||Chapter 1: The Evolution of Mind in the Twenty-First Century
An analysis of the history of technology shows that technological change is exponential, contrary to the common-sense "intuitive linear" view. So we won't experience 100 years of progress in the 21st century -- it will be approximately 20,000 years of progress (at today's rate). The "returns," such as chip speed and cost-effectiveness, also increase exponentially. There's even exponential growth in the rate of exponential growth. This exponential growth is not restricted to hardware, but with accelerating gains in brain reverse engineering, also applies to software. Within a few decades, machine intelligence will surpass human intelligence, allowing nonbiological intelligence to combine the subtleties of human intelligence with the speed and knowledge sharing ability of machines. The results will include the merger of biological and nonbiological intelligence, downloading the brain and immortal software-based humans -- the next step in evolution.
Originally published in print June 18, 2002 in Are
We Spiritual Machines? Ray Kurzweil vs. the Critics of Strong AI
by the Discovery
Institute. Published on KurzweilAI.net on June 18, 2002.
An Overview of the Next Several Decades
The intelligence of machines—nonbiological entities—will
exceed human intelligence early in this century. By intelligence,
I include all the diverse and subtle ways in which humans are intelligent—including
musical and artistic aptitude, creativity, physically moving through
the world, and even responding to emotion. By 2019, a $1,000 computer
will match the processing power of the human brain—about 20
million billion calculations per second. This level of processing
power is a necessary but not sufficient condition for achieving
human-level intelligence in a machine. Organizing these resources—the
“software” of intelligence—will take us to 2029,
by which time your average personal computer will be equivalent
to a thousand human brains.
Once a computer achieves a level of intelligence comparable to
human intelligence, it will necessarily soar past it. A key advantage
of nonbiological intelligence is that machines can easily share
their knowledge. If I learn French, or read War and Peace, I can’t
readily download that learning to you. You have to acquire that
scholarship the same painstaking way that I did. My knowledge, embedded
in a vast pattern of neurotransmitter concentrations and interneuronal
connections, cannot be quickly accessed or transmitted. But we won’t
leave out quick downloading ports in our nonbiological equivalents
of human neuron clusters. When one computer learns a skill or gains
an insight, it can immediately share that wisdom with billions of
As a contemporary example, we spent years teaching one research
computer how to recognize continuous human speech. We exposed it
to thousands of hours of recorded speech, corrected its errors,
and patiently improved its performance. Finally, it became quite
adept at recognizing speech (I dictated most of my recent book to
it). Now if you want your own personal computer to recognize speech,
it doesn’t have to go through the same process; you can just
download the fully trained program in seconds. Ultimately, billions
of nonbiological entities can be the master of all human and machine
acquired knowledge. Computers are also potentially millions of times
faster than human neural circuits, and have far more reliable memories.
One approach to designing intelligent computers will be to copy
the human brain, so these machines will seem very human. And through
nanotechnology, which is the ability to create physical objects
atom by atom, they will have humanlike—albeit greatly enhanced—bodies
as well. Having human origins, they will claim to be human, and
to have human feelings. And being immensely intelligent, they’ll
be very convincing when they tell us these things. But are these
feelings “real,” or just apparently real? I will discuss
this subtle but vital distinction below. First it is important to
understand the nature of nonbiological intelligence, and how it
Keep in mind that this is not an alien invasion of intelligent
machines. It is emerging from within our human-machine civilization.
There will not be a clear distinction between human and machine
as we go through the twenty-first century. First of all, we will
be putting computers—neural implants—directly into our
brains. We’ve already started down this path. We have ventral
posterior nucleus, subthalmic nucleus, and ventral lateral thalamus
neural implants to counteract Parkinson’s Disease and tremors
from other neurological disorders. I have a deaf friend who now
can hear what I am saying because of his cochlear implant. Under
development is a retina implant that will perform a similar function
for blind individuals, basically replacing certain visual processing
circuits of the retina and nervous system. Recently scientists from
Emory University placed a chip in the brain of a paralyzed stroke
victim who can now begin to communicate and control his environment
directly from his brain.
In the 2020s, neural implants will not be just for disabled people,
and introducing these implants into the brain will not require surgery,
but more about that later. There will be ubiquitous use of neural
implants to improve our sensory experiences, perception, memory,
and logical thinking.
These “noninvasive” implants will also plug us in directly
to the World Wide Web. By 2030, “going to a web site”
will mean entering a virtual reality environment. The implant will
generate the streams of sensory input that would otherwise come
from our real senses, thus creating an all-encompassing virtual
environment that responds to the behavior of our own virtual body
(and those of others) in the virtual environment. This technology
will enable us to have virtual reality experiences with other people—or
simulated people—without requiring any equipment not already
in our heads. And virtual reality will not be the crude experience
that one can experience in today’s arcade games. Virtual reality
will be as realistic, detailed, and subtle as real reality. So instead
of just phoning a friend, you can meet in a virtual French café
in Paris, or take a walk on a virtual Mediterranean beach, and it
will seem very real. People will be able to have any type of experience
with anyone—business, social, romantic, sexual—regardless
of physical proximity.
The Growth of Computing
To see into the future, we need insight into the past. We need
to discern the relevant trends and their interactions. Many projections
of the future suffer from three common failures. The first is that
people often consider only one or two iterations of advancement
in a technology, as if progress would then come to a halt.
A second is focusing on only one aspect of technology, without
considering the interactions and synergies from developments in
multiple fields (e.g. computational substrates and architectures,
software and artificial intelligence, communication, nanotechnology,
brain scanning and reverse engineering, amongst others).
By far the most important failure is to fail to adequately take
into consideration the accelerating pace of technology. Many predictions
do not factor this in with any consistent methodology, if at all.
Ten thousand years ago, there was little salient technological change
in even a thousand years. A thousand years ago, progress was much
faster and a paradigm shift required only a century or two. In the
nineteenth century, we saw more technological change than in the
nine centuries preceding it. Then in the first twenty years of the
twentieth century, we saw more advancement than in all of the nineteenth
century. Now, paradigm shifts (and new business models) take place
in only a few years time (lately it appears to be even less than
that). Just a decade ago, the Internet was in a formative stage
and the World Wide Web had yet to emerge
The fact that the successful application of certain innovations
(e.g., the laser) may have taken several decades in the past half
century does not mean that, going forward, comparable changes will
take nearly as long. The type of transformation that required thirty
years during the last half century will take only five to seven
years going forward. And the pace will continue to accelerate. It
is vital to consider the implications of this phenomenon; progress
is not linear, but exponential.
This “law of accelerating returns,” as I call it, is
true of any evolutionary process. It was true of the evolution of
life forms, which required billions of years for the first steps
(e.g. primitive cells); later on progress accelerated. During the
Cambrian explosion, major paradigm shifts took only tens of millions
of years. Later on, humanoids developed over a period of only millions
of years, and Homo sapiens over a period of hundreds of thousands
With the advent of a technology creating species, the exponential
pace became too fast for evolution through DNA-guided protein synthesis
and moved on to human-created technology. The first technological
steps — sharp edges, fire, the wheel—took tens of thousands
of years, and have accelerated ever since.
