #iew
/ost long range #orecasts o# technical #easibility in
#uture time periods dramatically underestimate the
power o# #uture technology because they are based on
what I call the 0intuiti%e linear0 %iew o# technological
progress rather than the 0historical e$ponential %iew.0
,o e$press this another way, it is not the case that we
will e$perience a hundred years o# progress in the
twenty1#irst century2 rather we will witness on the
order o# twenty thousand years o# progress (at today's
rate o# progress, that is&.
,his disparity in outloo! comes up #re3uently in a
%ariety o# conte$ts, #or e$ample, the discussion o# the
ethical issues that -ill 4oy raised in his contro%ersial
5I678 co%er story, 5hy ,he Future 8oesnt 'eed
"s. -ill and I ha%e been #re3uently paired in a %ariety
o# %enues as pessimist and optimist respecti%ely.
*lthough Im e$pected to critici.e -ills position, and
indeed I do ta!e issue with his prescription o#
relin3uishment, I nonetheless usually end up de#ending
4oy on the !ey issue o# #easibility. 6ecently a 'oble
9ri.e winning panelist dismissed -ills concerns,
e$claiming that, 0were not going to see sel#1replicating
nanoengineered entities #or a hundred years.0 I pointed
out that :00 years was indeed a reasonable estimate o#
the amount o# technical progress re3uired to achie%e
this particular milestone at today's rate of progress.
-ut because were doubling the rate o# progress e%ery
decade, well see a century o# progress11at today's
rate11in only ;< calendar years.
5hen people thin! o# a #uture period, they
intuiti%ely assume that the current rate o# progress will
continue #or #uture periods. =owe%er, care#ul
consideration o# the pace o# technology shows that the
rate o# progress is not constant, but it is human nature
to adapt to the changing pace, so the intuiti%e %iew is
that the pace will continue at the current rate. 7%en #or
those o# us who ha%e been around long enough to
e$perience how the pace increases o%er time, our
une$amined intuition nonetheless pro%ides the
impression that progress changes at the rate that we
ha%e e$perienced recently. From the mathematicians
perspecti%e, a primary reason #or this is that an
e$ponential cur%e appro$imates a straight line when
%iewed #or a brie# duration. +o e%en though the rate o#
progress in the %ery recent past (e.g., this past year& is
#ar greater than it was ten years ago (let alone a
hundred or a thousand years ago&, our memories are
nonetheless dominated by our %ery recent e$perience.
It is typical, there#ore, that e%en sophisticated
commentators, when considering the #uture,
e$trapolate the current pace o# change o%er the ne$t :0
years or :00 years to determine their e$pectations. ,his
is why I call this way o# loo!ing at the #uture the
0intuiti%e linear0 %iew.
-ut a serious assessment o# the history o#
technology shows that technological change is
e$ponential. In e$ponential growth, we #ind that a !ey
measurement such as computational power is
multiplied by a constant #actor #or each unit o# time
(e.g., doubling e%ery year& rather than just being added
to incrementally. 7$ponential growth is a #eature o#
any e%olutionary process, o# which technology is a
primary e$ample. )ne can e$amine the data
in di##erent ways, on di##erent time scales, and #or a
wide %ariety o# technologies ranging #rom electronic to
biological, and the acceleration o# progress and growth
applies. Indeed, we #ind not just simple e$ponential
growth, but 0double0 e$ponential growth, meaning that
the rate o# e$ponential growth is itsel# growing
e$ponentially. ,hese obser%ations do not rely merely
on an assumption o# the continuation o# /oores law
(i.e., the e$ponential shrin!ing o# transistor si.es on an
integrated circuit&, but is based on a rich model o#
di%erse technological processes. 5hat it clearly shows
is that technology, particularly the pace o#
technological change, ad%ances (at least&
e$ponentially, not linearly, and has been doing so since
the ad%ent o# technology, indeed since the ad%ent o#
e%olution on 7arth.
I emphasi.e this point because it is the most
important #ailure that would1be prognosticators ma!e
in considering #uture trends. /ost technology #orecasts
ignore altogether this 0historical e$ponential %iew0 o#
technological progress. ,hat is why people tend to
o%erestimate what can be achie%ed in the short term
(because we tend to lea%e out necessary details&, but
underestimate what can be achie%ed in the long term
(because the e$ponential growth is ignored&.
The "aw o' (ccelerating Returns
5e can organi.e these obser%ations into what I call the
law o# accelerating returns as #ollows(
7%olution applies positi%e #eedbac! in that the
more capable methods resulting #rom one stage o#
e%olutionary progress are used to create the ne$t
stage. *s a result, the rate o# progress o# an
e%olutionary process increases e$ponentially o%er
time. )%er time, the 0order0 o# the in#ormation
embedded in the e%olutionary process (i.e., the
measure o# how well the in#ormation #its a purpose,
which in e%olution is sur%i%al& increases.
* correlate o# the abo%e obser%ation is that the
0returns0 o# an e%olutionary process (e.g., the
speed, cost1e##ecti%eness, or o%erall 0power0 o# a
process& increase e$ponentially o%er time.
In another positi%e #eedbac! loop, as a particular
e%olutionary process (e.g., computation& becomes
more e##ecti%e (e.g., cost e##ecti%e&, greater
resources are deployed toward the #urther progress
o# that process. ,his results in a second le%el o#
e$ponential growth (i.e., the rate o# e$ponential
growth itsel# grows e$ponentially&.
-iological e%olution is one such e%olutionary
process.
,echnological e%olution is another such
e%olutionary process. Indeed, the emergence o# the
#irst technology creating species resulted in the
new e%olutionary process o# technology. ,here#ore,
technological e%olution is an outgrowth o#11and a
continuation o#11biological e%olution.
* speci#ic paradigm (a method or approach to
sol%ing a problem, e.g., shrin!ing transistors on an
integrated circuit as an approach to ma!ing more
power#ul computers& pro%ides e$ponential growth
until the method e$hausts its potential. 5hen this
happens, a paradigm shi#t (i.e., a #undamental
change in the approach& occurs, which enables
e$ponential growth to continue.
I# we apply these principles at the highest le%el o#
e%olution on 7arth, the #irst step, the creation o# cells,
introduced the paradigm o# biology. ,he subse3uent
emergence o# 8'* pro%ided a digital method to
record the results o# e%olutionary e$periments. ,hen,
the e%olution o# a species who combined rational
thought with an opposable appendage (i.e., the thumb&
caused a #undamental paradigm shi#t #rom biology to
technology. ,he upcoming primary paradigm shi#t will
be #rom biological thin!ing to a hybrid combining
biological and nonbiological thin!ing. ,his hybrid will
include 0biologically inspired0 processes resulting
#rom the re%erse engineering o# biological brains.
I# we e$amine the timing o# these steps, we see that
the process has continuously accelerated. ,he
e%olution o# li#e #orms re3uired billions o# years #or the
#irst steps (e.g., primiti%e cells&2 later on progress
accelerated. 8uring the >ambrian e$plosion, major
paradigm shi#ts too! only tens o# millions o# years.
?ater on, =umanoids de%eloped o%er a period o#
millions o# years, and =omo sapiens o%er a period o#
only hundreds o# thousands o# years.
5ith the ad%ent o# a technology1creating species,
the e$ponential pace became too #ast #or e%olution
through 8'*1guided protein synthesis and mo%ed on
to human1created technology. ,echnology goes
beyond mere tool ma!ing2 it is a process o# creating
e%er more power#ul technology using the tools #rom
the pre%ious round o# inno%ation. In this way, human
technology is distinguished #rom the tool ma!ing o#
other species. ,here is a record o# each stage o#
technology, and each new stage o# technology builds
on the order o# the pre%ious stage.
,he #irst technological steps@sharp edges, #ire, the
wheel@too! tens o# thousands o# years. For people
li%ing in this era, there was little noticeable
technological change in e%en a thousand years. -y
:000 *.8., progress was much #aster and a paradigm
shi#t re3uired only a century or two. In the nineteenth
century, we saw more technological change than in the
nine centuries preceding it. ,hen in the #irst twenty
years o# the twentieth century, we saw more
ad%ancement than in all o# the nineteenth century.
'ow, paradigm shi#ts occur in only a #ew years time.
,he 5orld 5ide 5eb did not e$ist in anything li!e its
present #orm just a #ew years ago2 it didnt e$ist at all a
decade ago.
,he paradigm shi#t rate (i.e., the o%erall rate o#
technical progress& is currently doubling
(appro$imately& e%ery decade2 that is, paradigm shi#t
times are hal%ing e%ery decade (and the rate o#
acceleration is itsel# growing e$ponentially&. +o, the
technological progress in the twenty1#irst century will
be e3ui%alent to what would re3uire (in the linear
%iew& on the order o# ;00 centuries. In contrast, the
twentieth century saw only about ;< years o# progress
(again at todays rate o# progress& since we ha%e been
speeding up to current rates. +o the twenty1#irst
century will see almost a thousand times greater
technological change than its predecessor.
,he +ingularity Is 'ear
,o appreciate the nature and signi#icance o# the
coming 0singularity,0 it is important to ponder the
nature o# e$ponential growth. ,oward this end, I am
#ond o# telling the tale o# the in%entor o# chess and his
patron, the emperor o# >hina. In response to the
emperors o##er o# a reward #or his new belo%ed game,
the in%entor as!ed #or a single grain o# rice on the #irst
s3uare, two on the second s3uare, #our on the third, and
so on. ,he 7mperor 3uic!ly granted this seemingly
benign and humble re3uest. )ne %ersion o# the story
has the emperor going ban!rupt as the AB doublings
ultimately totaled :C million trillion grains o# rice. *t
ten grains o# rice per s3uare inch, this re3uires rice
#ields co%ering twice the sur#ace area o# the 7arth,
oceans included. *nother %ersion o# the story has the
in%entor losing his head.
It should be pointed out that as the emperor and the
in%entor went through the #irst hal# o# the chess board,
things were #airly une%ent#ul. ,he in%entor was gi%en
spoon#uls o# rice, then bowls o# rice, then barrels. -y
the end o# the #irst hal# o# the chess board, the in%entor
had accumulated one large #ields worth (4 billion
grains&, and the emperor did start to ta!e notice. It was
as they progressed through the second hal# o# the
chessboard that the situation 3uic!ly deteriorated.
Incidentally, with regard to the doublings o#
computation, thats about where we stand now11there
ha%e been slightly more than B; doublings o#
per#ormance since the #irst programmable computers
were in%ented during 5orld 5ar II.
,his is the nature o# e$ponential growth. *lthough
technology grows in the e$ponential domain, we
humans li%e in a linear world. +o technological trends
are not noticed as small le%els o# technological power
are doubled. ,hen seemingly out o# nowhere, a
technology e$plodes into %iew. For e$ample, when the
Internet went #rom ;0,000 to C0,000 nodes o%er a two
year period during the :DC0s, this progress remained
hidden #rom the general public. * decade later, when it
went #rom ;0 million to C0 million nodes in the same
amount o# time, the impact was rather conspicuous.
*s e$ponential growth continues to accelerate into
the #irst hal# o# the twenty1#irst century, it will appear
to e$plode into in#inity, at least #rom the limited and
linear perspecti%e o# contemporary humans. ,he
progress will ultimately become so #ast that it will
rupture our ability to #ollow it. It will literally get out
o# our control. ,he illusion that we ha%e our hand 0on
the plug,0 will be dispelled.
>an the pace o# technological progress continue to
speed up inde#initelyE Is there not a point where
humans are unable to thin! #ast enough to !eep up with
itE 5ith regard to unenhanced humans, clearly so. -ut
what would a thousand scientists, each a thousand
times more intelligent than human scientists today, and
each operating a thousand times #aster than
contemporary humans (because the in#ormation
processing in their primarily nonbiological brains is
#aster& accomplishE )ne year would be li!e a
millennium. 5hat would they come up withE
5ell, #or one thing, they would come up with
technology to become e%en more intelligent (because
their intelligence is no longer o# #i$ed capacity&. ,hey
would change their own thought processes to thin!
e%en #aster. 5hen the scientists e%ol%e to be a million
times more intelligent and operate a million times
#aster, then an hour would result in a century o#
progress (in todays terms&.
,his, then, is the +ingularity. ,he +ingularity is
technological change so rapid and so pro#ound that it
represents a rupture in the #abric o# human history.
+ome would say that we cannot comprehend the
+ingularity, at least with our current le%el o#
understanding, and that it is impossible, there#ore, to
loo! past its 0e%ent hori.on0 and ma!e sense o# what
lies beyond.
/y %iew is that despite our pro#ound limitations o#
thought, constrained as we are today to a mere hundred
trillion interneuronal connections in our biological
brains, we nonetheless ha%e su##icient powers o#
abstraction to ma!e meaning#ul statements about the
nature o# li#e a#ter the +ingularity. /ost importantly, it
is my %iew that the intelligence that will emerge will
continue to represent the human ci%ili.ation, which is
already a human1machine ci%ili.ation. ,his will be the
ne$t step in e%olution, the ne$t high le%el paradigm
shi#t.
,o put the concept o# +ingularity into perspecti%e,
lets e$plore the history o# the word itsel#. +ingularity
is a #amiliar word meaning a uni3ue e%ent with
pro#ound implications. In mathematics, the term
implies in#inity, the e$plosion o# %alue that occurs
when di%iding a constant by a number that gets closer
and closer to .ero. In physics, similarly, a singularity
denotes an e%ent or location o# in#inite power. *t the
center o# a blac! hole, matter is so dense that its
gra%ity is in#inite. *s nearby matter and energy are
drawn into the blac! hole, an e%ent hori.on separates
the region #rom the rest o# the "ni%erse. It constitutes a
rupture in the #abric o# space and time. ,he "ni%erse
itsel# is said to ha%e begun with just such a +ingularity.
In the :D<0s, 4ohn Fon 'eumann was 3uoted as
saying that 0the e%er accelerating progress o#
technology...gi%es the appearance o# approaching some
essential singularity in the history o# the race beyond
which human a##airs, as we !now them, could not
continue.0 In the :DA0s, I. 4. Good wrote o# an
0intelligence e$plosion,0 resulting #rom intelligent
machines designing their ne$t generation without
human inter%ention. In :DCA, Fernor Finge, a
mathematician and computer scientist at +an 8iego
+tate "ni%ersity, wrote about a rapidly approaching
technological 0singularity0 in his science #iction no%el,
Marooned in Realtime. ,hen in :DDB, Finge presented
a paper to a '*+*1organi.ed symposium which
described the +ingularity as an impending e%ent
resulting primarily #rom the ad%ent o# 0entities with
greater than human intelligence,0 which Finge saw as
the harbinger o# a run1away phenomenon.
From my perspecti%e, the +ingularity has many
#aces. It represents the nearly %ertical phase o#
e$ponential growth where the rate o# growth is so
e$treme that technology appears to be growing at
in#inite speed. )# course, #rom a mathematical
perspecti%e, there is no discontinuity, no rupture, and
the growth rates remain #inite, albeit e$traordinarily
large. -ut #rom our currently limited perspecti%e, this
imminent e%ent appears to be an acute and abrupt
brea! in the continuity o# progress. =owe%er, I
emphasi.e the word 0currently,0 because one o# the
salient implications o# the +ingularity will be a change
in the nature o# our ability to understand. In other
words, we will become %astly smarter as we merge
with our technology.
5hen I wrote my #irst boo!, The *ge o# Intelligent
/achines, in the :DC0s, I ended the boo! with the
specter o# the emergence o# machine intelligence
greater than human intelligence, but #ound it di##icult
to loo! beyond this e%ent hori.on. 'ow ha%ing thought
about its implications #or the past ;0 years, I #eel that
we are indeed capable o# understanding the many
#acets o# this threshold, one that will trans#orm all
spheres o# human li#e.