Technology is evolution by other means. It is the cutting edge
of evolution today, moving far faster than DNA-based evolution.
However, unlike biological evolution, technology is not a “blind
watchmaker.” (Actually, I would prefer the phrase “mindless
watchmaker” as more descriptive and less insensitive to the
visually impaired.) The process of humans creating technology, then,
is a “mindful watchmaker.” An implication is that we do
have the ability (and the responsibility) to guide this evolutionary
process in a constructive direction.
Technology goes beyond mere toolmaking; it is a process of creating
ever more powerful technology using the tools from the previous
round of innovation. In this way, human technology is distinguished
from the toolmaking of other species. There is a record of each
stage of technology, and each new stage of technology builds on
the order of the previous stage. Technology, therefore, is a continuation
of the evolutionary process that gave rise to the technology creating
species in the first place.
It is critical when considering the future to use a systematic
methodology that considers these three issues: (i) iterations of
technological progress do not just stop at an arbitrary point, (ii)
diverse developments interact, and, most importantly, (iii) the
pace of technological innovation accelerates. This third item can
be quantified, which I discuss in the sidebar below. Although these
formulas are not perfect models, they do provide a framework for
considering future developments. I’ve used this methodology
for the past twenty years, and the predictions derived from this
method in the 1980s have held up rather well.
One very important trend is referred to as “Moore’s Law.”
Gordon Moore, one of the inventors of integrated circuits, and then
chairman of Intel, noted in the mid-1970s that we could squeeze
twice as many transistors on an integrated circuit every twenty-four
months. The implication is that computers, which are built from
integrated circuits, are doubling in power every two years. Lately,
the rate has been even faster.
After sixty years of devoted service, Moore’s Law will die
a dignified death no later than the year 2019. By that time, transistor
features will be just a few atoms in width, and the strategy of
ever finer photolithography will have run its course. So, will that
be the end of the exponential growth of computing?
Don’t bet on it.
If we plot the speed (in instructions per second) per $1000 (in
constant dollars) of 49 famous calculators and computers spanning
the entire twentieth century, we note some interesting observations.
It is important to note that Moore’s Law of Integrated Circuits
was not the first, but the fifth paradigm to provide accelerating
price-performance. Computing devices have been consistently multiplying
in power (per unit of time) from the mechanical calculating devices
used in the 1890 U.S. census, to Turing’s relay-based “Robinson”
machine that cracked the Nazi enigma code, to the CBS vacuum tube
computer that predicted the election of Eisenhower, to the transistor-based
machines used in the first space launches, to the integrated-circuit-based
personal computer which I used to dictate (and automatically transcribe)
But I noticed something else surprising. When I plotted the 49
machines on a logarithmic graph (where a straight line means exponential
growth), I didn’t get a straight line. What I got was another
exponential curve. In other words, there’s exponential growth
in the rate of exponential growth. Computer speed (per unit cost)
doubled every three years between 1910 and 1950, doubled every two
years between 1950 and 1966, and is now doubling every year.
Wherefrom Moore’s Law:
Where does Moore’s Law come from? What is behind this remarkably
predictable phenomenon? I have seen relatively little written about
the ultimate source of this trend. Is it just “a set of industry
expectations and goals,” as Randy Isaac, head of basic science
at IBM, contends?
In my view, it is one manifestation (among many) of the exponential
growth of the evolutionary process that is technology. Just as the
pace of an evolutionary process accelerates, the “returns”
(i.e., the output, the products) of an evolutionary process grow
exponentially. The exponential growth of computing is a marvelous
quantitative example of the exponentially growing returns from an
evolutionary process. We can also express the exponential growth
of computing in terms of an accelerating pace: It took 90 years
to achieve the first Multiple in Power (MIP) per thousand dollars;
now we add a MIP per thousand dollars every day.
Moore’s Law narrowly refers to the number of transistors on
an integrated circuit of fixed size, and sometimes has been expressed
even more narrowly in terms of transistor feature size. But rather
than feature size (which is only one contributing factor), or even
number of transistors, I think the most salient measure to track
is computational speed per unit cost. This takes into account many
levels of “cleverness” (i.e., innovation, which is to
say technological evolution). In addition to all of the innovation
in integrated circuits, there are multiple layers of innovation
in computer design, e.g., pipelining, parallel processing, instruction
look-ahead, instruction and memory caching, etc.
From the above chart, we see that the exponential growth of computing
didn’t start with integrated circuits (around 1958), or even
transistors (around 1947), but goes back to the electromechanical
calculators used in the 1890 and 1900 U.S. census. This chart spans
at least five distinct paradigms of computing (electromechanical
calculators, relay-based computers, vacuum tube-based computers,
discrete transistor-based computers, and finally microprocessors),
of which Moore’s Law pertains to only the latest one.
It’s obvious what the sixth paradigm will be after Moore’s
Law runs out of steam before 2019 (because before then transistor
feature sizes will be just a few atoms in width). Chips today are
flat (although it does require up to 20 layers of material to produce
one layer of circuitry). Our brain, in contrast, is organized in
three dimensions. We live in a three-dimensional world; why not
use the third dimension? There are many technologies in the wings
that build circuitry in three dimensions. Nanotubes, for example,
which are already working in laboratories, build circuits from pentagonal
arrays of carbon atoms. One cubic inch of nanotube circuitry would
be a million times more powerful than the human brain.
Thus the (double) exponential growth of computing is broader than
Moore’s Law. And this accelerating growth of computing is,
in turn, part of a yet broader phenomenon discussed above, the accelerating
pace of any evolutionary process. In my book, I discuss the link
between the pace of a process and the degree of chaos versus order
in the process. For example, in cosmological history, the Universe
started with little chaos, so the first three major paradigm shifts
(the emergence of gravity, the emergence of matter, and the emergence
of the four fundamental forces) all occurred in the first billionth
of a second; now with vast chaos, cosmological paradigm shifts take
billions of years.
Observers are quick to criticize extrapolations of an exponential
trend on the basis that the trend is bound to run out of “resources.”
The classical example is when a species happens upon a new habitat
(e.g., rabbits in Australia), the species’ numbers will grow
exponentially for a time, but then hit a limit when resources such
as food and space run out. But the resources underlying the exponential
growth of an evolutionary process are relatively unbounded: (i)
the (ever growing) order of the evolutionary process itself, and
(ii) the chaos of the environment in which the evolutionary process
takes place and which provides the options for further diversity.
We also need to distinguish between the S curve (very slow virtually
unnoticeable growth followed by very rapid growth followed by the
growth leveling off and reaching an asymptote) that is characteristic
of any specific technological paradigm and the continuing exponential
growth that is characteristic of the ongoing evolutionary process
of technology. Specific paradigms, such as Moore’s Law (i.e.,
achieving faster and denser computation through shrinking transistor
sizes on an integrated circuit), do ultimately reach levels at which
exponential growth is no longer feasible. Thus Moore’s Law
is an S curve. But the growth of computation is an ongoing exponential.
What turns the S curve (of any specific paradigm) into a continuing
exponential is paradigm shift (also called innovation), in which
a new paradigm (e.g., three-dimensional circuits) takes over when
the old paradigm approaches its natural limit. This has already
happened at least four times in the history of computation. This
difference also distinguishes the toolmaking of non-human species,
in which the mastery of a toolmaking (or using) skill by each animal
is characterized by an S-shaped learning curve, and human-created
technology, which has been following an exponential pattern of growth
and acceleration since its inception.
I discuss all of this in more detail in the first couple of chapters
of my book.
In the sidebar below, I include a mathematical model of the law
of accelerating returns as it pertains to the exponential growth
of computing. The formulas below result in the following graph of
the continued growth of computation. This graph matches the available
data for the twentieth century and provides projections for the
twenty-first century. Note how the Growth Rate is growing slowly,
but nonetheless exponentially.
Another technology trend that will have important implications
for the twenty-first century is miniaturization. A related analysis
can be made of this trend which shows that the salient implementation
sizes of a broad range of technology are shrinking, also at a double
exponential rate. At present, we are shrinking technology by a factor
of approximately 5.6 per linear dimension per decade.
The following provides a brief overview of the law of accelerating
returns as it applies to the double exponential growth of computation.
This model considers the impact of the growing power of the technology
to foster its own next generation. For example, with more powerful
computers and related technology, we have the tools and the knowledge
to design yet more powerful computers, and to do so more quickly.
Note that the data for the year 2000 and beyond assume neural net
connection calculations as it is expected that this type of calculation
will dominate, particularly in emulating human brain functions.
This type of calculation is less expensive than conventional (e.g.,
Pentium III) calculations by a factor of 10 (particularly if implemented
using digital controlled analog electronics, which would correspond
well to the brain’s digital controlled analog electrochemical
processes). A factor of 10 translates into approximately 3 years
(today) and less than 3 years later in the twenty-first century.
My estimate of brain capacity is 100 billion neurons times an average
1,000 connections per neuron (with the calculations taking place
primarily in the connections) times 200 calculations per second.
Although these estimates are conservatively high, one can find higher
and lower estimates. However, even much higher (or lower) estimates
by orders of magnitude only shift the prediction by a relatively
small number of years.
Some salient dates from this analysis include the following:
We achieve one Human Brain capability (2 * 10^16 cps) for $1,000
around the year 2023.
We achieve one Human Brain capability (2 * 10^16 cps) for one cent
around the year 2037.
We achieve one Human Race capability (2 * 10^26 cps) for $1,000
around the year 2049.
We achieve one Human Race capability (2 * 10^26 cps) for one cent
around the year 2059.
The Model considers the following variables:
V: Velocity (i.e., power) of computing (measured in CPS/unit cost)
W: World Knowledge as it pertains to designing and building computational
The assumptions of the model are:
(1) V = C1 * W
In other words, computer power is a linear function of the knowledge
of how to build computers. This is actually a conservative assumption.
In general, innovations improve V (computer power) by a multiple,
not in an additive way. Independent innovations multiply each other’s
effect. For example, a circuit advance such as CMOS, a more efficient
IC wiring methodology, and a processor innovation such as pipelining
all increase V by independent multiples.
(2) W = C2 * Integral (0 to t) V
In other words, W (knowledge) is cumulative, and the instantaneous
increment to knowledge is proportional to V.
This gives us:
W = C1 * C2 * Integral (0 to t) W
W = C1 * C2 * C3 ^ (C4 * t)
V = C1 ^ 2 * C2 * C3 ^ (C4 * t)
(Note on notation: a^b means a raised to the b power.)
Simplifying the constants, we get:
V = Ca * Cb ^ (Cc * t)
So this is a formula for “accelerating” (i.e., exponentially
growing) returns, a “regular Moore’s Law.”
As I mentioned above, the data shows exponential growth in the
rate of exponential growth. (We doubled computer power every three
years early in the twentieth century, every two years in the middle
of the century, and close to every one year during the 1990s.)
Let’s factor in another exponential phenomenon, which is the
growing resources for computation. Not only is each (constant cost)
device getting more powerful as a function of W, but the resources
deployed for computation are also growing exponentially.
We now have:
N: Expenditures for computation
V = C1 * W (as before)
N = C4 ^ (C5 * t) (Expenditure for computation is growing at its
own exponential rate)
W = C2 * Integral (0 to t) (N * V)
As before, world knowledge is accumulating, and the instantaneous
increment is proportional to the amount of computation, which equals
the resources deployed for computation (N) * the power of each (constant
This gives us:
W = C1 * C2 * Integral(0 to t) (C4 ^ (C5 * t) * W)
W = C1 * C2 * (C3 ^ (C6 * t)) ^ (C7 * t)
V = C1 ^ 2 * C2 * (C3 ^ (C6 * t)) ^ (C7 * t)
Simplifying the constants, we get:
V = Ca * (Cb ^ (Cc * t)) ^ (Cd * t)
This is a double exponential—an exponential curve in which
the rate of exponential growth is growing at a different exponential
Now let’s consider real-world data. Considering the data for
actual calculating devices and computers during the twentieth century:
CPS/$1K: Calculations Per Second for $1,000
Twentieth century computing data matches:
CPS/$1K = 10^(6.00*((20.40/6.00)^((A13-1900)/100))-11.00)
We can determine the growth rate over a period of time:
Growth Rate =10^((LOG(CPS/$1K for Current Year)—LOG(CPS/$1K
for Previous Year))/(Current Year—Previous Year))
Human Brain = 100 Billion (10^11) neurons * 1000 (10^3) Connections/Neuron
* 200 (2 * 10^2) Calculations Per Second Per Connection = 2 * 10^16
Calculations Per Second
Human Race = 10 Billion (10^10) Human Brains = 2 * 10^26 Calculations
These formulas produce the graph below.
In a process, the time interval between salient events expands
or contracts along with the amount of chaos. This relationship is
one key to understanding the reason that the exponential growth
of computing will survive the demise of Moore’s Law. Evolution
started with vast chaos and little effective order, so early progress
was slow. But evolution creates ever-increasing order. That is,
after all, the essence of evolution. Order is the opposite of chaos,
so when order in a process increases—as is the case for evolution—time
speeds up. I call this important sub-law the “law of accelerating
returns,” to contrast it with a better known law in which returns
Computation represents the essence of order in technology. Being
subject to the evolutionary process that is technology, it too grows
exponentially. There are many examples of the exponential growth
in technological speeds and capacities. For example, when the human
genome scan started twelve years ago, genetic sequencing speeds
were so slow that without speed increases the project would have
required thousands of years, yet it is now completing on schedule
in under fifteen years. Other examples include the accelerating
price-performance of all forms of computer memory, the exponential
growing bandwidth of communication technologies (electronic, optical,
as well as wireless), the rapidly increasing speed and resolution
of human brain scanning, the miniaturization of technology, and
many others. If we view the exponential growth of computation in
its proper perspective, as one example of many of the law of accelerating
returns, then we can confidently predict its continuation.
A sixth paradigm will take over from Moore’s Law, just as
Moore’s Law took over from discrete transistors, and vacuum
tubes before that. There are many emerging technologies for new
computational substrates. In addition to nanotubes, several forms
of computing at the molecular level are working in laboratories.