>onsider a #ew e$amples o# the implications. ,he
bul! o# our e$periences will shi#t #rom real reality to
%irtual reality. /ost o# the intelligence o# our
ci%ili.ation will ultimately be nonbiological, which by
the end o# this century will be trillions o# trillions o#
times more power#ul than human intelligence.
=owe%er, to address o#ten e$pressed concerns, this
does not imply the end o# biological intelligence, e%en
i# thrown #rom its perch o# e%olutionary superiority.
/oreo%er, it is important to note that the nonbiological
#orms will be deri%ati%e o# biological design. In other
words, our ci%ili.ation will remain human, indeed in
many ways more e$emplary o# what we regard as
human than it is today, although our understanding o#
the term will mo%e beyond its strictly biological
origins.
/any obser%ers ha%e nonetheless e$pressed alarm
at the emergence o# #orms o# nonbiological intelligence
superior to human intelligence. ,he potential to
augment our own intelligence through intimate
connection with other thin!ing mediums does not
necessarily alle%iate the concern, as some people ha%e
e$pressed the wish to remain 0unenhanced0 while at
the same time !eeping their place at the top o# the
intellectual #ood chain. /y %iew is that the li!ely
outcome is that on the one hand, #rom the perspecti%e
o# biological humanity, these superhuman intelligences
will appear to be their transcendent ser%ants, satis#ying
their needs and desires. )n the other hand, #ul#illing
the wishes o# a re%ered biological legacy will occupy
only a tri%ial portion o# the intellectual power that the
+ingularity will bring.
'eedless to say, the +ingularity will trans#orm all
aspects o# our li%es, social, se$ual, and economic,
which I e$plore herewith.
Wherefrom Moore's Law
-e#ore considering #urther the implications o# the
+ingularity, lets e$amine the wide range o#
technologies that are subject to the law o# accelerating
returns. ,he e$ponential trend that has gained the
greatest public recognition has become !nown as
0/oores ?aw.0 Gordon /oore, one o# the in%entors o#
integrated circuits, and then >hairman o# Intel, noted
in the mid :DH0s that we could s3uee.e twice as many
transistors on an integrated circuit e%ery ;4 months.
Gi%en that the electrons ha%e less distance to tra%el,
the circuits also run twice as #ast, pro%iding an o%erall
3uadrupling o# computational power.
*#ter si$ty years o# de%oted ser%ice, /oores ?aw
will die a digni#ied death no later than the year ;0:D.
-y that time, transistor #eatures will be just a #ew
atoms in width, and the strategy o# e%er #iner
photolithography will ha%e run its course. +o, will that
be the end o# the e$ponential growth o# computingE
8ont bet on it.
I# we plot the speed (in instructions per second& per
$:000 (in constant dollars& o# 4D #amous calculators
and computers spanning the entire twentieth century,
we note some interesting obser%ations.
Moore's Law Was Not the First, but the Fifth
Paradigm To Provide Exponential rowth of
!omputing
7ach time one paradigm runs out o# steam, another
pic!s up the pace
It is important to note that /oores ?aw o# Integrated
>ircuits was not the #irst, but the #i#th paradigm to
pro%ide accelerating price1per#ormance. >omputing
de%ices ha%e been consistently multiplying in power
(per unit o# time& #rom the mechanical calculating
de%ices used in the :CD0 ".+. >ensus, to ,urings
relay1based 06obinson0 machine that crac!ed the 'a.i
enigma code, to the >-+ %acuum tube computer that
predicted the election o# 7isenhower, to the transistor1
based machines used in the #irst space launches, to the
integrated1circuit1based personal computer which I
used to dictate (and automatically transcribe& this boo!
prcis.
-ut I noticed something else surprising. 5hen I
plotted the 4D machines on an e$ponential graph
(where a straight line means e$ponential growth&, I
didnt get a straight line. 5hat I got was another
e$ponential cur%e. In other words, theres e$ponential
growth in the rate o# e$ponential growth. >omputer
speed (per unit cost& doubled e%ery three years
between :D:0 and :D<0, doubled e%ery two years
between :D<0 and :DAA, and is now doubling e%ery
year.
-ut where does /oores ?aw come #romE 5hat is
behind this remar!ably predictable phenomenonE I
ha%e seen relati%ely little written about the ultimate
source o# this trend. Is it just 0a set o# industry
e$pectations and goals,0 as 6andy Isaac, head o# basic
science at I-/ contendsE )r is there something more
pro#ound going onE
In my %iew, it is one mani#estation (among many&
o# the e$ponential growth o# the e%olutionary process
that is technology. ,he e$ponential growth o#
computing is a mar%elous 3uantitati%e e$ample o# the
e$ponentially growing returns #rom an e%olutionary
process. 5e can also e$press the e$ponential growth o#
computing in terms o# an accelerating pace( it too!
ninety years to achie%e the #irst /I9+ (million
instructions per second& per thousand dollars, now we
add one /I9+ per thousand dollars e%ery day.
/oores ?aw narrowly re#ers to the number o#
transistors on an integrated circuit o# #i$ed si.e, and
sometimes has been e$pressed e%en more narrowly in
terms o# transistor #eature si.e. -ut rather than #eature
si.e (which is only one contributing #actor&, or e%en
number o# transistors, I thin! the most appropriate
measure to trac! is computational speed per unit cost.
,his ta!es into account many le%els o# 0cle%erness0
(i.e., inno%ation, which is to say, technological
e%olution&. In addition to all o# the inno%ation in
integrated circuits, there are multiple layers o#
inno%ation in computer design, e.g., pipelining, parallel
processing, instruction loo!1ahead, instruction and
memory caching, and many others.
From the abo%e chart, we see that the e$ponential
growth o# computing didnt start with integrated
circuits (around :D<C&, or e%en transistors (around
:D4H&, but goes bac! to the electromechanical
calculators used in the :CD0 and :D00 ".+. >ensus.
,his chart spans at least #i%e distinct paradigms o#
computing, o# which /oores ?aw pertains to only the
latest one.
Its ob%ious what the si$th paradigm will be a#ter
/oores ?aw runs out o# steam during the second
decade o# this century. >hips today are #lat (although it
does re3uire up to ;0 layers o# material to produce one
layer o# circuitry&. )ur brain, in contrast, is organi.ed
in three dimensions. 5e li%e in a three dimensional
world, why not use the third dimensionE ,he human
brain actually uses a %ery ine##icient electrochemical
digital controlled analog computational process. ,he
bul! o# the calculations are done in the interneuronal
connections at a speed o# only about ;00 calculations
per second (in each connection&, which is about ten
million times slower than contemporary electronic
circuits. -ut the brain gains its prodigious powers #rom
its e$tremely parallel organi.ation in three dimensions.
,here are many technologies in the wings that build
circuitry in three dimensions. 'anotubes, #or e$ample,
which are already wor!ing in laboratories, build
circuits #rom pentagonal arrays o# carbon atoms. )ne
cubic inch o# nanotube circuitry would be a million
times more power#ul than the human brain. ,here are
more than enough new computing technologies now
being researched, including three1dimensional silicon
chips, optical computing, crystalline computing, 8'*
computing, and 3uantum computing, to !eep the law o#
accelerating returns as applied to computation going
#or a long time.
,hus the (double& e$ponential growth o#
computing is broader than /oores ?aw, which re#ers
to only one o# its paradigms. *nd this accelerating
growth o# computing is, in turn, part o# the yet broader
phenomenon o# the accelerating pace o# any
e%olutionary process. )bser%ers are 3uic! to critici.e
e$trapolations o# an e$ponential trend on the basis that
the trend is bound to run out o# 0resources.0 ,he
classical e$ample is when a species happens upon a
new habitat (e.g., rabbits in *ustralia&, the species
numbers will grow e$ponentially #or a time, but then
hit a limit when resources such as #ood and space run
out.
-ut the resources underlying the e$ponential
growth o# an e%olutionary process are relati%ely
unbounded(
:. (i& ,he (e%er growing& order o# the e%olutionary
process itsel#. 7ach stage o# e%olution pro%ides
more power#ul tools #or the ne$t. In biological
e%olution, the ad%ent o# 8'* allowed more
power#ul and #aster e%olutionary
0e$periments.0 ?ater, setting the 0designs0 o#
animal body plans during the >ambrian
e$plosion allowed rapid e%olutionary
de%elopment o# other body organs such as the
brain. )r to ta!e a more recent e$ample, the
ad%ent o# computer assisted design tools allows
rapid de%elopment o# the ne$t generation o#
computers.
;. (ii& ,he 0chaos0 o# the en%ironment in which
the e%olutionary process ta!es place and which
pro%ides the options #or #urther di%ersity. In
biological e%olution, di%ersity enters the
process in the #orm o# mutations and e%er
changing en%ironmental conditions. In
technological e%olution, human ingenuity
combined with e%er changing mar!et
conditions !eep the process o# inno%ation
going.
,he ma$imum potential o# matter and energy to
contain intelligent processes is a %alid issue. -ut
according to my models, we wont approach those
limits during this century (but this will become an
issue within a couple o# centuries&.
5e also need to distinguish between the 0+0 cur%e
(an 0+0 stretched to the right, comprising %ery slow,
%irtually unnoticeable growth11#ollowed by %ery rapid
growth11#ollowed by a #lattening out as the process
approaches an asymptote& that is characteristic o# any
speci#ic technological paradigm and the continuing
e$ponential growth that is characteristic o# the ongoing
e%olutionary process o# technology. +peci#ic
paradigms, such as /oores ?aw, do ultimately reach
le%els at which e$ponential growth is no longer
#easible. ,hus /oores ?aw is an + cur%e. -ut the
growth o# computation is an ongoing e$ponential (at
least until we 0saturate0 the "ni%erse with the
intelligence o# our human1machine ci%ili.ation, but
that will not be a limit in this coming century&. In
accordance with the law o# accelerating returns,
paradigm shi#t, also called inno%ation, turns the +
cur%e o# any speci#ic paradigm into a continuing
e$ponential. * new paradigm (e.g., three1dimensional
circuits& ta!es o%er when the old paradigm approaches
its natural limit. ,his has already happened at least
#our times in the history o# computation. ,his
di##erence also distinguishes the tool ma!ing o# non1
human species, in which the mastery o# a tool1ma!ing
(or using& s!ill by each animal is characteri.ed by an
abruptly ending + shaped learning cur%e, %ersus
human1created technology, which has #ollowed an
e$ponential pattern o# growth and acceleration since its
inception.
"N# $e%uen&ing, Memor', !ommuni&ations, the
(nternet, and Miniaturi)ation
,his 0law o# accelerating returns0 applies to all o#
technology, indeed to any true e%olutionary process,
and can be measured with remar!able precision in
in#ormation based technologies. ,here are a great
many e$amples o# the e$ponential growth implied by
the law o# accelerating returns in technologies as
%aried as 8'* se3uencing, communication speeds,
electronics o# all !inds, and e%en in the rapidly
shrin!ing si.e o# technology. ,he +ingularity results
not #rom the e$ponential e$plosion o# computation
alone, but rather #rom the interplay and myriad
synergies that will result #rom mani#old intertwined
technological re%olutions. *lso, !eep in mind that
e%ery point on the e$ponential growth cur%es
underlying these panoply o# technologies (see the
graphs below& represents an intense human drama o#
inno%ation and competition. It is remar!able there#ore
that these chaotic processes result in such smooth and
predictable e$ponential trends.
For e$ample, when the human genome scan started
#ourteen years ago, critics pointed out that gi%en the
speed with which the genome could then be scanned, it
would ta!e thousands o# years to #inish the project. Yet
the #i#teen year project was nonetheless completed
slightly ahead o# schedule.
)# course, we e$pect to see e$ponential growth in
electronic memories such as 6*/.
Noti&e *ow Exponential rowth !ontinued
through Paradigm $hifts from +a&uum Tubes
to "is&rete Transistors to (ntegrated !ir&uits
=owe%er, growth in magnetic memory is not primarily
a matter o# /oores law, but includes ad%ances in
mechanical and electromagnetic systems.
7$ponential growth in communications technology has
been e%en more e$plosi%e than in computation and is
no less signi#icant in its implications. *gain, this
progression in%ol%es #ar more than just shrin!ing
transistors on an integrated circuit, but includes
accelerating ad%ances in #iber optics, optical switching,
electromagnetic technologies, and others.
'otice >ascade o# smaller 0+0 >ur%es
'ote that in the abo%e two charts we can actually see
the progression o# 0+0 cur%es( the acceleration #ostered
by a new paradigm, #ollowed by a le%eling o## as the
paradigm runs out o# steam, #ollowed by renewed
acceleration through paradigm shi#t.
,he #ollowing two charts show the o%erall growth o#
the Internet based on the number o# hosts. ,hese two
charts plot the same data, but one is on an e$ponential
a$is and the other is linear. *s I pointed out earlier,
whereas technology progresses in the e$ponential
domain, we e$perience it in the linear domain. +o #rom
the perspecti%e o# most obser%ers, nothing was
happening until the mid :DD0s when seemingly out o#
nowhere, the world wide web and email e$ploded into
%iew. -ut the emergence o# the Internet into a
worldwide phenomenon was readily predictable much
earlier by e$amining the e$ponential trend data.
'otice how the e$plosion o# the Internet appears to be
a surprise #rom the ?inear >hart, but was per#ectly
predictable #rom the 7$ponential >hart
"ltimately we will get away #rom the tangle o# wires
in our cities and in our li%es through wireless
communication, the power o# which is doubling e%ery
:0 to :: months.
*nother technology that will ha%e pro#ound
implications #or the twenty1#irst century is the
per%asi%e trend toward ma!ing things smaller, i.e.,
miniaturi.ation. ,he salient implementation si.es o# a
broad range o# technologies, both electronic and
mechanical, are shrin!ing, also at a double e$ponential
rate. *t present, we are shrin!ing technology by a
#actor o# appro$imately <.A per linear dimension per
decade.
The Exponential rowth of !omputation ,evisited
I# we %iew the e$ponential growth o# computation in
its proper perspecti%e as one e$ample o# the
per%asi%eness o# the e$ponential growth o# in#ormation
based technology, that is, as one e$ample o# many o#
the law o# accelerating returns, then we can con#idently
predict its continuation.
In the accompanying sidebar, I include a simpli#ied
mathematical model o# the law o# accelerating returns
as it pertains to the (double& e$ponential growth o#
computing. ,he #ormulas below result in the abo%e
graph o# the continued growth o# computation. ,his
graph matches the a%ailable data #or the twentieth
century through all #i%e paradigms and pro%ides
projections #or the twenty1#irst century. 'ote how the
Growth 6ate is growing slowly, but nonetheless
e$ponentially.
The Law of #&&elerating ,eturns #pplied to the
rowth of !omputation
,he #ollowing pro%ides a brie# o%er%iew o# the law o#
accelerating returns as it applies to the double
e$ponential growth o# computation. ,his model
considers the impact o# the growing power o# the
technology to #oster its own ne$t generation. For
e$ample, with more power#ul computers and related
technology, we ha%e the tools and the !nowledge to
design yet more power#ul computers, and to do so
more 3uic!ly.
'ote that the data #or the year ;000 and beyond
assume neural net connection calculations as it is
e$pected that this type o# calculation will ultimately
dominate, particularly in emulating human brain
#unctions. ,his type o# calculation is less e$pensi%e
than con%entional (e.g., 9entium III I IF& calculations
by a #actor o# at least :00 (particularly i# implemented
using digital controlled analog electronics, which
would correspond well to the brains digital controlled
analog electrochemical processes&. * #actor o# :00
translates into appro$imately A years (today& and less
than A years later in the twenty1#irst century.
/y estimate o# brain capacity is :00 billion neurons
times an a%erage :,000 connections per neuron (with
the calculations ta!ing place primarily in the
connections& times ;00 calculations per second.