There are more than enough new computing technologies now being
researched, including three-dimensional chips, optical computing,
crystalline computing, DNA computing, and quantum computing, to
keep the law of accelerating returns going for a long time.
So where will this take us?
IBM’s “Blue Gene” supercomputer, scheduled to be
completed by 2005, is projected to provide 1 million billion calculations
per second, already one-twentieth of the capacity of the human brain.
By the year 2019, your $1,000 personal computer will have the processing
power of the human brain—20 million billion calculations per
second (100 billion neurons times 1,000 connections per neuron times
200 calculations per second per connection). By 2029, it will take
a village of human brains (about a thousand) to match $1,000 of
computing. By 2050, $1,000 of computing will equal the processing
power of all human brains on Earth. Of course, this only includes
those brains still using carbon-based neurons. While human neurons
are wondrous creations in a way, we wouldn’t design computing
circuits the same way. Our electronic circuits are already more
than 10 million times faster than a neuron’s electrochemical
processes. Most of the complexity of a human neuron is devoted to
maintaining its life support functions, not its information processing
capabilities. Ultimately, we will need to port our mental processes
to a more suitable computational substrate. Then our minds won’t
have to stay so small, being constrained as they are today to a
mere hundred trillion neural connections each operating at a ponderous
200 digitally controlled analog calculations per second.
A careful consideration of the law of time and chaos, and its key
sublaw, the law of accelerating returns, shows that the exponential
growth of computing is not like those other exponential trends that
run out of resources. The two resources it needs—the growing
order of the evolving technology itself, and the chaos from which
an evolutionary process draws its options for further diversity—are
without practical limits, at least not limits that we will encounter
in the 21st century.
Many long range forecasts of technical feasibility in future time
periods dramatically underestimate the power of future technology
because they are based on what I call the “intuitive linear”
view of technological progress rather than the “historical
exponential view.” To express this another way, it is not the
case that we will experience a hundred years of progress in the
twenty-first century; rather we will witness on the order of twenty
thousand years of progress (from the linear perspective, that is).
When people think of a future period, they intuitively assume that
the current rate of progress will continue for the period being
considered. However, careful consideration of the pace of technology
shows that the rate of progress is not constant, but it is human
nature to adapt to the changing pace, so the intuitive view is that
the pace will continue at the current rate. Even for those of us
who have lived through a sufficiently long period of technological
progress to experience how the pace increases over time, our unexamined
intuition nonetheless provides the impression that progress changes
at the rate that we have experienced recently. A salient reason
for this is that an exponential curve approximates a straight line
when viewed for a brief duration. So even though the rate of progress
in the very recent past (e.g., this past year) is far greater than
it was ten years ago (let alone a hundred or a thousand years ago),
our memories are nonetheless dominated by our very recent experience.
Since the rate has not changed significantly in the very recent
past (because a very small piece of an exponential curve is approximately
straight), it is an understandable misperception to view the pace
of change as a constant. It is typical, therefore, that even sophisticated
commentators, when considering the future, extrapolate the current
pace of change over the next ten years or hundred years to determine
their expectations. This is why I call this way of looking at the
future the “intuitive linear” view.
But any serious consideration of the history of technology shows
that technological change is at least exponential, not linear. There
are a great many examples of this which I have discussed above.
One can examine this data in many different ways, and on many different
time scales, and for a wide variety of different phenomena, and
the (at least) double exponential growth implied by the law of accelerating
returns applies. The law of accelerating returns does not rely on
an assumption of the continuation of Moore’s law, but is based
on a rich model of diverse technological processes. What it clearly
shows is that technology, particularly the pace of technological
change, advances (at least) exponentially, not linearly, and has
been doing so since the advent of technology, indeed since the advent
of evolution on Earth.
Most technology forecasts ignore altogether this “historical
exponential view” of technological progress and assume instead
the “intuitive linear view.” Although the evidence is
compelling, it still requires study and modeling of many diverse
events to see this exponential aspect. That is why people tend to
overestimate what can be achieved in the short term (because we
tend to leave out necessary details), but underestimate what can
be achieved in the long term (because the exponential growth is
This observation also applies to paradigm shift rates, which are
currently doubling (approximately) every decade; that is paradigm
shift times are halving every decade (and this rate is also changing
slowly, but nonetheless exponentially). So, the technological progress
in the twenty-first century will be equivalent to what would require
(in the linear view) on the order of twenty thousand years. In terms
of the growth of computing, the comparison is even more dramatic.
So far, I’ve been talking about the hardware of computing.
The software is even more salient. Achieving the computational capacity
of the human brain, or even villages and nations of human brains
will not automatically produce human levels of capability. It is
a necessary but not sufficient condition. The organization and content
of these resources—the software of intelligence—is also
There are a number of compelling scenarios to capture higher levels
of intelligence in our computers, and ultimately human levels and
beyond. We will be able to evolve and train a system combining massively
parallel neural nets with other paradigms to understand language
and model knowledge, including the ability to read and model the
knowledge contained in written documents. Unlike many contemporary
“neural net” machines, which use mathematically simplified
models of human neurons, more advanced neural nets are already using
highly detailed models of human neurons, including detailed nonlinear
analog activation functions and other salient details. Although
the ability of today’s computers to extract and learn knowledge
from natural language documents is limited, their capabilities in
this domain are improving rapidly. Computers will be able to read
on their own, understanding and modeling what they have read, by
the second decade of the twenty-first century. We can then have
our computers read all of the world’s literature—books,
magazines, scientific journals, and other available material. Ultimately,
the machines will gather knowledge on their own by venturing into
the physical world, drawing from the full spectrum of media and
information services, and sharing knowledge with each other (which
machines can do far more easily than their human creators).
Once a computer achieves a human level of intelligence, it will
necessarily soar past it. Since their inception, computers have
significantly exceeded human mental dexterity in their ability to
remember and process information. A computer can remember billions
or even trillions of facts perfectly, while we are hard pressed
to remember a handful of phone numbers. A computer can quickly search
a data base with billions of records in fractions of a second. As
I mentioned earlier, computers can readily share their knowledge.
The combination of human level intelligence in a machine with a
computer’s inherent superiority in the speed, accuracy and
sharing ability of its memory will be formidable.
The most compelling scenario for mastering the software of intelligence
is to tap into the blueprint of the best example we can get our
hands on of an intelligent process. There is no reason why we cannot
reverse engineer the human brain, and essentially copy its design.
It took its original designer several billion years to develop.
And it’s not even copyrighted.
The most immediately accessible way to accomplish this is through
destructive scanning: we take a frozen brain, preferably one frozen
just slightly before rather than slightly after it was going to
die anyway, and examine one brain layer—one very thin slice—at
a time. We can readily see every neuron and every connection and
every neurotransmitter concentration represented in each synapse-thin
Human brain scanning has already started. A condemned killer allowed
his brain and body to be scanned and you can access all 10 billion
bytes of him on the Internet. He has a 25 billion byte female companion
on the site as well in case he gets lonely. This scan is not high
enough resolution for our purposes, but then we probably don’t
want to base our templates of machine intelligence on the brain
of a convicted killer, anyway.