*lthough these estimates are conser%ati%ely high, one
can #ind higher and lower estimates. =owe%er, e%en
much higher (or lower& estimates by orders o#
magnitude only shi#t the prediction by a relati%ely
small number o# years.
+ome prominent dates #rom this analysis include the
#ollowing(
5e achie%e one =uman -rain capability (; J
:0K:A cps& #or $:,000 around the year ;0;B.
5e achie%e one =uman -rain capability (; J
:0K:A cps& #or one cent around the year ;0BH.
5e achie%e one =uman 6ace capability (; J
:0K;A cps& #or $:,000 around the year ;04D.
5e achie%e one =uman 6ace capability (; J
:0K;A cps& #or one cent around the year ;0<D.
,he /odel considers the #ollowing %ariables(
F( Felocity (i.e., power& o# computing
(measured in >9+Iunit cost&
5( 5orld Lnowledge as it pertains to
designing and building computational de%ices
t( ,ime
,he assumptions o# the model are(
:. (:& F M >: J 5
In other words, computer power is a linear #unction o#
the !nowledge o# how to build computers. ,his is
actually a conser%ati%e assumption. In general,
inno%ations impro%e F (computer power& by a
multiple, not in an additi%e way. Independent
inno%ations multiply each others e##ect. For e$ample,
a circuit ad%ance such as >/)+, a more e##icient I>
wiring methodology, and a processor inno%ation such
as pipelining all increase F by independent multiples.
(;& 5 M >; J Integral (0 to t& F
In other words, 5 (!nowledge& is cumulati%e, and the
instantaneous increment to !nowledge is proportional
to F.
,his gi%es us(
5 M >: J >; J Integral (0 to t& 5
5 M >: J >; J >B K (>4 J t&
F M >: K ; J >; J >B K (>4 J t&
('ote on notation( aKb means a raised to the b
power.&
+impli#ying the constants, we get(
F M >a J >b K (>c J t&
+o this is a #ormula #or 0accelerating0 (i.e.,
e$ponentially growing& returns, a 0regular /oores
?aw.0
*s I mentioned abo%e, the data shows e$ponential
growth in the rate o# e$ponential growth. (5e doubled
computer power e%ery three years early in the
twentieth century, e%ery two years in the middle o# the
century, and close to e%ery one year during the :DD0s.&
?ets #actor in another e$ponential phenomenon, which
is the growing resources #or computation. 'ot only is
each (constant cost& de%ice getting more power#ul as a
#unction o# 5, but the resources deployed #or
computation are also growing e$ponentially.
5e now ha%e(
'( 7$penditures #or computation
F M >: J 5 (as be#ore&
' M >4 K (>< J t& (7$penditure #or computation
is growing at its own e$ponential rate&
5 M >; J Integral(0 to t& (' J F&
*s be#ore, world !nowledge is accumulating, and the
instantaneous increment is proportional to the amount
o# computation, which e3uals the resources deployed
#or computation ('& J the power o# each (constant
cost& de%ice.
,his gi%es us(
5 M >: J >; J Integral(0 to t& (>4 K (>< J t& J
5&
5 M >: J >; J (>B K (>A J t&& K (>H J t&
F M >: K ; J >; J (>B K (>A J t&& K (>H J t&
+impli#ying the constants, we get(
F M >a J (>b K (>c J t&& K (>d J t&
,his is a double e$ponential11an e$ponential cur%e in
which the rate o# e$ponential growth is growing at a
di##erent e$ponential rate.
'ow lets consider real1world data. >onsidering the
data #or actual calculating de%ices and computers
during the twentieth century(
>9+I$:L( >alculations 9er +econd #or $:,000
,wentieth century computing data matches(
>9+I$:L M :0K(A.00J((;0.40IA.00&K((*:B1
:D00&I:00&&1::.00&
5e can determine the growth rate o%er a period o#
time(
Growth 6ate M:0K((?)G(>9+I$:L #or >urrent
Year& 1 ?)G(>9+I$:L #or 9re%ious Year&&I
(>urrent Year 1 9re%ious Year&&
=uman -rain M :00 -illion (:0K::& neurons J
:000 (:0KB& >onnectionsI'euron J ;00 (; J
:0K;& >alculations 9er +econd 9er >onnection
M ; J :0K:A >alculations 9er +econd
=uman 6ace M :0 -illion (:0K:0& =uman
-rains M ; J :0K;A >alculations 9er +econd
,hese #ormulas produce the graph abo%e.
*lready, I-/s 0-lue Gene0 supercomputer, now
being built and scheduled to be completed by ;00<, is
projected to pro%ide : million billion calculations per
second (i.e., one billion mega#lops&. ,his is already
one twentieth o# the capacity o# the human brain,
which I estimate at a conser%ati%ely high ;0 million
billion calculations per second (:00 billion neurons
times :,000 connections per neuron times ;00
calculations per second per connection&. In line with
my earlier predictions, supercomputers will achie%e
one human brain capacity by ;0:0, and personal
computers will do so by around ;0;0. -y ;0B0, it will
ta!e a %illage o# human brains (around a thousand& to
match $:000 o# computing. -y ;0<0, $:000 o#
computing will e3ual the processing power o# all
human brains on 7arth. )# course, this only includes
those brains still using carbon1based neurons. 5hile
human neurons are wondrous creations in a way, we
wouldnt (and dont& design computing circuits the
same way. )ur electronic circuits are already more
than ten million times #aster than a neurons
electrochemical processes. /ost o# the comple$ity o# a
human neuron is de%oted to maintaining its li#e support
#unctions, not its in#ormation processing capabilities.
"ltimately, we will need to port our mental processes
to a more suitable computational substrate. ,hen our
minds wont ha%e to stay so small, being constrained as
they are today to a mere hundred trillion neural
connections each operating at a ponderous ;00
digitally controlled analog calculations per second.
The $oftware of (ntelligen&e
+o #ar, I%e been tal!ing about the hardware o#
computing. ,he so#tware is e%en more salient. )ne o#
the principal assumptions underlying the e$pectation
o# the +ingularity is the ability o# nonbiological
mediums to emulate the richness, subtlety, and depth
o# human thin!ing. *chie%ing the computational
capacity o# the human brain, or e%en %illages and
nations o# human brains will not automatically produce
human le%els o# capability. -y human le%els I include
all the di%erse and subtle ways in which humans are
intelligent, including musical and artistic aptitude,
creati%ity, physically mo%ing through the world, and
understanding and responding appropriately to
emotion. ,he re3uisite hardware capacity is a
necessary but not su##icient condition. ,he
organi.ation and content o# these resources11the
so#tware o# intelligence11is also critical.
-e#ore addressing this issue, it is important to note that
once a computer achie%es a human le%el o#
intelligence, it will necessarily soar past it. * !ey
ad%antage o# nonbiological intelligence is that
machines can easily share their !nowledge. I# I learn
French, or read 5ar and 9eace, I cant readily
download that learning to you. You ha%e to ac3uire
that scholarship the same painsta!ing way that I did.
/y !nowledge, embedded in a %ast pattern o#
neurotransmitter concentrations and interneuronal
connections, cannot be 3uic!ly accessed or
transmitted. -ut we wont lea%e out 3uic! downloading
ports in our nonbiological e3ui%alents o# human
neuron clusters. 5hen one computer learns a s!ill or
gains an insight, it can immediately share that wisdom
with billions o# other machines.
*s a contemporary e$ample, we spent years teaching
one research computer how to recogni.e continuous
human speech. 5e e$posed it to thousands o# hours o#
recorded speech, corrected its errors, and patiently
impro%ed its per#ormance. Finally, it became 3uite
adept at recogni.ing speech (I dictated most o# my
recent boo! to it&. 'ow i# you want your own personal
computer to recogni.e speech, it doesnt ha%e to go
through the same process2 you can just download the
#ully trained patterns in seconds. "ltimately, billions o#
nonbiological entities can be the master o# all human
and machine ac3uired !nowledge.
In addition, computers are potentially millions o# times
#aster than human neural circuits. * computer can also
remember billions or e%en trillions o# #acts per#ectly,
while we are hard pressed to remember a hand#ul o#
phone numbers. ,he combination o# human le%el
intelligence in a machine with a computers inherent
superiority in the speed, accuracy, and sharing ability
o# its memory will be #ormidable.
,here are a number o# compelling scenarios to achie%e
higher le%els o# intelligence in our computers, and
ultimately human le%els and beyond. 5e will be able
to e%ol%e and train a system combining massi%ely
parallel neural nets with other paradigms to understand
language and model !nowledge, including the ability
to read and model the !nowledge contained in written
documents. "nli!e many contemporary 0neural net0
machines, which use mathematically simpli#ied models
o# human neurons, some contemporary neural nets are
already using highly detailed models o# human
neurons, including detailed nonlinear analog acti%ation
#unctions and other rele%ant details. *lthough the
ability o# todays computers to e$tract and learn
!nowledge #rom natural language documents is
limited, their capabilities in this domain are impro%ing
rapidly. >omputers will be able to read on their own,
understanding and modeling what they ha%e read, by
the second decade o# the twenty1#irst century. 5e can
then ha%e our computers read all o# the worlds
literature11boo!s, maga.ines, scienti#ic journals, and
other a%ailable material. "ltimately, the machines will
gather !nowledge on their own by %enturing out on the
web, or e%en into the physical world, drawing #rom the
#ull spectrum o# media and in#ormation ser%ices, and
sharing !nowledge with each other (which machines
can do #ar more easily than their human creators&.
,everse Engineering the *uman -rain
,he most compelling scenario #or mastering the
so#tware o# intelligence is to tap into the blueprint o#
the best e$ample we can get our hands on o# an
intelligent process. ,here is no reason why we cannot
re%erse engineer the human brain, and essentially copy
its design. *lthough it too! its original designer se%eral
billion years to de%elop, its readily a%ailable to us, and
not (yet& copyrighted. *lthough theres a s!ull around
the brain, it is not hidden #rom our %iew.
,he most immediately accessible way to accomplish
this is through destructi%e scanning( we ta!e a #ro.en
brain, pre#erably one #ro.en just slightly be#ore rather
than slightly a#ter it was going to die anyway, and
e$amine one brain layer11one %ery thin slice11at a time.
5e can readily see e%ery neuron and e%ery connection
and e%ery neurotransmitter concentration represented
in each synapse1thin layer.
=uman brain scanning has already started. *
condemned !iller allowed his brain and body to be
scanned and you can access all :0 billion bytes o# him
on the Internet
http(IIwww.nlm.nih.go%IresearchI%isibleI%isibleNhuma
n.html.
=e has a ;< billion byte #emale companion on the site
as well in case he gets lonely. ,his scan is not high
enough in resolution #or our purposes, but then, we
probably dont want to base our templates o# machine
intelligence on the brain o# a con%icted !iller, anyway.
-ut scanning a #ro.en brain is #easible today, albeit not
yet at a su##icient speed or bandwidth, but again, the
law o# accelerating returns will pro%ide the re3uisite
speed o# scanning, just as it did #or the human genome
scan. >arnegie /ellon "ni%ersitys *ndreas 'owat.y!
plans to scan the ner%ous system o# the brain and body
o# a mouse with a resolution o# less than ;00
nanometers, which is getting %ery close to the
resolution needed #or re%erse engineering.
5e also ha%e nonin%asi%e scanning techni3ues today,
including high1resolution magnetic resonance imaging
(/6I& scans, optical imaging, near1in#rared scanning,
and other technologies which are capable in certain
instances o# resol%ing indi%idual somas, or neuron cell
bodies. -rain scanning technologies are also increasing
their resolution with each new generation, just what we
would e$pect #rom the law o# accelerating returns.
Future generations will enable us to resol%e the
connections between neurons and to peer inside the
synapses and record the neurotransmitter
concentrations.
5e can peer inside someones brain today with
nonin%asi%e scanners, which are increasing their
resolution with each new generation o# this
technology. ,here are a number o# technical challenges
in accomplishing this, including achie%ing suitable
resolution, bandwidth, lac! o# %ibration, and sa#ety.
For a %ariety o# reasons it is easier to scan the brain o#
someone recently deceased than o# someone still
li%ing. It is easier to get someone deceased to sit still,
#or one thing. -ut nonin%asi%ely scanning a li%ing
brain will ultimately become #easible as /6I, optical,
and other scanning technologies continue to impro%e in
resolution and speed.
$&anning from (nside
*lthough nonin%asi%e means o# scanning the brain
#rom outside the s!ull are rapidly impro%ing, the most
practical approach to capturing e%ery salient neural
detail will be to scan it #rom inside. -y ;0B0,
0nanobot0 (i.e., nano robot& technology will be %iable,
and brain scanning will be a prominent application.
'anobots are robots that are the si.e o# human blood
cells, or e%en smaller. -illions o# them could tra%el
through e%ery brain capillary and scan e%ery rele%ant
#eature #rom up close. "sing high speed wireless
communication, the nanobots would communicate
with each other, and with other computers that are
compiling the brain scan data base (in other words, the
nanobots will all be on a wireless local area networ!&.
,his scenario in%ol%es only capabilities that we can
touch and #eel today. 5e already ha%e technology
capable o# producing %ery high resolution scans,
pro%ided that the scanner is physically pro$imate to the
neural #eatures. ,he basic computational and
communication methods are also essentially #easible
today. ,he primary #eatures that are not yet practical
are nanobot si.e and cost. *s I discussed abo%e, we
can project the e$ponentially declining cost o#
computation, and the rapidly declining si.e o# both
electronic and mechanical technologies. 5e can
conser%ati%ely e$pect, there#ore, the re3uisite nanobot
technology by around ;0B0. -ecause o# its ability to
place each scanner in %ery close physical pro$imity to
e%ery neural #eature, nanobot1based scanning will be
more practical than scanning the brain #rom outside.
*ow to .se /our -rain $&an
=ow will we apply the thousands o# trillions o# bytes
o# in#ormation deri%ed #rom each brain scanE )ne
approach is to use the results to design more intelligent
parallel algorithms #or our machines, particularly those
based on one o# the neural net paradigms. 5ith this
approach, we dont ha%e to copy e%ery single
connection. ,here is a great deal o# repetition and
redundancy within any particular brain region.
*lthough the in#ormation contained in a human brain
would re3uire thousands o# trillions o# bytes o#
in#ormation (on the order o# :00 billion neurons times
an a%erage o# :,000 connections per neuron, each with
multiple neurotransmitter concentrations and
connection data&, the design o# the brain is
characteri.ed by a human genome o# only about a
billion bytes.
Furthermore, most o# the genome is redundant, so the
initial design o# the brain is characteri.ed by
appro$imately one hundred million bytes, about the
si.e o# /icroso#t 5ord. )# course, the comple$ity o#
our brains greatly increases as we interact with the
world (by a #actor o# more than ten million&. -ecause
o# the highly repetiti%e patterns #ound in each speci#ic
brain region, it is not necessary to capture each detail
in order to re%erse engineer the signi#icant digital1
analog algorithms. 5ith this in#ormation, we can
design simulated nets that operate similarly. ,here are
already multiple e##orts under way to scan the human
brain and apply the insights deri%ed to the design o#
intelligent machines.
,he pace o# brain re%erse engineering is only slightly
behind the a%ailability o# the brain scanning and
neuron structure in#ormation. * contemporary e$ample
is a comprehensi%e model o# a signi#icant portion o#
the human auditory processing system that ?loyd
5atts (www.lloydwatts.com& has de%eloped #rom both
neurobiology studies o# speci#ic neuron types and brain
interneuronal connection in#ormation. 5atts model
includes #i%e parallel paths and includes the actual
intermediate representations o# auditory in#ormation at
each stage o# neural processing. 5atts has
implemented his model as real1time so#tware which
can locate and identi#y sounds with many o# the same
properties as human hearing. *lthough a wor! in
progress, the model illustrates the #easibility o#
con%erting neurobiological models and brain
connection data into wor!ing simulations. *lso, as
=ans /ora%ec and others ha%e speculated, these
e##icient simulations re3uire about :,000 times less
computation than the theoretical potential o# the
biological neurons being simulated.
,everse Engineering the *uman -rain0 Five
Parallel #uditor' Pathwa's
Chart by ?loyd 5atts
!o&hlea( +ense organ o# hearing. B0,000 #ibers
con%erts motion o# the stapes into spectro1temporal
representation o# sound.
M!( /ultipolar >ells. /easure spectral energy.
-!( Globular -ushy >ells. 6elays spi!es #rom the
auditory ner%e to the ?ateral +uperior.
1livar' !omplex (includes L$1 and M$12.
7ncoding o# timing and amplitude o# signals #or
binaural comparison o# le%el.
$-!( +pherical -ushy >ells. 9ro%ide temporal
sharpening o# time o# arri%al, as a pre1processor #or
interaural time di##erence calculation.
1!( )ctopus >ells. 8etection o# transients.
"!N( 8orsal >ochlear 'ucleus. 8etection o# spectral
edges and calibrating #or noise le%els.
+NT-( Fentral 'ucleus o# the ,rape.oid -ody.
Feedbac! signals to modulate outer hair cell #unction
in the cochlea.
+NLL, P1N( Fentral 'ucleus o# the ?ateral
?emniscus, 9eri1)li%ary 'uclei. 9rocessing transients
#rom the )ctopus >ells.
M$1( /edial +uperior )li%e. >omputing inter1aural
time di##erence (di##erence in time o# arri%al between
the two ears, used to tell where a sound is coming
#rom&.
L$1( ?ateral +uperior )li%e. *lso in%ol%ed in
computing inter1aural le%el di##erence.
(!!( >entral 'ucleus o# the In#erior >olliculus. ,he
site o# major integration o# multiple representations o#
sound.
(!x( 7$terior 'ucleus o# the In#erior >olliculus.
Further re#inement o# sound locali.ation.
$!( +uperior >olliculus. ?ocation o# auditoryI%isual
merging.
M-( /edial Geniculate -ody. ,he auditory portion
o# the thalamus.
L$( ?imbic +ystem. >omprising many structures
associated with emotion, memory, territory, etc.
#!( *uditory >orte$.
,he brain is not one huge 0tabula rasa0 (i.e.,
undi##erentiated blan! slate&, but rather an intricate and
intertwined collection o# hundreds o# speciali.ed
regions. ,he process o# 0peeling the onion0 to
understand these interlea%ed regions is well underway.
*s the re3uisite neuron models and brain
interconnection data becomes a%ailable, detailed and
implementable models such as the auditory e$ample
abo%e will be de%eloped #or all brain regions.
*#ter the algorithms o# a region are understood, they
can be re#ined and e$tended be#ore being implemented
in synthetic neural e3ui%alents. For one thing, they can
be run on a computational substrate that is already
more than ten million times #aster than neural circuitry.
*nd we can also throw in the methods #or building
intelligent machines that we already understand.
"ownloading the *uman -rain
* more contro%ersial application than this scanning1
the1brain1to1understand1it scenario is scanning-the-
brain-to-download-it. =ere we scan someones brain to
map the locations, interconnections, and contents o# all
the somas, a$ons, dendrites, presynaptic %esicles,
neurotransmitter concentrations, and other neural
components and le%els. Its entire organi.ation can then
be re1created on a neural computer o# su##icient
capacity, including the contents o# its memory.
,o do this, we need to understand local brain
processes, although not necessarily all o# the higher
le%el processes. +canning a brain with su##icient detail
to download it may sound daunting, but so did the
human genome scan. *ll o# the basic technologies
e$ist today, just not with the re3uisite speed, cost, and
si.e, but these are the attributes that are impro%ing at a
double e$ponential pace.
,he computationally pertinent aspects o# indi%idual
neurons are complicated, but de#initely not beyond our
ability to accurately model. For e$ample, ,ed -erger
and his colleagues at =edco 'eurosciences ha%e built
integrated circuits that precisely match the digital and
analog in#ormation processing characteristics o#
neurons, including clusters with hundreds o# neurons.
>ar%er /ead and his colleagues at >al,ech ha%e built
a %ariety o# integrated circuits that emulate the digital1
analog characteristics o# mammalian neural circuits.
* recent e$periment at +an 8iegos Institute #or
'onlinear +cience demonstrates the potential #or
electronic neurons to precisely emulate biological
ones. 'eurons (biological or otherwise& are a prime
e$ample o# what is o#ten called 0chaotic computing.0
7ach neuron acts in an essentially unpredictable
#ashion. 5hen an entire networ! o# neurons recei%es
input (#rom the outside world or #rom other networ!s
o# neurons&, the signaling amongst them appears at
#irst to be #ren.ied and random. )%er time, typically a
#raction o# a second or so, the chaotic interplay o# the
neurons dies down, and a stable pattern emerges. ,his
pattern represents the 0decision0 o# the neural networ!.
I# the neural networ! is per#orming a pattern
recognition tas! (which, incidentally, comprises the
bul! o# the acti%ity in the human brain&, then the
emergent pattern represents the appropriate
recognition.
+o the 3uestion addressed by the +an 8iego
researchers was whether electronic neurons could
engage in this chaotic dance alongside biological ones.
,hey hoo!ed up their arti#icial neurons with those
#rom spiney lobsters in a single networ!, and their
hybrid biological1nonbiological networ! per#ormed in
the same way (i.e., chaotic interplay #ollowed by a
stable emergent pattern& and with the same type o#
results as an all biological net o# neurons. 7ssentially,
the biological neurons accepted their electronic peers.
It indicates that their mathematical model o# these
neurons was reasonably accurate.
,here are many projects around the world which are
creating nonbiological de%ices to recreate in great
detail the #unctionality o# human neuron clusters. ,he
accuracy and scale o# these neuron1cluster replications
are rapidly increasing. 5e started with #unctionally
e3ui%alent recreations o# single neurons, then clusters
o# tens, then hundreds, and now thousands. +caling up
technical processes at an e$ponential pace is what
technology is good at.
*s the computational power to emulate the human
brain becomes a%ailable11were not there yet, but we
will be there within a couple o# decades11projects
already under way to scan the human brain will be
accelerated, with a %iew both to understand the human
brain in general, as well as pro%iding a detailed
description o# the contents and design o# speci#ic
brains. -y the third decade o# the twenty1#irst century,
we will be in a position to create highly detailed and
complete maps o# all rele%ant #eatures o# all neurons,
neural connections and synapses in the human brain,
all o# the neural details that play a role in the beha%ior
and #unctionality o# the brain, and to recreate these
designs in suitably ad%anced neural computers.
(s the *uman -rain "ifferent from a !omputer3
Is the human brain di##erent #rom a computerE
,he answer depends on what we mean by the word
0computer.0 >ertainly the brain uses %ery di##erent
methods #rom con%entional contemporary computers.
/ost computers today are all digital and per#orm one
(or perhaps a #ew& computations at a time at e$tremely
high speed. In contrast, the human brain combines
digital and analog methods with most computations
per#ormed in the analog domain. ,he brain is
massi%ely parallel, per#orming on the order o# a
hundred trillion computations at the same time, but at
e$tremely slow speeds.
5ith regard to digital %ersus analog computing, we
!now that digital computing can be #unctionally
e3ui%alent to analog computing (although the re%erse
is not true&, so we can per#orm all o# the capabilities o#
a hybrid digital11analog networ! with an all digital
computer. )n the other hand, there is an engineering
ad%antage to analog circuits in that analog computing
is potentially thousands o# times more e##icient. *n
analog computation can be per#ormed by a #ew
transistors, or, in the case o# mammalian neurons,
speci#ic electrochemical processes. * digital
computation, in contrast, re3uires thousands or tens o#
thousands o# transistors. +o there is a signi#icant
engineering ad%antage to emulating the brains analog
methods.
,he massi%e parallelism o# the human brain is the !ey
to its pattern recognition abilities, which re#lects the
strength o# human thin!ing. *s I discussed abo%e,
mammalian neurons engage in a chaotic dance, and i#
the neural networ! has learned its lessons well, then a
stable pattern will emerge re#lecting the networ!s
decision. ,here is no reason why our nonbiological
#unctionally e3ui%alent recreations o# biological neural
networ!s cannot be built using these same principles,
and indeed there are do.ens o# projects around the
world that ha%e succeeded in doing this. /y own
technical #ield is pattern recognition, and the projects
that I ha%e been in%ol%ed in #or o%er thirty years use
this #orm o# chaotic computing. 9articularly success#ul
e$amples are >ar%er /eads neural chips, which are
highly parallel, use digital controlled analog
computing, and are intended as #unctionally similar
recreations o# biological networ!s.
1b4e&tive and $ub4e&tive
,he +ingularity en%isions the emergence o# human1
li!e intelligent entities o# astonishing di%ersity and
scope. *lthough these entities will be capable o#
passing the 0,uring test0 (i.e., able to #ool humans that
they are human&, the 3uestion arises as to whether
these 0people0 are conscious, or just appear that way.
,o gain some insight as to why this is an e$tremely
subtle 3uestion (albeit an ultimately important one& it
is use#ul to consider some o# the parado$es that emerge
#rom the concept o# downloading speci#ic human
brains.
*lthough I anticipate that the most common
application o# the !nowledge gained #rom re%erse
engineering the human brain will be creating more
intelligent machines that are not necessarily modeled
on speci#ic biological human indi%iduals, the scenario
o# scanning and reinstantiating all o# the neural details
o# a specific person raises the most immediate
3uestions o# identity. ?ets consider the 3uestion o#
what we will #ind when we do this.
5e ha%e to consider this 3uestion on both the objecti%e
and subjecti%e le%els. 0)bjecti%e0 means e%eryone
e$cept me, so lets start with that. )bjecti%ely, when
we scan someones brain and reinstantiate their
personal mind #ile into a suitable computing medium,
the newly emergent 0person0 will appear to other
obser%ers to ha%e %ery much the same personality,
history, and memory as the person originally scanned.
,hat is, once the technology has been re#ined and
per#ected. ?i!e any new technology, it wont be per#ect
at #irst. -ut ultimately, the scans and recreations will
be %ery accurate and realistic.
Interacting with the newly instantiated person will #eel
li!e interacting with the original person. ,he new
person will claim to be that same old person and will
ha%e a memory o# ha%ing been that person. ,he new
person will ha%e all o# the patterns o# !nowledge, s!ill,
and personality o# the original. 5e are already creating
#unctionally e3ui%alent recreations o# neurons and
neuron clusters with su##icient accuracy that biological
neurons accept their nonbiological e3ui%alents and
wor! with them as i# they were biological. ,here are
no natural limits that pre%ent us #rom doing the same
with the hundred billion neuron cluster o# clusters we
call the human brain.
+ubjecti%ely, the issue is more subtle and pro#ound,
but #irst we need to re#lect on one additional objecti%e
issue( our physical sel#.
The (mportan&e of *aving a -od'
>onsider how many o# our thoughts and thin!ing are
directed toward our body and its sur%i%al, security,
nutrition, and image, not to mention a##ection,
se$uality, and reproduction. /any, i# not most, o# the
goals we attempt to ad%ance using our brains ha%e to
do with our bodies( protecting them, pro%iding them
with #uel, ma!ing them attracti%e, ma!ing them #eel
good, pro%iding #or their myriad needs and desires.
+ome philosophers maintain that achie%ing human
le%el intelligence is impossible without a body. I# were
going to port a humans mind to a new computational
medium, wed better pro%ide a body. * disembodied
mind will 3uic!ly get depressed.
,here are a %ariety o# bodies that we will pro%ide #or
our machines, and that they will pro%ide #or
themsel%es( bodies built through nanotechnology (i.e.,
building highly comple$ physical systems atom by
atom&, %irtual bodies (that e$ist only in %irtual reality&,
bodies comprised o# swarms o# nanobots, and other
technologies.
* common scenario will be to enhance a persons
biological brain with intimate connection to
nonbiological intelligence. In this case, the body
remains the good old human body that were #amiliar
with, although this too will become greatly enhanced
through biotechnology (gene enhancement and
replacement& and, later on, through nanotechnology. *
detailed e$amination o# twenty1#irst century bodies is
beyond the scope o# this prcis, but recreating and
enhancing our bodies will be (and has been& an easier
tas! than recreating our minds.
$o 5ust Who #re These People3
,o return to the issue o# subjecti%ity, consider( is the
reinstantiated mind the same consciousness as the
person we just scannedE *re these 0people0 conscious
at allE Is this a mind or just a brainE
>onsciousness in our twenty1#irst century machines
will be a critically important issue. -ut it is not easily
resol%ed, or e%en readily understood. 9eople tend to
ha%e strong %iews on the subject, and o#ten just cant
understand how anyone else could possibly see the
issue #rom a di##erent perspecti%e. /ar%in /ins!y
obser%ed that 0theres something 3ueer about
describing consciousness. 5hate%er people mean to
say, they just cant seem to ma!e it clear.0
5e dont worry, at least not yet, about causing pain
and su##ering to our computer programs. -ut at what
point do we consider an entity, a process, to be
conscious, to #eel pain and discom#ort, to ha%e its own
intentionality, its own #ree willE =ow do we determine
i# an entity is conscious2 i# it has subjecti%e
e$perienceE =ow do we distinguish a process that is
conscious #rom one that just acts as if it is consciousE
5e cant simply as! it. I# it says 0=ey Im conscious,0
does that settle the issueE 'o, we ha%e computer
games today that e##ecti%ely do that, and theyre not
terribly con%incing.
=ow about i# the entity is %ery con%incing and
compelling when it says 0Im lonely, please !eep me
company.0 8oes that settle the issueE
I# we loo! inside its circuits, and see essentially the
identical !inds o# #eedbac! loops and other
mechanisms in its brain that we see in a human brain
(albeit implemented using nonbiological e3ui%alents&,
does that settle the issueE
*nd just who are these people in the machine,
anywayE ,he answer will depend on who you as!. I#
you as! the people in the machine, they will
strenuously claim to be the original persons. For
e$ample, i# we scan11lets say mysel#11and record the
e$act state, le%el, and position o# e%ery
neurotransmitter, synapse, neural connection, and
e%ery other rele%ant detail, and then reinstantiate this
massi%e data base o# in#ormation (which I estimate at
thousands o# trillions o# bytes& into a neural computer
o# su##icient capacity, the person who then emerges in
the machine will thin! that 0he0 is (and had been& me,
or at least he will act that way. =e will say 0I grew up
in Oueens, 'ew Yor!, went to college at /I,, stayed
in the -oston area, started and sold a #ew arti#icial
intelligence companies, wal!ed into a scanner there,
and wo!e up in the machine here. =ey, this technology
really wor!s.0
-ut wait.
Is this really meE For one thing, old biological 6ay
(thats me& still e$ists. Ill still be here in my carbon1
cell1based brain. *las, I will ha%e to sit bac! and watch
the new 6ay succeed in endea%ors that I could only
dream o#.
# Thought Experiment
?ets consider the issue o# just who I am, and who the
new 6ay is a little more care#ully. First o# all, am I the
stu## in my brain and bodyE
>onsider that the particles ma!ing up my body and
brain are constantly changing. 5e are not at all
permanent collections o# particles. ,he cells in our
bodies turn o%er at di##erent rates, but the particles
(e.g., atoms and molecules& that comprise our cells are
e$changed at a %ery rapid rate. I am just not the same
collection o# particles that I was e%en a month ago. It is
the patterns o# matter and energy that are
semipermanent (that is, changing only gradually&, but
our actual material content is changing constantly, and
%ery 3uic!ly. 5e are rather li!e the patterns that water
ma!es in a stream. ,he rushing water around a
#ormation o# roc!s ma!es a particular, uni3ue pattern.