But scanning a frozen brain is feasible today, albeit not yet at
a sufficient speed or bandwidth, but again, the law of accelerating
returns will provide the requisite speed of scanning, just as it
did for the human genome scan.
We also have noninvasive scanning techniques today, including high-resolution
magnetic resonance imaging (MRI) scans, optical imaging, near-infrared
scanning, and other noninvasive scanning technologies, that are
capable in certain instances of resolving individual somas, or neuron
cell bodies. Brain scanning technologies are increasing their resolution
with each new generation, just what we would expect from the law
of accelerating returns. Future generations will enable us to resolve
the connections between neurons, and to peer inside the synapses
and record the neurotransmitter concentrations.
We can peer inside someone’s brain today with noninvasive
scanners, which are increasing their resolution with each new generation
of this technology. There are a number of technical challenges in
accomplishing this, including achieving suitable resolution, bandwidth,
lack of vibration, and safety. For a variety of reasons it is easier
to scan the brain of someone recently deceased than of someone still
living. It is easier to get someone deceased to sit still, for one
thing. But noninvasively scanning a living brain will ultimately
become feasible as MRI, optical, and other scanning technologies
continue to improve in resolution and speed.
In fact, the driving force behind the rapidly improving capability
of noninvasive scanning technologies is again the law of accelerating
returns, because it requires massive computational ability to build
the high-resolution three-dimensional images. The exponentially
increasing computational ability provided by the law of accelerating
returns (and for another 10 to 20 years, Moore’s Law) will
enable us to continue to rapidly improve the resolution and speed
of these scanning technologies.
Scanning from Inside
To capture every salient neural detail of the human brain, the
most practical approach will be to scan it from inside. By 2030,
“nanobot” (i.e., nano-robot) technology will be viable,
and brain scanning will be a prominent application. Nanobots are
robots that are the size of human blood cells, or even smaller.
Billions of them could travel through every brain capillary and
scan every salient neural detail from up close. Using high-speed
wireless communication, the nanobots would communicate with each
other, and with other computers that are compiling the brain scan
database (in other words, the nanobots will all be on a wireless
local area network).
This scenario involves only capabilities we can touch and feel
today. We already have technology capable of producing very high-resolution
scans provided that the scanner is physically proximate to the neural
features. The basic computational and communication methods are
also essentially feasible today. The primary features that are not
yet practical are nanobot size and cost. As I discussed above, we
can project the exponentially declining cost of computation. Miniaturization
is another readily predictable aspect of the law of accelerating
returns. Already being developed at the University of California
at Berkeley are tiny flying robots called “smart dust,”
which are approximately one millimeter wide (about the size of a
grain of sand), and capable of flying, sensing, computing, and communicating
using tiny lasers and hinged micro-flaps and micro-mirrors. The
size of electronics and robotics will continue to shrink at an exponential
rate, currently by a factor of 5.6 per linear dimension per decade.
We can conservatively expect, therefore, the requisite nanobot technology
by around 2030. Because of its ability to place each scanner in
very close physical proximity to every neural feature, nanobot-based
scanning will be more practical than scanning the brain from outside.
How will we apply the thousands of trillions of bytes of information
derived from each brain scan? One approach is to use the results
to design more intelligent parallel algorithms for our machines,
particularly those based on one of the neural net paradigms. With
this approach, we don’t have to copy every single connection.
There is a great deal of repetition and redundancy within any particular
brain region. Although the information contained in a human brain
would require thousands of trillions of bytes of information (on
the order of 100 billion neurons times an average of 1,000 connections
per neuron, each with multiple neurotransmitter concentrations and
connection data), the design of the brain is characterized by a
human genome of only about a billion bytes.
Furthermore, most of the genome is redundant, so the initial design
of the brain is characterized by approximately one hundred million
bytes, about the size of Microsoft Word. Of course, the complexity
of our brains greatly increases as we interact with the world (by
a factor of more than ten million). It is not necessary, however,
to capture each detail in order to reverse engineer the salient
digital-analog algorithms. With this information, we can design
simulated nets that operate similarly. There are already multiple
efforts under way to scan the human brain and apply the insights
derived to the design of intelligent machines. The ATR (Advanced
Telecommunications Research) Lab in Kyoto, Japan, for example, is
building a silicon brain with 1 billion neurons. Although this is
1% of the number of neurons in a human brain, the ATR neurons operate
at much faster speeds.
After the algorithms of a region are understood, they can be refined
and extended before being implemented in synthetic neural equivalents.
For one thing, they can be run on a computational substrate that
is already more than ten million times faster than neural circuitry.
And we can also throw in the methods for building intelligent machines
that we already understand.
Perhaps a more interesting approach than this scanning-the-brain-to-understand-it
scenario is scanning-the-brain-to-download-it. Here we scan someone’s
brain to map the locations, interconnections, and contents of all
the somas, axons, dendrites, presynaptic vesicles, neurotransmitter
concentrations, and other neural components and levels. Its entire
organization can then be re-created on a neural computer of sufficient
capacity, including the contents of its memory.
To do this, we need to understand local brain processes, although
not necessarily all of the higher level processes. Scanning a brain
with sufficient detail to download it may sound daunting, but so
did the human genome scan. All of the basic technologies exist today,
just not with the requisite speed, cost, and size, but these are
the attributes that are improving at a double exponential pace.
The computationally salient aspects of individual neurons are complicated,
but definitely not beyond our ability to accurately model. For example,
Ted Berger and his colleagues at Hedco Neurosciences have built
integrated circuits that precisely match the digital and analog
information processing characteristics of neurons, including clusters
with hundreds of neurons. Carver Mead and his colleagues at CalTech
have built a variety of integrated circuits that emulate the digital-analog
characteristics of mammalian neural circuits.
A recent experiment at San Diego’s Institute for Nonlinear
Science demonstrates the potential for electronic neurons to precisely
emulate biological ones. Neurons (biological or otherwise) are a
prime example of what is often called “chaotic computing.”
Each neuron acts in an essentially unpredictable fashion. When an
entire network of neurons receives input (from the outside world
or from other networks of neurons), the signaling amongst them appears
at first to be frenzied and random. Over time, typically a fraction
of a second or so, the chaotic interplay of the neurons dies down,
and a stable pattern emerges. This pattern represents the “decision”
of the neural network. If the neural network is performing a pattern
recognition task (which, incidentally, comprises more than 95% of
the activity in the human brain), then the emergent pattern represents
the appropriate recognition.
So the question addressed by the San Diego researchers was whether
electronic neurons could engage in this chaotic dance alongside
biological ones. They hooked up their artificial neurons with those
from spiney lobsters in a single network, and their hybrid biological-nonbiological
network performed in the same way (i.e., chaotic interplay followed
by a stable emergent pattern) and with the same type of results
as an all biological net of neurons. Essentially, the biological
neurons accepted their electronic peers. It indicates that their
mathematical model of these neurons was reasonably accurate.
There are many projects around the world, which are creating nonbiological
devices and which recreate in great detail the functionality of
human neuron clusters, and the accuracy and scale of these neuron
clusters replications are rapidly increasing. We started with functionally
equivalent recreations of single neurons, then clusters of tens,
then hundreds, and now thousands. Scaling up technical processes
at an exponential pace is what technology is good at.