,his pattern may remain relati%ely unchanged #or
hours, e%en years. )# course, the actual material
constituting the pattern11the water11is replaced in
milliseconds. ,he same is true #or 6ay Lur.weil. ?i!e
the water in a stream, my particles are constantly
changing, but the pattern that people recogni.e as 6ay
has a reasonable le%el o# continuity. ,his argues that
we should not associate our #undamental identity with
a speci#ic set o# particles, but rather the pattern o#
matter and energy that we represent. /any
contemporary philosophers seem partial to this
0identi#y #rom pattern0 argument.
-ut (again& wait.
I# you were to scan my brain and reinstantiate new 6ay
while I was sleeping, I would not necessarily e%en
!now about it (with the nanobots, this will be a #easible
scenario&. I# you then come to me, and say, 0good
news, 6ay, we%e success#ully reinstantiated your mind
#ile, so we wont be needing your old brain anymore,0 I
may suddenly reali.e the #law in this 0identity #rom
pattern0 argument. I may wish new 6ay well, and
reali.e that he shares my 0pattern,0 but I would
nonetheless conclude that hes not me, because Im still
here. =ow could he be meE *#ter all, I would not
necessarily !now that he e%en e$isted.
?ets consider another perple$ing scenario. +uppose I
replace a small number o# biological neurons with
#unctionally e3ui%alent nonbiological ones (they may
pro%ide certain bene#its such as greater reliability and
longe%ity, but thats not rele%ant to this thought
e$periment&. *#ter I ha%e this procedure per#ormed,
am I still the same personE /y #riends certainly thin!
so. I still ha%e the same sel#1deprecating humor, the
same silly grin11yes, Im still the same guy.
It should be clear where Im going with this. -it by bit,
region by region, I ultimately replace my entire brain
with essentially identical (perhaps impro%ed&
nonbiological e3ui%alents (preser%ing all o# the
neurotransmitter concentrations and other details that
represent my learning, s!ills, and memories&. *t each
point, I #eel the procedures were success#ul. *t each
point, I #eel that I am the same guy. *#ter each
procedure, I claim to be the same guy. /y #riends
concur. ,here is no old 6ay and new 6ay, just one
6ay, one that ne%er appears to #undamentally change.
-ut consider this. ,his gradual replacement o# my
brain with a nonbiological e3ui%alent is essentially
identical to the #ollowing se3uence(
:. (i& scan 6ay and reinstantiate 6ays mind #ile
into new (nonbiological& 6ay, and, then
;. (ii& terminate old 6ay. -ut we concluded abo%e
that in such a scenario new 6ay is not the same
as old 6ay. *nd i# old 6ay is terminated, well
then thats the end o# 6ay. +o the gradual
replacement scenario essentially ends with the
same result( 'ew 6ay has been created, and old
6ay has been destroyed, e%en i# we ne%er saw
him missing. +o what appears to be the
continuing e$istence o# just one 6ay is really
the creation o# new 6ay and the termination o#
old 6ay.
On yet another hand (were running out o#
philosophical hands here&, the gradual replacement
scenario is not altogether di##erent #rom what happens
normally to our biological sel%es, in that our particles
are always rapidly being replaced. +o am I constantly
being replaced with someone else who just happens to
be %ery similar to my old sel#E
I am trying to illustrate why consciousness is not an
easy issue. I# we tal! about consciousness as just a
certain type o# intelligent s!ill( the ability to re#lect on
ones own sel# and situation, #or e$ample, then the
issue is not di##icult at all because any s!ill or
capability or #orm o# intelligence that one cares to
de#ine will be replicated in nonbiological entities (i.e.,
machines& within a #ew decades. 5ith this type o#
objective %iew o# consciousness, the conundrums do
go away. -ut a #ully objecti%e %iew does not penetrate
to the core o# the issue, because the essence o#
consciousness is subjective e$perience, not objecti%e
correlates o# that e$perience.
5ill these #uture machines be capable o# ha%ing
spiritual e$periencesE
,hey certainly will claim to. ,hey will claim to be
people, and to ha%e the #ull range o# emotional and
spiritual e$periences that people claim to ha%e. *nd
these will not be idle claims2 they will e%idence the
sort o# rich, comple$, and subtle beha%ior one
associates with these #eelings. =ow do the claims and
beha%iors11compelling as they will be11relate to the
subjecti%e e$perience o# these reinstantiated peopleE
5e !eep coming bac! to the %ery real but ultimately
unmeasurable issue o# consciousness.
9eople o#ten tal! about consciousness as i# it were a
clear property o# an entity that can readily be
identi#ied, detected, and gauged. I# there is one crucial
insight that we can ma!e regarding why the issue o#
consciousness is so contentious, it is the #ollowing(
,here e$ists no objecti%e test that can conclusi%ely
determine its presence.
+cience is about objecti%e measurement and logical
implications there#rom, but the %ery nature o#
objecti%ity is that you cannot measure subjecti%e
e$perience1you can only measure correlates o# it, such
as beha%ior (and by beha%ior, I include the actions o#
components o# an entity, such as neurons&. ,his
limitation has to do with the %ery nature o# the
concepts 0objecti%e0 and 0subjecti%e.0 Fundamentally,
we cannot penetrate the subjecti%e e$perience o#
another entity with direct objecti%e measurement. 5e
can certainly ma!e arguments about it( i.e., 0loo!
inside the brain o# this nonhuman entity, see how its
methods are just li!e a human brain.0 )r, 0see how its
beha%ior is just li!e human beha%ior.0 -ut in the end,
these remain just arguments. 'o matter how
con%incing the beha%ior o# a reinstantiated person,
some obser%ers will re#use to accept the consciousness
o# an entity unless it s3uirts neurotransmitters, or is
based on 8'*1guided protein synthesis, or has some
other speci#ic biologically human attribute.
5e assume that other humans are conscious, but that is
still an assumption, and there is no consensus amongst
humans about the consciousness o# nonhuman entities,
such as higher non1human animals. ,he issue will be
e%en more contentious with regard to #uture
nonbiological entities with human1li!e beha%ior and
intelligence.
+o how will we resol%e the claimed consciousness o#
nonbiological intelligence (claimed, that is, by the
machines&E From a practical perspecti%e, well accept
their claims. Leep in mind that nonbiological entities
in the twenty1#irst century will be e$tremely
intelligent, so theyll be able to con%ince us that they
are conscious. ,heyll ha%e all the delicate and
emotional cues that con%ince us today that humans are
conscious. ,hey will be able to ma!e us laugh and cry.
*nd theyll get mad i# we dont accept their claims. -ut
#undamentally this is a political prediction, not a
philosophical argument.
1n Tubules and 6uantum !omputing
)%er the past se%eral years, 6oger 9enrose, a noted
physicist and philosopher, has suggested that #ine
structures in the neurons called tubules per#orm an
e$otic #orm o# computation called 03uantum
computing.0 Ouantum computing is computing using
what are called 03u bits0 which ta!e on all possible
combinations o# solutions simultaneously. It can be
considered to be an e$treme #orm o# parallel
processing (because e%ery combination o# %alues o# the
3u bits are tested simultaneously&. 9enrose suggests
that the tubules and their 3uantum computing
capabilities complicate the concept o# recreating
neurons and reinstantiating mind #iles.
=owe%er, there is little to suggest that the tubules
contribute to the thin!ing process. 7%en generous
models o# human !nowledge and capability are more
than accounted #or by current estimates o# brain si.e,
based on contemporary models o# neuron #unctioning
that do not include tubules. In #act, e%en with these
tubule1less models, it appears that the brain is
conser%ati%ely designed with many more connections
(by se%eral orders o# magnitude& than it needs #or its
capabilities and capacity. 6ecent e$periments (e.g., the
+an 8iego Institute #or 'onlinear +cience
e$periments& showing that hybrid biological1
nonbiological networ!s per#orm similarly to all
biological networ!s, while not de#initi%e, are strongly
suggesti%e that our tubule1less models o# neuron
#unctioning are ade3uate. ?loyd 5atts so#tware
simulation o# his intricate model o# human auditory
processing uses orders o# magnitude less computation
than the networ!s o# neurons he is simulating, and
there is no suggestion that 3uantum computing is
needed.
=owe%er, e%en i# the tubules are important, it doesnt
change the projections I ha%e discussed abo%e to any
signi#icant degree. *ccording to my model o#
computational growth, i# the tubules multiplied neuron
comple$ity by a #actor o# a thousand (and !eep in mind
that our current tubule1less neuron models are already
comple$, including on the order o# a thousand
connections per neuron, multiple nonlinearities and
other details&, this would delay our reaching brain
capacity by only about D years. I# were o## by a #actor
o# a million, thats still only a delay o# :H years. *
#actor o# a billion is around ;4 years (!eep in mind
computation is growing by a double e$ponential&.
5ith regard to 3uantum computing, once again there is
nothing to suggest that the brain does 3uantum
computing. 4ust because 3uantum technology may be
#easible does not suggest that the brain is capable o# it.
*#ter all, we dont ha%e lasers or e%en radios in our
brains. *lthough some scientists ha%e claimed to
detect 3uantum wa%e collapse in the brain, no one has
suggested human capabilities that actually re3uire a
capacity #or 3uantum computing.
=owe%er, e%en i# the brain does do 3uantum
computing, this does not signi#icantly change the
outloo! #or human1le%el computing (and beyond& nor
does it suggest that brain downloading is in#easible.
First o# all, i# the brain does do 3uantum computing
this would only %eri#y that 3uantum computing is
#easible. ,here would be nothing in such a #inding to
suggest that 3uantum computing is restricted to
biological mechanisms. -iological 3uantum computing
mechanisms, i# they e$ist, could be replicated. Indeed,
recent e$periments with small scale 3uantum
computers appear to be success#ul. 7%en the
con%entional transistor relies on the 3uantum e##ect o#
electron tunneling.
9enrose suggests that it is impossible to per#ectly
replicate a set o# 3uantum states, so there#ore, per#ect
downloading is impossible. 5ell, how per#ect does a
download ha%e to beE I am at this moment in a %ery
di##erent 3uantum state (and di##erent in non13uantum
ways as well& than I was a minute ago (certainly in a
%ery di##erent state than I was be#ore I wrote this
paragraph&. I# we de%elop downloading technology to
the point where the 0copies0 are as close to the original
as the original person changes anyway in the course o#
one minute, that would be good enough #or any
concei%able purpose, yet does not re3uire copying
3uantum states. *s the technology impro%es, the
accuracy o# the copy could become as close as the
original changes within e%er brie#er periods o# time
(e.g., one second, one millisecond, one microsecond&.
5hen it was pointed out to 9enrose that neurons (and
e%en neural connections& were too big #or 3uantum
computing, he came up with the tubule theory as a
possible mechanism #or neural 3uantum computing. +o
the concern with 3uantum computing and tubules ha%e
been introduced together. I# one is searching #or
barriers to replicating brain #unction, it is an ingenious
theory, but it #ails to introduce any genuine barriers.
,here is no e%idence #or it, and e%en i# true, it only
delays matters by a decade or two. ,here is no reason
to belie%e that biological mechanisms (including
3uantum computing& are inherently impossible to
replicate using nonbiological materials and
mechanisms. 8o.ens o# contemporary e$periments are
success#ully per#orming just such replications.
The Noninvasive $urger'7Free ,eversible
Programmable "istributed -rain (mplant,
Full7(mmersion $hared +irtual ,ealit'
Environments, Experien&e -eamers, and
-rain Expansion
=ow will we apply technology that is more intelligent
than its creatorsE )ne might be tempted to respond
0>are#ullyP0 -ut lets ta!e a loo! at some e$amples.
>onsider se%eral e$amples o# the nanobot technology,
which, based on miniaturi.ation and cost reduction
trends, will be #easible within B0 years. In addition to
scanning your brain, the nanobots will also be able to
e$pand our e$periences and our capabilities.
'anobot technology will pro%ide #ully immersi%e,
totally con%incing %irtual reality in the #ollowing way.
,he nanobots ta!e up positions in close physical
pro$imity to e%ery interneuronal connection coming
#rom all o# our senses (e.g., eyes, ears, s!in&. 5e
already ha%e the technology #or electronic de%ices to
communicate with neurons in both directions that
re3uires no direct physical contact with the neurons.
For e$ample, scientists at the /a$ 9lanc! Institute
ha%e de%eloped 0neuron transistors0 that can detect the
#iring o# a nearby neuron, or alternati%ely, can cause a
nearby neuron to #ire, or suppress it #rom #iring. ,his
amounts to two1way communication between neurons
and the electronic1based neuron transistors. ,he
Institute scientists demonstrated their in%ention by
controlling the mo%ement o# a li%ing leech #rom their
computer. *gain, the primary aspect o# nanobot1based
%irtual reality that is not yet #easible is si.e and cost.
5hen we want to e$perience real reality, the nanobots
just stay in position (in the capillaries& and do nothing.
I# we want to enter %irtual reality, they suppress all o#
the inputs coming #rom the real senses, and replace
them with the signals that would be appropriate #or the
%irtual en%ironment. You (i.e., your brain& could
decide to cause your muscles and limbs to mo%e as you
normally would, but the nanobots again intercept these
interneuronal signals, suppress your real limbs #rom
mo%ing, and instead cause your %irtual limbs to mo%e
and pro%ide the appropriate mo%ement and
reorientation in the %irtual en%ironment.
,he web will pro%ide a panoply o# %irtual
en%ironments to e$plore. +ome will be recreations o#
real places, others will be #anci#ul en%ironments that
ha%e no 0real0 counterpart. +ome indeed would be
impossible in the physical world (perhaps, because
they %iolate the laws o# physics&. 5e will be able to
0go0 to these %irtual en%ironments by oursel%es, or we
will meet other people there, both real people and
simulated people. )# course, ultimately there wont be
a clear distinction between the two.
-y ;0B0, going to a web site will mean entering a #ull
immersion %irtual reality en%ironment. In addition to
encompassing all o# the senses, these shared
en%ironments can include emotional o%erlays as the
nanobots will be capable o# triggering the neurological
correlates o# emotions, se$ual pleasure, and other
deri%ati%es o# our sensory e$perience and mental
reactions.
In the same way that people today beam their li%es
#rom web cams in their bedrooms, 0e$perience
beamers0 circa ;0B0 will beam their entire #low o#
sensory e$periences, and i# so desired, their emotions
and other secondary reactions. 5ell be able to plug in
(by going to the appropriate web site& and e$perience
other peoples li%es as in the plot concept o# -eing
4ohn /al!o%ich. 9articularly interesting e$periences
can be archi%ed and reli%ed at any time.
5e wont need to wait until ;0B0 to e$perience shared
%irtual reality en%ironments, at least #or the %isual and
auditory senses. Full immersion %isual1auditory
en%ironments will be a%ailable by the end o# this
decade with images written directly onto our retinas by
our eyeglasses and contact lenses. *ll o# the
electronics #or the computation, image reconstruction,
and %ery high bandwidth wireless connection to the
Internet will be embedded in our glasses and wo%en
into our clothing, so computers as distinct objects will
disappear.
In my %iew, the most signi#icant implication o# the
+ingularity will be the merger o# biological and
nonbiological intelligence. First, it is important to point
out that well be#ore the end o# the twenty1#irst century,
thin!ing on nonbiological substrates will dominate.
-iological thin!ing is stuc! at :0
;A
calculations per
second (#or all biological human brains&, and that
#igure will not appreciably change, e%en with
bioengineering changes to our genome. 'onbiological
intelligence, on the other hand, is growing at a double
e$ponential rate and will %astly e$ceed biological
intelligence well be#ore the middle o# this century.
=owe%er, in my %iew, this nonbiological intelligence
should still be considered human as it is #ully
deri%ati%e o# the human1machine ci%ili.ation. ,he
merger o# these two worlds o# intelligence is not
merely a merger o# biological and nonbiological
thin!ing mediums, but more importantly one o#
method and organi.ation o# thin!ing.