As the computational power to emulate the human brain becomes available—we’re
not there yet, but we will be there within a couple of decades—projects
already under way to scan the human brain will be accelerated, with
a view both to understand the human brain in general, as well as
providing a detailed description of the contents and design of specific
brains. By the third decade of the twenty-first century, we will
be in a position to create highly detailed and complete maps of
all relevant features of all neurons, neural connections and synapses
in the human brain, all of the neural details that play a role in
the behavior and functionality of the brain, and to recreate these
designs in suitably advanced neural computers.
The answer depends on what we mean by the word “computer.”
Certainly the brain uses very different methods from conventional
contemporary computers. Most computers today are all digital and
perform one (or perhaps a few) computation(s) at a time at extremely
high speed. In contrast, the human brain combines digital and analog
methods with most computations performed in the analog domain. The
brain is massively parallel, performing on the order of a hundred
trillion computations at the same time, but at extremely slow speeds.
With regard to digital versus analog computing, we know that digital
computing can be functionally equivalent to analog computing (although
the reverse is not true), so we can perform all of the capabilities
of a hybrid digital—analog network with an all digital computer.
On the other hand, there is an engineering advantage to analog circuits
in that analog computing is potentially thousands of times more
efficient. An analog computation can be performed by a few transistors,
or, in the case of mammalian neurons, specific electrochemical processes.
A digital computation in contrast requires thousands or tens of
thousands of transistors. So there is a significant efficiency advantage
to emulating the brain’s analog methods.
The massive parallelism of the human brain is the key to its pattern
recognition abilities, which reflects the strength of human thinking.
As I discussed above, mammalian neurons engage in a chaotic dance,
and if the neural network has learned its lessons well, then a stable
pattern will emerge reflecting the network’s decision. There
is no reason why our nonbiological functionally-equivalent recreations
of biological neural networks cannot be built using these same principles,
and indeed there are dozens of projects around the world that have
succeeded in doing this. My own technical field is pattern recognition,
and the projects that I have been involved in for over thirty years
use this form of chaotic computing. Particularly successful examples
are Carver Mead’s neural chips, which are highly parallel,
use digital controlled analog computing, and are intended as functionally
similar recreations of biological networks.
As we create nonbiological but functionally equivalent recreations
of biological neural networks ranging from clusters of dozens of
neurons up to entire human brains and beyond, we can combine the
qualities of human thinking with certain advantages of machine intelligence.
My human knowledge and skills exist in my brain as vast patterns
of interneuronal connections, neurotransmitter concentrations, and
other neural elements. As I mentioned at the beginning of this chapter,
there are no quick downloading ports for these patterns in our biological
neural networks, but as we build nonbiological equivalents, we will
not leave out the ability to quickly load patterns representing
knowledge and skills.
Although it is remarkable that as complex and capable an entity
as the human brain evolved through natural selection, aspects of
its design are nonetheless extremely inefficient. Neurons are very
bulky devices and at least ten million times slower in their information
processing than electronic circuits. As we combine the brain’s
pattern recognition methods derived from high-resolution brain scans
and reverse engineering efforts with the knowledge sharing, speed,
and memory accuracy advantages of nonbiological intelligence, the
combination will be formidable.
Objective and Subjective
Although I anticipate that the most common application of the knowledge
gained from reverse engineering the human brain will be creating
more intelligent machines that are not necessarily modeled on specific
individuals, the scenario of scanning and reinstantiating all of
the neural details of a particular person raises the most immediate
questions of identity. Let’s consider the question of what
we will find when we do this.
We have to consider this question on both the objective and subjective
levels. “Objective” means everyone except me, so let’s
start with that. Objectively, when we scan someone’s brain
and reinstantiate their personal mind file into a suitable computing
medium, the newly emergent “person” will appear to other
observers to have very much the same personality, history, and memory
as the person originally scanned. That is, once the technology has
been refined and perfected. Like any new technology, it won’t
be perfect at first. But ultimately, the scans and recreations will
be very accurate and realistic.
Interacting with the newly instantiated person will feel like interacting
with the original person. The new person will claim to be that same
old person and will have a memory of having been that person. The
new person will have all of the patterns of knowledge, skill, and
personality of the original. We are already creating functionally
equivalent recreations of neurons and neuron clusters with sufficient
accuracy that biological neurons accept their nonbiological equivalents
and work with them as if they were biological. There are no natural
limits that prevent us from doing the same with the hundred billion
neuron cluster we call the human brain.
Subjectively, the issue is more subtle and profound, but first
we need to reflect on one additional objective issue: our physical
The Importance of Having a Body
Consider how many of our thoughts and thinking is directed towards
our body and its survival, security, nutrition, image, not to mention
affection, sexuality, and reproduction. Many, if not most, of the
goals we attempt to advance using our brains have to do with our
bodies: protecting them, providing them with fuel, making them attractive,
making them feel good, providing for their myriad needs and desires.
Some philosophers maintain that achieving human level intelligence
is impossible without a body. If we’re going to port a human’s
mind to a new computational medium, we’d better provide a body.
A disembodied mind will quickly get depressed.
There are a variety of bodies that we will provide for our machines,
and that they will provide for themselves: bodies built through
nanotechnology (an emerging field devoted to building highly complex
physical entities atom by atom), virtual bodies (that exist only
in virtual reality), bodies comprised of swarms of nanobots.
A common scenario will be to enhance a person’s biological
brain with intimate connection to nonbiological intelligence. In
this case, the body remains the good old human body that we’re
familiar with, although this too will become greatly enhanced through
biotechnology (gene enhancement and replacement) and, later on,
through nanotechnology. A detailed examination of twenty-first century
bodies is beyond the scope of this chapter, but is examined in chapter
seven of my recent book The Age of Spiritual Machines.
So Just Who are These People?
To return to the issue of subjectivity, consider: Is the reinstantiated
mind the same consciousness as the person we just scanned? Are these
“people” conscious at all? Is this a mind or just a brain?
Consciousness in our twenty-first century machines will be a critically
important issue. But it is not easily resolved, or even readily
understood. People tend to have strong views on the subject, and
often just can’t understand how anyone else could possibly
see the issue from a different perspective. Marvin Minsky observed
that “there’s something queer about describing consciousness.
Whatever people mean to say, they just can’t seem to make it
We don’t worry, at least not yet, about causing pain and suffering
to our computer programs. But at what point do we consider an entity,
a process, to be conscious, to feel pain and discomfort, to have
its own intentionality, its own free will? How do we determine if
an entity is conscious; if it has subjective experience? How do
we distinguish a process that is conscious from one that just acts
as if it is conscious?
We can’t simply ask it. If it says, “Hey I’m conscious,”
does that settle the issue? No, we have computer games today that
effectively do that, and they’re not terribly convincing.
How about if the entity is very convincing and compelling when
it says, “I’m lonely, please keep me company”? Does
that settle the issue?
If we look inside its circuits, and see essentially the identical
kinds of feedback loops and other mechanisms in its brain that we
see in a human brain (albeit implemented using nonbiological equivalents),
does that settle the issue?
And just who are these people in the machine, anyway? The answer
will depend on who you ask. If you ask the people in the machine,
they will strenuously claim to be the original persons. For example,
if we scan—let’s say myself—and record the exact
state, level, and position of every neurotransmitter, synapse, neural
connection, and every other relevant detail, and then reinstantiate
this massive data base of information (which I estimate at thousands
of trillions of bytes) into a neural computer of sufficient capacity,
the person that then emerges in the machine will think that he is
(and had been) me. He will say “I grew up in Queens, New York,
went to college at MIT, stayed in the Boston area, sold a few artificial
intelligence companies, walked into a scanner there, and woke up
in the machine here. Hey, this technology really works.”
But wait. Is this really me? For one thing, old biological Ray
(that’s me) still exists. I’ll still be here in my carbon-cell-based
brain. Alas, I will have to sit back and watch the new Ray succeed
in endeavors that I could only dream of.
Let’s consider the issue of just who I am, and who the new
Ray is a little more carefully. First of all, am I the stuff in
my brain and body?
Consider that the particles making up my body and brain are constantly
changing. We are not at all permanent collections of particles.
The cells in our bodies turn over at different rates, but the particles
(e.g., atoms and molecules) that comprise our cells are exchanged
at a very rapid rate. I am just not the same collection of particles
that I was even a month ago. It is the patterns of matter and energy
that are semipermanent (that is, changing only gradually), but our
actual material content is changing constantly, and very quickly.
We are rather like the patterns that water makes in a stream. The
rushing water around a formation of rocks makes a particular, unique
pattern. This pattern may remain relatively unchanged for hours,
even years. Of course, the actual material constituting the pattern—the
water—is replaced in milliseconds. The same is true for Ray
Kurzweil. Like the water in a stream, my particles are constantly
changing, but the pattern that people recognize as Ray has a reasonable
level of continuity. This argues that we should not associate our
fundamental identity with a specific set of particles, but rather
the pattern of matter and energy that we represent. Many contemporary
philosophers seem partial to this “identify from pattern”
But wait. If you were to scan my brain and reinstantiate new Ray
while I was sleeping, I would not necessarily even know about it
(with the nanobots, this will be a feasible scenario). If you then
come to me, and say, “Good news, Ray, we’ve successfully
reinstantiated your mind file, so we won’t be needing your
old brain anymore,” I may suddenly realize the flaw in the
“identity from pattern” argument. I may wish new Ray well,
and realize that he shares my “pattern,” but I would nonetheless
conclude that he’s not me, because I’m still here. How
could he be me? After all, I would not necessarily know that he
Let’s consider another perplexing scenario. Suppose I replace
a small number of biological neurons with functionally equivalent
nonbiological ones (they may provide certain benefits such as greater
reliability and longevity, but that’s not relevant to this
thought experiment). After I have this procedure performed, am I
still the same person? My friends certainly think so. I still have
the same self-deprecating humor, the same silly grin—yes, I’m
still the same guy.
It should be clear where I’m going with this. Bit by bit,
region by region, I ultimately replace my entire brain with essentially
identical (perhaps improved) nonbiological equivalents (preserving
all of the neurotransmitter concentrations and other details that
represent my learning, skills, and memories). At each point, I feel
the procedures were successful. At each point, I feel that I am
same guy. After each procedure, I claim to be the same guy. My friends
concur. There is no old Ray and new Ray, just one Ray, one that
never appears to fundamentally change.
But consider this. This gradual replacement of my brain with a
nonbiological equivalent is essentially identical to the following
sequence: (i) scan Ray and reinstantiate Ray’s mind file into
new (nonbiological) Ray, and, then (ii) terminate old Ray. But we
concluded above that in such a scenario new Ray is not the same
as old Ray. And if old Ray is terminated, well then that’s
the end of Ray. So the gradual replacement scenario essentially
results in the same result: New Ray has been created, and old Ray
has been terminated, even if we never saw him missing. So what appears
to be the continuing existence of just one Ray is really the creation
of new Ray and the end of old Ray.
On yet another hand (we’re running out of philosophical hands
here), the gradual replacement scenario is not altogether different
from what happens normally to our biological selves, in that our
particles are always rapidly being replaced. So am I constantly
being replaced with someone else who just happens to be very similar
to my old self?
I am trying to illustrate why consciousness is not an easy issue.
If we talk about consciousness as just a certain type of intelligent
skill: the ability to reflect on one’s own self and situation,
for example, then the issue is not difficult at all because any
skill or capability or form of intelligence that one cares to define
will be replicated in nonbiological entities (i.e., machines) within
a few decades. With this type of objective view of consciousness,
the conundrums do go away. But a fully objective view does not penetrate
to the core of the issue, because the essence of consciousness is
subjective experience, not objective correlates of that experience.
Will these future machines be capable of having spiritual experiences?
They certainly will claim to. They will claim to be people, and
to have the full range of emotional and spiritual experiences that
people claim to have. And these will not be idle claims; they will
evidence the sort of rich, complex, and subtle behavior one associates
with these feelings. How do the claims and behaviors—compelling
as they will be—relate to the subjective experience of these
reinstantiated people? We keep coming back to the very real but
ultimately unmeasurable issue of consciousness.
People often talk about consciousness as if it were a clear property
of an entity that can readily be identified, detected, and gauged.
If there is one crucial insight that we can make regarding why the
issue of consciousness is so contentious, it is the following:
There exists no objective test that can absolutely determine its
Science is about objective measurement and logical implications
therefrom, but the very nature of objectivity is that you cannot
measure subjective experience—you can only measure correlates
of it, such as behavior (and by behavior, I include the actions
of components of an entity, such as neurons). This limitation has
to do with the very nature of the concepts “objective”
and “subjective.” Fundamentally, we cannot penetrate the
subjective experience of another entity with direct objective measurement.
We can certainly make arguments about it: i.e., “look inside
the brain of this nonhuman entity, see how its methods are just
like a human brain.” Or, “see how its behavior is just
like human behavior.” But in the end, these remain just arguments.
No matter how convincing the behavior of a reinstantiated person,
some observers will refuse to accept the consciousness of an entity
unless it squirts neurotransmitters, or is based on DNA-guided protein
synthesis, or has some other specific biologically human attribute.
We assume that other humans are conscious, but that is still an
assumption, and there is no consensus amongst humans about the consciousness
of nonhuman entities, such as other higher non-human animals. The
issue will be even more contentious with regard to future nonbiological
entities with human-like behavior and intelligence.
From a practical perspective, we’ll accept their claims. Keep
in mind that nonbiological entities in the twenty-first century
will be extremely intelligent, so they’ll be able to convince
us that they are conscious. They’ll have all the subtle cues
that convince us today that humans are conscious. They will be able
to make us laugh and cry. And they’ll get mad if we don’t
accept their claims. But this is a political prediction, not a philosophical
Over the past several years, Roger Penrose, a noted physicist and
philosopher, has suggested that fine structures in the neurons called
tubules perform an exotic form of computation called “quantum
computing.” Quantum computing is computing using what are known
as “qu bits,” which take on all possible combinations
of solutions simultaneously. It can be considered to be an extreme
form of parallel processing (because every combination of values
of the qu bits is tested simultaneously). Penrose suggests that
the tubules and their quantum computing capabilities complicate
the concept of recreating neurons and reinstantiating mind files.