)ne o# the !ey ways in which the two worlds can
interact will be through the nanobots. 'anobot
technology will be able to e$pand our minds in
%irtually any imaginable way. )ur brains today are
relati%ely #i$ed in design. *lthough we do add patterns
o# interneuronal connections and neurotransmitter
concentrations as a normal part o# the learning process,
the current o%erall capacity o# the human brain is
highly constrained, restricted to a mere hundred trillion
connections. -rain implants based on massi%ely
distributed intelligent nanobots will ultimately e$pand
our memories a trillion #old, and otherwise %astly
impro%e all o# our sensory, pattern recognition, and
cogniti%e abilities. +ince the nanobots are
communicating with each other o%er a wireless local
area networ!, they can create any set o# new neural
connections, can brea! e$isting connections (by
suppressing neural #iring&, can create new hybrid
biological1nonbiological networ!s, as well as add %ast
new nonbiological networ!s.
"sing nanobots as brain e$tenders is a signi#icant
impro%ement o%er the idea o# surgically installed
neural implants, which are beginning to be used today
(e.g., %entral posterior nucleus, subthalmic nucleus,
and %entral lateral thalamus neural implants to
counteract 9ar!insons 8isease and tremors #rom other
neurological disorders, cochlear implants, and others.&
'anobots will be introduced without surgery,
essentially just by injecting or e%en swallowing them.
,hey can all be directed to lea%e, so the process is
easily re%ersible. ,hey are programmable, in that they
can pro%ide %irtual reality one minute, and a %ariety o#
brain e$tensions the ne$t. ,hey can change their
con#iguration, and clearly can alter their so#tware.
9erhaps most importantly, they are massi%ely
distributed and there#ore can ta!e up billions or
trillions o# positions throughout the brain, whereas a
surgically introduced neural implant can only be
placed in one or at most a #ew locations.
The "ouble Exponential rowth of the E&onom'
"uring the 899:s Was Not a -ubble
Yet another mani#estation o# the law o# accelerating
returns as it rushes toward the +ingularity can be #ound
in the world o# economics, a world %ital to both the
genesis o# the law o# accelerating returns, and to its
implications. It is the economic imperati%e o# a
competiti%e mar!etplace that is dri%ing technology
#orward and #ueling the law o# accelerating returns. In
turn, the law o# accelerating returns, particularly as it
approaches the +ingularity, is trans#orming economic
relationships.
Firtually all o# the economic models taught in
economics classes, used by the Federal 6eser%e -oard
to set monetary policy, by Go%ernment agencies to set
economic policy, and by economic #orecasters o# all
!inds are #undamentally #lawed because they are based
on the intuiti%e linear %iew o# history rather than the
historically based e$ponential %iew. ,he reason that
these linear models appear to wor! #or a while is #or
the same reason that most people adopt the intuiti%e
linear %iew in the #irst place( e$ponential trends appear
to be linear when %iewed (and e$perienced& #or a brie#
period o# time, particularly in the early stages o# an
e$ponential trend when not much is happening. -ut
once the 0!nee o# the cur%e0 is achie%ed and the
e$ponential growth e$plodes, the linear models brea!
down. ,he e$ponential trends underlying producti%ity
growth are just beginning this e$plosi%e phase.
,he economy (%iewed either in total or per capita& has
been growing e$ponentially throughout this century(
,here is also a second le%el o# e$ponential growth, but
up until recently the second e$ponent has been in the
early phase so that the growth in the growth rate has
not been noticed. =owe%er, this has changed in this
past decade, during which the rate o# growth has been
noticeably e$ponential.
9roducti%ity (economic output per wor!er& has also
been growing e$ponentially. 7%en these statistics are
greatly understated because they do not #ully re#lect
signi#icant impro%ements in the 3uality and #eatures o#
products and ser%ices. It is not the case that 0a car is a
car20 there ha%e been signi#icant impro%ements in
sa#ety, reliability, and #eatures. ,here are a myriad o#
such e$amples. 9harmaceutical drugs are increasingly
e##ecti%e. Groceries ordered in #i%e minutes on the web
and deli%ered to your door are worth more than
groceries on a supermar!et shel# that you ha%e to #etch
yoursel#. >lothes custom manu#actured #or your uni3ue
body scan are worth more than clothes you happen to
#ind le#t on a store rac!. ,hese sorts o# impro%ements
are true #or most product categories, and none o# them
are re#lected in the producti%ity statistics.
,he statistical methods underlying the producti%ity
measurements tend to #actor out gains by essentially
concluding that we still only get one dollar o# products
and ser%ices #or a dollar despite the #act that we get
much more #or a dollar (e.g., compare a $:,000
computer today to one ten years ago&. "ni%ersity o#
>hicago 9ro#essor 9ete Llenow and "ni%ersity o#
6ochester 9ro#essor /ar! -ils estimate that the %alue
o# e$isting goods has been increasing at :.<Q per year
#or the past ;0 years because o# 3ualitati%e
impro%ements. ,his still does not account #or the
introduction o# entirely new products and product
categories. ,he -ureau o# ?abor +tatistics, which is
responsible #or the in#lation statistics, uses a model that
incorporates an estimate o# 3uality growth at only
0.<Q per year, re#lecting a systematic underestimate o#
3uality impro%ement and a resulting o%erestimate o#
in#lation by at least : percent per year.
8espite these wea!nesses in the producti%ity statistical
methods, the gains in producti%ity are now reaching
the steep part o# the e$ponential cur%e. ?abor
producti%ity grew at :.AQ per year until :DD4, then
rose at ;.4Q per year, and is now growing e%en more
rapidly. In the 3uarter ending 4uly B0, ;000, labor
producti%ity grew at <.BQ. /anu#acturing producti%ity
grew at 4.4Q annually #rom :DD< to :DDD, durables
manu#acturing at A.<Q per year.
,he :DD0s ha%e seen the most power#ul de#lationary
#orces in history. ,his is why we are not seeing
in#lation. Yes, its true that low unemployment, high
asset %alues, economic growth, and other such #actors
are in#lationary, but these #actors are o##set by the
double e$ponential trends in the price1per#ormance o#
all in#ormation based technologies( computation,
memory, communications, biotechnology,
miniaturi.ation, and e%en the o%erall rate o# technical
progress. ,hese technologies deeply a##ect all
industries.
5e are also undergoing massi%e disintermediation in
the channels o# distribution through the web and other
new communication technologies, as well as escalating
e##iciencies in operations and administration.
*ll o# the technology trend charts in this prcis e
represent massi%e de#lation. ,here are many e$amples
o# the impact o# these escalating e##iciencies. -9
*mocos cost #or #inding oil is now less than $: per
barrel, down #rom nearly $:0 in :DD:. 9rocessing an
internet transaction costs a ban! one penny, compared
to o%er $: using a teller ten years ago. * 6oland
-erger I 8eutsche -an! study estimates a cost sa%ings
o# $:;00 per 'orth *merican car o%er the ne$t #i%e
years. * more optimistic /organ +tanley study
estimates that Internet1based procurement will sa%e
Ford, G/, and 8aimler>hrysler about $;H00 per
%ehicle. +o#tware prices are de#lating e%en more
3uic!ly than computer hardware.
$oftware Pri&e7Performan&e *as #lso (mproved at
an Exponential ,ate
(7$ample( *utomatic +peech 6ecognition +o#tware
:DC<
:DD<
;000
9rice
$<,000
$<00
$<0
Focabulary +i.e (R words&
:,000
:0,000
:00,000
>ontinuous +peechE
'o
'o
Yes
"ser ,raining 6e3uired (/inutes&
:C0
A0
<
*ccuracy
9oor
Fair
Good
>urrent economic policy is based on outdated models
which include energy prices, commodity prices, and
capital in%estment in plant and e3uipment as !ey
dri%ing #actors, but do not ade3uately model
bandwidth, /I9s, megabytes, intellectual property,
!nowledge, and other increasingly %ital (and
increasingly increasing& constituents that are dri%ing
the economy.
,he economy 0wants0 to grow more than the B.<Q per
year, which constitutes the current 0speed limit0 that
the Federal 6eser%e ban! and other policy ma!ers ha%e
established as 0sa#e,0 meaning nonin#lationary. -ut in
!eeping with the law o# accelerating returns, the
economy is capable o# 0sa#ely0 establishing this le%el
o# growth in less than a year, implying a growth rate in
an entire year o# greater than B.<Q. 6ecently, the
growth rate has e$ceeded <Q.
'one o# this means that cycles o# recession will
disappear immediately. ,he economy still has some o#
the underlying dynamics that historically ha%e caused
cycles o# recession, speci#ically e$cessi%e
commitments such as capital intensi%e projects and the
o%erstoc!ing o# in%entories. =owe%er, the rapid
dissemination o# in#ormation, sophisticated #orms o#
online procurement, and increasingly transparent
mar!ets in all industries ha%e diminished the impact o#
this cycle. +o 0recessions0 are li!ely to be shallow and
short li%ed. ,he underlying long1term growth rate will
continue at a double e$ponential rate.
,he o%erall growth o# the economy re#lects completely
new #orms and layers o# wealth and %alue that did not
pre%iously e$ist, or least that did not pre%iously
constitute a signi#icant portion o# the economy (but do
now&( intellectual property, communication portals,
web sites, bandwidth, so#tware, data bases, and many
other new technology based categories.
,here is no need #or high interest rates to counter an
in#lation that doesnt e$ist. ,he in#lationary pressures
which e$ist are counterbalanced by all o# the
de#lationary #orces I%e mentioned. ,he current high
interest rates #ostered by the Federal 6eser%e -an! are
destructi%e, are causing trillions o# dollars o# lost
wealth, are regressi%e, hurt business and the middle
class, and are completely unnecessary.
,he Feds monetary policy is only in#luential because
people belie%e it to be. It has little real power. ,he
economy today is largely bac!ed by pri%ate capital in
the #orm o# a growing %ariety o# e3uity instruments.
,he portion o# a%ailable li3uidity in the economy that
the Fed actually controls is relati%ely insigni#icant. ,he
reser%es that ban!s and #inancial institutions maintain
with the Federal 6eser%e +ystem are less than $<0
billion, which is only 0.AQ o# the G89, and 0.;<Q o#
the li3uidity a%ailable in stoc!s.
6estricting the growth rate o# the economy to an
arbitrary limit ma!es as much sense as restricting the
rate at which a company can grow its re%enues11or its
mar!et cap. +peculati%e #e%er will certainly occur and
there will necessarily continue to be high pro#ile
#ailures and mar!et corrections. =owe%er the ability o#
technology companies to rapidly create new11real11
wealth is just one o# the #actors that will continue to
#uel ongoing double e$ponential growth in the
economy. ,hese policies ha%e led to an 0*lice in
5onderland0 situation in which the mar!et goes up on
bad economic news (because it means that more
unnecessary punishment will be a%oided& and goes
down on good economic news.
+pea!ing o# mar!et speculati%e #e%er and mar!et
corrections, the stoc! mar!et %alues #or so1called 0- to
-0 (-usiness to -usiness& and 0- to >0 (-usiness to
>onsumer& web portals and enabling technologies is
li!ely to come bac! strongly as it becomes clear that
economic transactions are indeed escalating toward e1
commerce, and that the (sur%i%ing& contenders are
capable o# demonstrating pro#itable business models.
,he intuiti%e linear assumption underlying economic
thin!ing reaches its most ludicrous conclusions in the
political debate surrounding the long1term #uture o# the
social security system. ,he economic models used #or
the social security projections are entirely linear, i.e.,
they re#lect #i$ed economic growth. ,his might be
%iewed as conser%ati%e planning i# we were tal!ing
about projections o# only a #ew years, but they become
utterly unrealistic #or the three to #our decades being
discussed. ,hese projections actually assume a #i$ed
rate o# growth o# B.<Q per year #or the ne$t #i#ty yearsP
,here are incredibly naS%e assumptions that bear on
both sides o# the argument. )n the one hand, there will
be radical e$tensions to human longe%ity, while on the
other hand, we will bene#it #rom #ar greater economic
e$pansion. ,hese #actors do not rule each other out,
howe%er, as the positi%e #actors are stronger, and will
ultimately dominate. /oreo%er, we are certain to
rethin! social security when we ha%e centenarians who
loo! and act li!e B0 year1olds (but who will thin!
much #aster than B0 year1olds circa the year ;000&.
*nother implication o# the law o# accelerating returns
is e$ponential growth in education and learning. )%er
the past :;0 years, we ha%e increased our in%estment
in L1:; education (per student and in constant dollars&
by a #actor o# ten. 5e ha%e a one hundred #old increase
in the number o# college students. *utomation started
by ampli#ying the power o# our muscles, and in recent
times has been ampli#ying the power o# our minds.
,hus, #or the past two centuries, automation has been
eliminating jobs at the bottom o# the s!ill ladder while
creating new (and better paying& jobs at the top o# the
s!ill ladder. +o the ladder has been mo%ing up, and
thus we ha%e been e$ponentially increasing
in%estments in education at all le%els.
)h, and about that 0o##er0 at the beginning o# this
prcis, consider that present stoc! %alues are based on
#uture e$pectations. Gi%en that the (literally& short
sighted linear intuiti%e %iew represents the ubi3uitous
outloo!, the common wisdom in economic
e$pectations are dramatically understated. *lthough
stoc! prices re#lect the consensus o# a buyer1seller
mar!et, it nonetheless re#lects the underlying linear
assumption regarding #uture economic growth. -ut the
law o# accelerating returns clearly implies that the
growth rate will continue to grow e$ponentially
because the rate o# progress will continue to accelerate.
*lthough (wea!ening& recessionary cycles will
continue to cause immediate growth rates to #luctuate,
the underlying rate o# growth will continue to double
appro$imately e%ery decade.
-ut wait a second, you said that I would get $40
trillion i# I read and understood this prcis .
,hats right. *ccording to my models, i# we replace the
linear outloo! with the more appropriate e$ponential
outloo!, current stoc! prices should triple. +ince
theres about $;0 trillion in the e3uity mar!ets, thats
$40 trillion in additional wealth.
-ut you said I would get that money.
'o, I said 0you0 would get the money, and thats why I
suggested reading the sentence care#ully. ,he 7nglish
word 0you0 can be singular or plural. I meant it in the
sense o# 0all o# you.0
I see, all o# us as in the whole world. -ut not e%eryone
will read this prcis .
5ell, but e%eryone could. +o i# all o# you read this
prcis and understand it, then economic e$pectations
would be based on the historical e$ponential model,
and thus stoc! %alues would increase.
You mean i# e%eryone understands it, and agrees with
it.
)!ay, I suppose I was assuming that.
Is that what you e$pect to happen.
5ell, actually, no. 9utting on my #uturist hat again, my
prediction is that indeed these %iews will pre%ail, but
only o%er time, as more and more e%idence o# the
e$ponential nature o# technology and its impact on the
economy becomes apparent. ,his will happen
gradually o%er the ne$t se%eral years, which will
represent a strong continuing updra#t #or the mar!et.
# !lear and Future "anger
,echnology has always been a double edged sword,
bringing us longer and healthier li#e spans, #reedom
#rom physical and mental drudgery, and many new
creati%e possibilities on the one hand, while
introducing new and salient dangers on the other. 5e
still li%e today with su##icient nuclear weapons (not all
o# which appear to be well accounted #or& to end all
mammalian li#e on the planet. -ioengineering is in the
early stages o# enormous strides in re%ersing disease
and aging processes. =owe%er, the means and
!nowledge will soon e$ist in a routine college
bioengineering lab (and already e$ists in more
sophisticated labs& to create un#riendly pathogens more
dangerous than nuclear weapons. *s technology
accelerates toward the +ingularity, we will see the
same intertwined potentials( a #east o# creati%ity
resulting #rom human intelligence e$panded a trillion1
#old combined with many gra%e new dangers.
>onsider unrestrained nanobot replication. 'anobot
technology re3uires billions or trillions o# such
intelligent de%ices to be use#ul. ,he most cost e##ecti%e
way to scale up to such le%els is through sel#1
replication, essentially the same approach used in the
biological world. *nd in the same way that biological
sel#1replication gone awry (i.e., cancer& results in
biological destruction, a de#ect in the mechanism
curtailing nanobot sel#1replication would endanger all
physical entities, biological or otherwise.
)ther primary concerns include 0who is controlling the
nanobotsE0 and 0who are the nanobots tal!ing toE0
)rgani.ations (e.g., go%ernments, e$tremist groups& or
just a cle%er indi%idual could put trillions o#
undetectable nanobots in the water or #ood supply o#
an indi%idual or o# an entire population. ,hese 0spy0
nanobots could then monitor, in#luence, and e%en
control our thoughts and actions. In addition to
introducing physical spy nanobots, e$isting nanobots
could be in#luenced through so#tware %iruses and other
so#tware 0hac!ing0 techni3ues. 5hen there is so#tware
running in our brains, issues o# pri%acy and security
will ta!e on a new urgency.
/y own e$pectation is that the creati%e and
constructi%e applications o# this technology will
dominate, as I belie%e they do today. -ut there will be
a %aluable (and increasingly %ocal& role #or a
concerned and constructi%e ?uddite mo%ement (i.e.,
anti1technologists inspired by early nineteenth century
wea%ers who destroyed labor1sa%ing machinery in
protest&.
I# we imagine describing the dangers that e$ist today to
people who li%ed a couple o# hundred years ago, they
would thin! it mad to ta!e such ris!s. )n the other
hand, how many people in the year ;000 would really
want to go bac! to the short, brutish, disease1#illed,
po%erty1stric!en, disaster1prone li%es that DD percent o#
the human race struggled through a couple o# centuries
agoE 5e may romantici.e the past, but up until #airly
recently, most o# humanity li%ed e$tremely #ragile
li%es where one all too common mis#ortune could spell
disaster. +ubstantial portions o# our species still li%e in
this precarious way, which is at least one reason to
continue technological progress and the economic
enhancement that accompanies it.
9eople o#ten go through three stages in e$amining the
impact o# #uture technology( awe and wonderment at
its potential to o%ercome age old problems, then a
sense o# dread at a new set o# gra%e dangers that
accompany these new technologies, #ollowed, #inally
and hope#ully, by the reali.ation that the only %iable
and responsible path is to set a care#ul course that can
reali.e the promise while managing the peril.
In his co%er story #or 5I678 Why The Future oesn't
!eed "s, -ill 4oy elo3uently described the plagues o#
centuries past, and how new sel#1replicating
technologies, such as mutant bioengineered pathogens,
and 0nanobots0 run amo!, may bring bac! long
#orgotten pestilence. Indeed these are real dangers. It is
also the case, which 4oy ac!nowledges, that it has been
technological ad%ances, such as antibiotics and
impro%ed sanitation, which has #reed us #rom the
pre%alence o# such plagues. +u##ering in the world
continues and demands our stead#ast attention. +hould
we tell the millions o# people a##licted with cancer and
other de%astating conditions that we are canceling the
de%elopment o# all bioengineered treatments because
there is a ris! that these same technologies may
someday be used #or male%olent purposesE =a%ing
as!ed the rhetorical 3uestion, I reali.e that there is a
mo%ement to do e$actly that, but I thin! most people
would agree that such broad based relin3uishment is
not the answer.
,he continued opportunity to alle%iate human distress
is one important moti%ation #or continuing
technological ad%ancement. *lso compelling are the
already apparent economic gains I discussed abo%e
which will continue to hasten in the decades ahead.
,he continued acceleration o# many intertwined
technologies are roads pa%ed with gold (I use the plural
here because technology is clearly not a single path&.
In a competiti%e en%ironment, it is an economic
imperati%e to go down these roads. 6elin3uishing
technological ad%ancement would be economic suicide
#or indi%iduals, companies, and nations.
5hich brings us to the issue o# relin3uishment, which
is -ill 4oys most contro%ersial recommendation and
personal commitment. I do #eel that relin3uishment at
the right le%el is part o# a responsible and constructi%e
response to these genuine perils. ,he issue, howe%er, is
e$actly this( at what le%el are we to relin3uish
technologyE
,ed Lac.yns!i would ha%e us renounce all o# it. ,his,
in my %iew, is neither desirable nor #easible, and the
#utility o# such a position is only underscored by the
senselessness o# Lac.yns!is deplorable tactics.
*nother le%el would be to #orego certain #ields2
nanotechnology, #or e$ample, that might be regarded
as too dangerous. -ut such sweeping stro!es o#
relin3uishment are e3ually untenable. 'anotechnology
is simply the ine%itable end result o# the persistent
trend toward miniaturi.ation which per%ades all o#
technology. It is #ar #rom a single centrali.ed e##ort,
but is being pursued by a myriad o# projects with many
di%erse goals.
)ne obser%er wrote(
0* #urther reason why industrial society
cannot be re#ormed. . . is that modern
technology is a uni#ied system in which
all parts are dependent on one another.
You cant get rid o# the 0bad0 parts o#
technology and retain only the 0good0
parts. ,a!e modern medicine, #or
e$ample. 9rogress in medical science
depends on progress in chemistry,
physics, biology, computer science and
other #ields. *d%anced medical
treatments re3uire e$pensi%e, high1tech
e3uipment that can be made a%ailable
only by a technologically progressi%e,
economically rich society. >learly you
cant ha%e much progress in medicine
without the whole technological system
and e%erything that goes with it.0
,he obser%er I am 3uoting is, again, ,ed Lac.yns!i.
*lthough one might properly resist Lac.yns!i as an
authority, I belie%e he is correct on the deeply
entangled nature o# the bene#its and ris!s. =owe%er,
Lac.yns!i and I clearly part company on our o%erall
assessment on the relati%e balance between the two.
-ill 4oy and I ha%e dialogued on this issue both
publicly and pri%ately, and we both belie%e that
technology will and should progress, and that we need
to be acti%ely concerned with the dar! side. I# -ill and
I disagree, its on the granularity o# relin3uishment that
is both #easible and desirable.
*bandonment o# broad areas o# technology will only
push them underground where de%elopment would
continue unimpeded by ethics and regulation. In such a
situation, it would be the less stable, less responsible
practitioners (e.g., the terrorists& who would ha%e all
the e$pertise.
I do thin! that relin3uishment at the right le%el needs
to be part o# our ethical response to the dangers o#
twenty #irst century technologies. )ne constructi%e
e$ample o# this is the proposed ethical guideline by the
Foresight Institute, #ounded by nanotechnology
pioneer 7ric 8re$ler, that nanotechnologists agree to
relin3uish the de%elopment o# physical entities that can
sel#1replicate in a natural en%ironment. *nother is a
ban on sel#1replicating physical entities that contain
their own codes #or sel#1replication. In what
nanotechnologist 6alph /er!le calls the 0-roadcast
*rchitecture,0 such entities would ha%e to obtain such
codes #rom a centrali.ed secure ser%er, which would
guard against undesirable replication. ,he -roadcast
*rchitecture is impossible in the biological world,
which represents at least one way in which
nanotechnology can be made sa#er than biotechnology.
In other ways, nanotech is potentially more dangerous
because nanobots can be physically stronger than
protein1based entities and more intelligent. It will
e%entually be possible to combine the two by ha%ing
nanotechnology pro%ide the codes within biological
entities (replacing 8'*&, in which case biological
entities can use the much sa#er -roadcast *rchitecture.
)ur ethics as responsible technologists should include
such 0#ine grained0 relin3uishment, among other
pro#essional ethical guidelines. )ther protections will
need to include o%ersight by regulatory bodies, the
de%elopment o# technology1speci#ic 0immune0
responses, as well as computer assisted sur%eillance by
law en#orcement organi.ations. /any people are not
aware that our intelligence agencies already use
ad%anced technologies such as automated word
spotting to monitor a substantial #low o# telephone
con%ersations. *s we go #orward, balancing our
cherished rights o# pri%acy with our need to be
protected #rom the malicious use o# power#ul twenty
#irst century technologies will be one o# many
pro#ound challenges. ,his is one reason that such
issues as an encryption 0trap door0 (in which law
en#orcement authorities would ha%e access to
otherwise secure in#ormation& and the F-I 0>arni%ore0
email1snooping system ha%e been so contentious.
*s a test case, we can ta!e a small measure o# com#ort
#rom how we ha%e dealt with one recent technological
challenge. ,here e$ists today a new #orm o# #ully
nonbiological sel# replicating entity that didnt e$ist
just a #ew decades ago( the computer %irus. 5hen this
#orm o# destructi%e intruder #irst appeared, strong
concerns were %oiced that as they became more
sophisticated, so#tware pathogens had the potential to
destroy the computer networ! medium they li%e in. Yet
the 0immune system0 that has e%ol%ed in response to
this challenge has been largely e##ecti%e. *lthough
destructi%e sel#1replicating so#tware entities do cause
damage #rom time to time, the injury is but a small
#raction o# the bene#it we recei%e #rom the computers
and communication lin!s that harbor them. 'o one
would suggest we do away with computers, local area
networ!s, and the Internet because o# so#tware %iruses.
)ne might counter that computer %iruses do not ha%e
the lethal potential o# biological %iruses or o#
destructi%e nanotechnology. *lthough true, this
strengthens my obser%ation. ,he #act that computer
%iruses are not usually deadly to humans only means
that more people are willing to create and release them.
It also means that our response to the danger is that
much less intense. >on%ersely, when it comes to sel#
replicating entities that are potentially lethal on a large
scale, our response on all le%els will be %astly more
serious.
,echnology will remain a double edged sword, and the
story o# the ,wenty First century has not yet been
written. It represents %ast power to be used #or all
human!inds purposes. 5e ha%e no choice but to wor!
hard to apply these 3uic!ening technologies to ad%ance
our human %alues, despite what o#ten appears to be a
lac! o# consensus on what those %alues should be.
Living Forever
)nce brain porting technology has been re#ined and
#ully de%eloped, will this enable us to li%e #ore%erE ,he
answer depends on what we mean by li%ing and dying.
>onsider what we do today with our personal
computer #iles. 5hen we change #rom one personal
computer to a less obsolete model, we dont throw all
our #iles away2 rather we copy them o%er to the new
hardware. *lthough our so#tware #iles do not necessary
continue their e$istence #ore%er, the longe%ity o# our
personal computer so#tware is completely separate and
disconnected #rom the hardware that it runs on. 5hen
it comes to our personal mind #ile, howe%er, when our
human hardware crashes, the so#tware o# our li%es dies
with it. =owe%er, this will not continue to be the case
when we ha%e the means to store and restore the
thousands o# trillions o# bytes o# in#ormation
represented in the pattern that we call our brains.
,he longe%ity o# ones mind #ile will not be dependent,
there#ore, on the continued %iability o# any particular
hardware medium. "ltimately so#tware1based humans,
albeit %astly e$tended beyond the se%ere limitations o#
humans as we !now them today, will li%e out on the
web, projecting bodies whene%er they need or want
them, including %irtual bodies in di%erse realms o#
%irtual reality, holographically projected bodies,
physical bodies comprised o# nanobot swarms, and
other #orms o# nanotechnology.
* so#tware1based human will be #ree, there#ore, #rom
the constraints o# any particular thin!ing medium.
,oday, we are each con#ined to a mere hundred trillion
connections, but humans at the end o# the twenty1#irst
century can grow their thin!ing and thoughts without
limit. 5e may regard this as a #orm o# immortality,
although it is worth pointing out that data and
in#ormation do not necessarily last #ore%er. *lthough
not dependent on the %iability o# the hardware it runs
on, the longe%ity o# in#ormation depends on its
rele%ance, utility, and accessibility. I# you%e e%er tried
to retrie%e in#ormation #rom an obsolete #orm o# data
storage in an old obscure #ormat (e.g., a reel o#
magnetic tape #rom a :DH0 minicomputer&, you will
understand the challenges in !eeping so#tware %iable.
=owe%er, i# we are diligent in maintaining our mind
#ile, !eeping current bac!ups, and porting to current
#ormats and mediums, then a #orm o# immortality can
be attained, at least #or so#tware1based humans. )ur
mind #ile11our personality, s!ills, memories11all o# that
is lost today when our biological hardware crashes.
5hen we can access, store, and restore that
in#ormation, then its longe%ity will no longer be tied to
our hardware permanence.
Is this #orm o# immortality the same concept as a
physical human, as we !now them today, li%ing
#ore%erE In one sense it is, because as I pointed out
earlier, our contemporary sel%es are not a constant
collection o# matter either. )nly our pattern o# matter
and energy persists, and e%en that gradually changes.
+imilarly, it will be the pattern o# a so#tware human
that persists and de%elops and changes gradually.
-ut is that person based on my mind #ile, who migrates
across many computational substrates, and who
outli%es any particular thin!ing medium, really meE
5e come bac! to the same 3uestions o# consciousness
and identity, issues that ha%e been debated since the
9latonic dialogues. *s we go through the twenty1#irst
century, these will not remain polite philosophical
debates, but will be con#ronted as %ital, practical,
political, and legal issues.
* related 3uestion is 0is death desirableE0 * great deal
o# our e##ort goes into a%oiding it. 5e ma!e
e$traordinary e##orts to delay it, and indeed o#ten
consider its intrusion a tragic e%ent. Yet we might #ind
it hard to li%e without it. 5e consider death as gi%ing
meaning to our li%es. It gi%es importance and %alue to
time. ,ime could become meaningless i# there were
too much o# it.
The Next $tep in Evolution and the Purpose of Life
-ut I regard the #reeing o# the human mind #rom its
se%ere physical limitations o# scope and duration as the
necessary ne$t step in e%olution. 7%olution, in my
%iew, represents the purpose o# li#e. ,hat is, the
purpose o# li#e11and o# our li%es11is to e%ol%e. ,he
+ingularity then is not a gra%e danger to be a%oided. In
my %iew, this ne$t paradigm shi#t represents the goal
o# our ci%ili.ation.
5hat does it mean to e%ol%eE 7%olution mo%es toward
greater comple$ity, greater elegance, greater
!nowledge, greater intelligence, greater beauty, greater
creati%ity, and more o# other abstract and subtle
attributes such as lo%e. *nd God has been called all
these things, only without any limitation( in#inite
!nowledge, in#inite intelligence, in#inite beauty,
in#inite creati%ity, in#inite lo%e, and so on. )# course,
e%en the accelerating growth o# e%olution ne%er
achie%es an in#inite le%el, but as it e$plodes
e$ponentially, it certainly mo%es rapidly in that
direction. +o e%olution mo%es ine$orably toward our
conception o# God, albeit ne%er 3uite reaching this
ideal. ,hus the #reeing o# our thin!ing #rom the se%ere
limitations o# its biological #orm may be regarded as an
essential spiritual 3uest.
In ma!ing this statement, it is important to emphasi.e
that terms li!e e%olution, destiny, and spiritual 3uest
are obser%ations about the end result, not the basis #or
these predictions. I am not saying that technology will
e%ol%e to human le%els and beyond simply because it
is our destiny and because o# the satis#action o# a
spiritual 3uest. 6ather my projections result #rom a
methodology based on the dynamics underlying the
(double& e$ponential growth o# technological
processes. ,he primary #orce dri%ing technology is
economic imperati%e. 5e are mo%ing toward machines
with human le%el intelligence (and beyond& as the
result o# millions o# small ad%ances, each with their
own particular economic justi#ication.
,o use an e$ample #rom my own e$perience at one o#
my companies (Lur.weil *pplied Intelligence&,
whene%er we came up with a slightly more intelligent
%ersion o# speech recognition, the new %ersion
in%ariably had greater %alue than the earlier generation
and, as a result, sales increased. It is interesting to note
that in the e$ample o# speech recognition so#tware, the
three primary sur%i%ing competitors stayed %ery close
to each other in the intelligence o# their so#tware. *
#ew other companies that #ailed to do so (e.g., +peech
+ystems& went out o# business. *t any point in time,
we would be able to sell the %ersion prior to the latest
%ersion #or perhaps a 3uarter o# the price o# the current
%ersion. *s #or %ersions o# our technology that were
two generations old, we couldnt e%en gi%e those away.
,his phenomenon is not only true #or pattern
recognition and other 0*I0 so#tware, but applies to all
products, #rom bread ma!ers to cars. *nd i# the
product itsel# doesnt e$hibit some le%el o# intelligence,
then intelligence in the manu#acturing and mar!eting
methods ha%e a major e##ect on the success and
pro#itability o# an enterprise.
,here is a %ital economic imperati%e to create more
intelligent technology. Intelligent machines ha%e
enormous %alue. ,hat is why they are being built.
,here are tens o# thousands o# projects that are
ad%ancing intelligent machines in di%erse incremental
ways. ,he support #or 0high tech0 in the business
community (mostly so#tware& has grown enormously.
5hen I started my optical character recognition ()>6&
and speech synthesis company (Lur.weil >omputer
9roducts, Inc.& in :DH4, there were only a hal#1do.en
high technology I9)s that year. ,he number o# such
deals has increased one hundred #old and the number
o# dollars in%ested has increased by more than one
thousand #old in the past ;< years. In the #our years
between :DD< and :DDD alone, high tech %enture
capital deals increased #rom just o%er $: billion to
appro$imately $:< billion.
5e will continue to build more power#ul
computational mechanisms because it creates
enormous %alue. 5e will re%erse1engineer the human
brain not simply because it is our destiny, but because
there is %aluable in#ormation to be #ound there that will
pro%ide insights in building more intelligent (and more
%aluable& machines. 5e would ha%e to repeal
capitalism and e%ery %isage o# economic competition
to stop this progression.
-y the second hal# o# this ne$t century, there will be
no clear distinction between human and machine
intelligence. )n the one hand, we will ha%e biological
brains %astly e$panded through distributed nanobot1
based implants. )n the other hand, we will ha%e #ully
nonbiological brains that are copies o# human brains,
albeit also %astly e$tended. *nd we will ha%e a myriad
o# other %arieties o# intimate connection between
human thin!ing and the technology it has #ostered.
"ltimately, nonbiological intelligence will dominate
because it is growing at a double e$ponential rate,
whereas #or all practical purposes biological
intelligence is at a standstill. =uman thin!ing is stuc!
at :0
;A
calculations per second (#or all biological
humans&, and that #igure will ne%er appreciably change
(e$cept #or a small increase resulting #rom genetic
engineering&. 'onbiological thin!ing is still millions o#
times less today, but the cross o%er will occur be#ore
;0B0. -y the end o# the twenty1#irst century,
nonbiological thin!ing will be trillions o# trillions o#
times more power#ul than that o# its biological
progenitors, although still o# human origin. It will
continue to be the human1machine ci%ili.ation ta!ing
the ne$t step in e%olution.
/ost #orecasts o# the #uture seem to ignore the
re%olutionary impact o# the +ingularity in our human
destiny( the ine%itable emergence o# computers that
match and ultimately %astly e$ceed the capabilities o#
the human brain, a de%elopment that will be no less
important than the e%olution o# human intelligence
itsel# some thousands o# centuries ago. *nd the
primary reason #or this #ailure is that they are based on
the intuiti%e but short sighted linear %iew o# history.
-e#ore the ne$t century is o%er, the 7arths technology1
creating species will merge with its computational
technology. ,here will not be a clear distinction
between human and machine. *#ter all, what is the
di##erence between a human brain enhanced a trillion
#old by nanobot1based implants, and a computer whose
design is based on high resolution scans o# the human
brain, and then e$tended a trillion1#oldE
Wh' $ET( Will Fail ;and wh' we are alone in the
.niverse2
,he law o# accelerating returns implies that by ;0DD,
the intelligence that will ha%e emerged #rom human1
machine ci%ili.ation will be trillions o# trillions o#
times more power#ul than it is today, dominated o#
course by its nonbiological #orm.
+o what does this ha%e to do with +7,I (the +earch #or
7$tra ,errestrial Intelligence&E ,he naS%e %iew, going
bac! to pre1>opernican days, was that the 7arth was at
the center o# the "ni%erse, and human intelligence its
greatest gi#t (ne$t to God&. ,he more in#ormed recent
%iew is that e%en i# the li!elihood o# a star ha%ing a
planet with a technology creating species is %ery low
(e.g., one in a million&, there are so many stars (i.e.,
billions o# trillions o# them&, that there are bound to be
many with ad%anced technology.
,his is the %iew behind +7,I, was my %iew until
recently, and is the common in#ormed %iew today.
*lthough +7,I has not yet loo!ed e%erywhere, it has
already co%ered a substantial portion o# the "ni%erse.
>hart by +cienti#ic *merican
In the abo%e diagram (courtesy o# +cienti#ic
*merican&, we can see that +7,I has already
thoroughly searched all star systems within :0
H
light1
years #rom 7arth #or alien ci%ili.ations capable (and
willing& to transmit at a power o# at least :0
;<
watts, a
so1called ,ype II ci%ili.ation (and all star systems
within :0
A
light1years #or transmission o# at least :0
:C

watts, and so on&. 'o sign o# intelligence has been
#ound as o# yet.
In a recent email to my research assistant, 8r. +eth
+hosta! o# the +7,I Institute points out that a new
comprehensi%e targeted search, called 9roject 9hoeni$,
which has up to :00 times the sensiti%ity and co%ers a
greater range o# the radio dial as compared to pre%ious
searches, has only been applied thus #ar to <00 star
systems, which is, o# course only a minute #raction o#
the hal# trillion star systems in just our own gala$y.
=owe%er, according to my model, once a ci%ili.ation
achie%es our own le%el (07arth1le%el0& o# radio
transmission, it ta!es no more than one century, two at
the most, to achie%e what +7,I calls a ,ype II
ci%ili.ation. I# the assumption that there are at least
millions o# radio capable ci%ili.ations out there, and
that these ci%ili.ations are spread out o%er millions
(indeed billions& o# years o# de%elopment, then surely
there ought to be millions that ha%e achie%ed ,ype II
status.
Incidentally, this is not an argument against the +7,I
project, which in my %iew should ha%e the highest
possible priority because the negati%e #inding is no less
signi#icant than a positi%e result.
It is odd that we #ind the cosmos so silent. 5here is
e%erybodyE ,here should be millions o# ci%ili.ations
%astly more ad%anced than our own, so we should be
noticing their broadcasts. * su##iciently ad%anced
ci%ili.ation would not be li!ely to restrict its
broadcasts to subtle signals on obscure #re3uencies.
5hy are they so silent, and so shyE
*s I ha%e studied the implications o# the law o#
accelerating returns, I ha%e come to a di##erent %iew.
-ecause e$ponential growth is so e$plosi%e, it is the
case that once a species de%elops computing
technology, it is only a matter o# a couple o# centuries
be#ore the nonbiological #orm o# their intelligence
e$plodes. It permeates %irtually all matter in their
%icinity, and then ine%itably e$pands outward close to
the ma$imum speed that in#ormation can tra%el. )nce
the nonbiological intelligence emerging #rom that
species technology has saturated its %icinity (and the
nature o# this saturation is another comple$ issue,
which I wont deal with in this prcis&, it has no other
way to continue to e%ol%e but to e$pand outwardly.
,he e$pansion does not start out at the ma$imum
speed, but 3uic!ly achie%es a speed within a
%anishingly small delta #rom the ma$imum speed.
5hat is the ma$imum speedE 5e currently understand
this to be the speed o# light, but there are already
tantali.ing hints that this may not be an absolute limit.
,here were recent e$periments that measured the #light
time o# photons at nearly twice the speed o# light, a
result o# 3uantum uncertainty on their position.
=owe%er, this result is actually not use#ul #or this
analysis, because it does not actually allow in#ormation
to be communicated at #aster than the speed o# light,
and we are #undamentally interested in communication
speed.
Ouantum disentanglement has been measured at many
times the speed o# light, but this is only
communicating randomness (pro#ound 3uantum
randomness& at speeds #ar greater than the speed o#
light2 again, this is not communication o# in#ormation
(but is o# great interest #or restoring encryption, a#ter
3uantum computing destroys it&. ,here is the potential
#or worm holes (or #olds o# the "ni%erse in dimensions
beyond the three %isible ones&, but this is not really
tra%eling at #aster than the speed o# light, it just means
that the topology o# the "ni%erse is not the simple
three dimensional space that naS%e physics implies. -ut
we already !new that. =owe%er, i# worm holes or #olds
in the "ni%erse are ubi3uitous, then perhaps these short
cuts would allow us to get e%erywhere 3uic!ly. 5ould
anyone be shoc!ed i# some subtle ways o# getting
around this speed limit were disco%eredE *nd no
matter how subtle, su##iciently subtle technology will
#ind ways to apply it. ,he point is that i# there are ways
around this limit (or any other currently understood
limit&, then the e$traordinary le%els o# intelligence that
our human1machine ci%ili.ation will achie%e will #ind
those ways and e$ploit them.
+o #or now, we can say that ultra high le%els o#
intelligence will e$pand outward at the speed o# light,
but recogni.e that this may not be the actual limit o#
the speed o# e$pansion, or e%en i# the limit is the speed
o# light that this limit may not restrict reaching other
locations 3uic!ly.
>onsider that the time spans #or biological e%olution
are measured in millions and billions o# years, so i#
there are other ci%ili.ations out there, they would be
spread out by huge spans o# time. I# there are a lot o#
them, as contemporary thin!ing implies, then it would
be %ery unli!ely that at least some o# them would not
be ahead o# us. ,hat at least is the +7,I assumption.
*nd i# they are ahead o# us, they li!ely would be ahead
o# us by huge spans o# time. ,he li!elihood that any
ci%ili.ation that is ahead o# us is ahead o# us by only a
#ew decades is e$tremely small.
I# the +7,I assumption that there are many (e.g.,
millions& o# technological (at least radio capable&
ci%ili.ations is correct, then at least some o# them (i.e.,
millions o# them& would be way ahead o# us. -ut it
ta!es only a #ew centuries at most #rom the ad%ent o#
computation #or that ci%ili.ation to e$pand outward at
at least light speed. Gi%en this, how can it be that we
ha%e not noticed themE
,he conclusion I reach is that it is li!ely that there are
no such other ci%ili.ations. In other words, we are in
the lead. ,hats right, our humble ci%ili.ation with its
8odge pic! up truc!s, #ried chic!en #ast #ood, and
ethnic cleansings (and computationP& is in the lead.
'ow how can that beE Isnt this e$tremely unli!ely
gi%en the billions o# trillions o# li!ely planetsE Indeed
it is %ery unli!ely. -ut e3ually unli!ely is the e$istence
o# our "ni%erse with a set o# laws o# physics so
e$3uisitely precisely what is needed #or the e%olution
o# li#e to be possible. -ut by the *nthropic principle, i#
the "ni%erse didnt allow the e%olution o# li#e we
wouldnt be here to notice it. Yet here we are. +o by
the same *nthropic principle, were here in the lead in
the "ni%erse. *gain, i# we werent here, we would not
be noticing it.
?ets consider some arguments against this perspecti%e.
9erhaps there are e$tremely ad%anced technological
ci%ili.ations out there, but we are outside their light
sphere o# intelligence. ,hat is, they ha%ent gotten here
yet. )!ay, in this case, +7,I will still #ail because we
wont be able to see (or hear& them, at least not be#ore
we reach +ingularity.
9erhaps they are amongst us, but ha%e decided to
remain in%isible to us. Incidentally, I ha%e always
considered the science #iction notion o# large space
ships with large s3uishy creatures similar to us to be
%ery unli!ely. *ny ci%ili.ation sophisticated enough to
ma!e the trip here would ha%e long since passed the
point o# merging with their technology and would not
need to send such physically bul!y organisms and
e3uipment. +uch a ci%ili.ation would not ha%e any
unmet material needs that re3uire it to steal physical
resources #rom us. ,hey would be here #or obser%ation
only, to gather !nowledge, which is the only resource
o# %alue to such a ci%ili.ation. ,he intelligence and
e3uipment needed #or such obser%ation would be
e$tremely small. In this case, +7,I will still #ail
because i# this ci%ili.ation decided that it did not want
us to notice it, then it would succeed in that desire.
Leep in mind that they would be %astly more
intelligent than we are today. 9erhaps they will re%eal
themsel%es to us when we achie%e the ne$t le%el o# our
e%olution, speci#ically merging our biological brains
with our technology, which is to say, a#ter the
+ingularity. /oreo%er, gi%en that the +7,I assumption
implies that there are millions o# such highly
de%eloped ci%ili.ations, it seems odd that all o# them
ha%e made the same decision to stay out o# our way.
Wh' (ntelligen&e is More Powerful than Ph'si&s
*s intelligence saturates the matter and energy
a%ailable to it, it turns dumb matter into smart matter.
*lthough smart matter still nominally #ollows the laws
o# physics, it is so e$3uisitely intelligent that it can
harness the most subtle aspects o# the laws to
manipulate matter and energy to its will. +o it would at
least appear that intelligence is more power#ul than
physics.
9erhaps what I should say is that intelligence is
more power#ul than cosmology. ,hat is, once matter
e%ol%es into smart matter (matter #ully saturated with
intelligence&, it can manipulate matter and energy to do
whate%er it wants. ,his perspecti%e has not been
considered in discussions o# #uture cosmology. It is
assumed that intelligence is irrele%ant to e%ents and
processes on a cosmological scale. +tars are born and
die2 gala$ies go through their cycles o# creation and
destruction. ,he "ni%erse itsel# was born in a big bang
and will end with a crunch or a whimper, were not yet
sure which. -ut intelligence has little to do with it.
Intelligence is just a bit o# #roth, an ebullition o# little
creatures darting in and out o# ine$orable uni%ersal
#orces. ,he mindless mechanism o# the "ni%erse is
winding up or down to a distant #uture, and theres
nothing intelligence can do about it.
,hats the common wisdom, but I dont agree with it.
Intelligence will be more power#ul than these
impersonal #orces. )nce a planet yields a technology
creating species and that species creates computation
(as has happened here on 7arth&, it is only a matter o# a
#ew centuries be#ore its intelligence saturates the
matter and energy in its %icinity, and it begins to
e$pand outward at the speed o# light or greater. It will
then o%ercome gra%ity (through e$3uisite and %ast
technology& and other cosmological #orces (or, to be
#ully accurate, will maneu%er and control these #orces&
and create the "ni%erse it wants. ,his is the goal o# the
+ingularity.
5hat !ind o# "ni%erse will that beE 5ell, just wait and
see.
Plan to $ti&< #round
/ost o# you (again Im using the plural #orm o# the
word& are li!ely to be around to see the +ingularity.
,he e$panding human li#e span is another one o# those
e$ponential trends. In the eighteenth century, we added
a #ew days e%ery year to human longe%ity2 during the
nineteenth century we added a couple o# wee!s each
year2 and now were adding almost a hal# a year e%ery
year. 5ith the re%olutions in genomics, proteomics,
rational drug design, therapeutic cloning o# our own
organs and tissues, and related de%elopments in bio1
in#ormation sciences, we will be adding more than a
year e%ery year within ten years. +o ta!e care o#
yoursel# the old #ashioned way #or just a little while
longer, and you may actually get to e$perience the ne$t
#undamental paradigm shi#t in our destiny.
>opyright (>& 6aymond Lur.weil ;00:
>hart Graphics by -rett 6ampataI8igital )rganism

Вам также может понравиться