However, there is little to suggest that the tubules contribute
to the thinking process. Even generous models of human knowledge
and capability are more than accounted for by current estimates
of brain size, based on contemporary models of neuron functioning
that do not include tubules. In fact, even with these tubule-less
models, it appears that the brain is conservatively designed with
many more connections (by several orders of magnitude) than it needs
for its capabilities and capacity. Recent experiments (e.g., the
San Diego Institute for Nonlinear Science experiments) showing that
hybrid biological-nonbiological networks perform similarly to all
biological networks, while not definitive, are strongly suggestive
that our tubule-less models of neuron functioning are adequate.
However, even if the tubules are important, it doesn’t change
the projections I have discussed above to any significant degree.
According to my model of computational growth, if the tubules multiplied
neuron complexity by a factor of a thousand (and keep in mind that
our current tubule-less neuron models are already complex, including
on the order of a thousand connections per neuron, multiple nonlinearities
and other details), this would delay our reaching brain capacity
by only about nine years. If we’re off by a factor of a million,
that’s still only a delay of 17 years. A factor of a billion
is around 24 years (keep in mind computation is growing by a double
With regard to quantum computing, once again there is nothing to
suggest that the brain does quantum computing. Just because quantum
technology may be feasible does not suggest that the brain is capable
of it. We don’t have lasers or even radios in our brains either.
No one has suggested human capabilities that would require a capacity
for quantum computing.
However, even if the brain does do quantum computing, this does
not significantly change the outlook for human-level computing (and
beyond) nor does it suggest that brain downloading is infeasible.
First of all, if the brain does do quantum computing this would
only verify that quantum computing is feasible. There would be nothing
in such a finding to suggest that quantum computing is restricted
to biological mechanisms. Biological quantum computing mechanisms,
if they exist, could be replicated. Indeed, recent experiments with
small-scale quantum computers appear to be successful.
Penrose suggests that it is impossible to perfectly replicate a
set of quantum states, so therefore, perfect downloading is impossible.
Well, how perfect does a download have to be? I am at this moment
in a very different quantum state (and different in non-quantum
ways as well) than I was a minute ago (certainly in a different
state than I was before I wrote this paragraph). If we develop downloading
technology to the point where the “copies” are as close
to the original as the original person changes anyway in the course
of one minute, that would be good enough for any conceivable purpose,
yet does not require copying quantum states. As the technology improves,
the accuracy of the copy could become as close as the original changes
within ever-briefer periods of time (e.g., one second, one millisecond).
When it was pointed out to Penrose that neurons (and even neural
connections) were too big for quantum computing, he came up with
the tubule theory as a possible mechanism for neural quantum computing.
So the concerns with quantum computing and tubules have been introduced
together. If one is searching for barriers to replicating brain
function, it is an ingenious theory, but it fails to introduce any
genuine barriers. There is no evidence for it, and even if true,
it only delays matters by a decade or two. There is no reason to
believe that biological mechanisms (including quantum computing)
are inherently impossible to replicate using nonbiological materials
and mechanisms. Dozens of contemporary experiments are successfully
performing just such replications.
The Noninvasive Surgery-Free Reversible Programmable Distributed
How will we apply technology that is more intelligent than its
creators? One might be tempted to respond “Carefully!”
But let’s take a look at some examples.
Consider several examples of the nanobot technology, which, based
on miniaturization and cost reduction trends, will be feasible within
30 years. In addition to scanning your brain, the nanobots will
also be able to expand your brain.
Nanobot technology will provide fully immersive, totally convincing
virtual reality in the following way. The nanobots take up positions
in close physical proximity to every interneuronal connection coming
from all of our senses (e.g., eyes, ears, skin). We already have
the technology for electronic devices to communicate with neurons
in both directions that requires no direct physical contact with
the neurons. For example, scientists at the Max Planck Institute
have developed “neuron transistors” that can detect the
firing of a nearby neuron, or alternatively, can cause a nearby
neuron to fire, or suppress it from firing. This amounts to two-way
communication between neurons and the electronic-based neuron transistors.
The Institute scientists demonstrated their invention by controlling
the movement of a living leech from their computer. Again, the primary
aspect of nanobot-based virtual reality that is not yet feasible
is size and cost.
When we want to experience real reality, the nanobots just stay
in position (in the capillaries) and do nothing. If we want to enter
virtual reality, they suppress all of the inputs coming from the
real senses, and replace them with the signals that would be appropriate
for the virtual environment. You (i.e., your brain) could decide
to cause your muscles and limbs to move as you normally would, but
the nanobots again intercept these interneuronal signals, suppress
your real limbs from moving, and instead cause your virtual limbs
to move and provide the appropriate movement and reorientation in
the virtual environment.
The web will provide a panoply of virtual environments to explore.
Some will be recreations of real places; others will be fanciful
environments that have no “real” counterpart. Some indeed
would be impossible in the physical world (perhaps, because they
violate the laws of physics). We will be able to “go”
to these virtual environments by ourselves, or we will meet other
people there, both real people and simulated people. Of course,
ultimately there won’t be a clear distinction between the two.
Nanobot technology will be able to expand our minds in virtually
any imaginable way. Our brains today are relatively fixed in design.
Although we do add patterns of interneuronal connections and neurotransmitter
concentrations as a normal part of the learning process, the current
overall capacity of the human brain is highly constrained, restricted
to a mere hundred trillion connections. Brain implants based on
massively distributed intelligent nanobots will ultimately expand
our memories a trillion fold, and otherwise vastly improve all of
our sensory, pattern recognition and cognitive abilities. Since
the nanobots are communicating with each other over a wireless local
area network, they can create any set of new neural connections,
can break existing connections (by suppressing neural firing), can
create new hybrid biological-nonbiological networks as well as adding
vast new nonbiological networks.
Using nanobots as brain extenders is a significant improvement
over the idea of surgically installed neural implants, which are
beginning to be used today. Nanobots will be introduced without
surgery, essentially just by injecting or even swallowing them.
They can all be directed to leave, so the process is easily reversible.
They are programmable, in that they can provide virtual reality
one minute, and a variety of brain extensions the next. They can
change their configuration, and clearly can alter their software.
Perhaps most importantly, they are massively distributed and therefore
can take up billions or trillions of positions throughout the brain,
whereas a surgically introduced neural implant can only be placed
in one or at most a few locations.
A Clear and Future Danger
Technology has always been a double-edged sword, bringing us longer
and healthier life spans, freedom from physical and mental drudgery,
and many new creative possibilities on the one hand, while introducing
new and salient dangers on the other. We still live today with sufficient
nuclear weapons (not all of which appear to be well accounted for)
to end all mammalian life on the planet. Bioengineering is in the
early stages of enormous strides in reversing disease and aging
processes. However, the means and knowledge exist in a routine college
bioengineering lab to create unfriendly pathogens more dangerous
than nuclear weapons. For the twenty-first century, we will see
the same intertwined potentials: a great feast of creativity resulting
from human intelligence expanded a trillion-fold combined with many
grave new dangers.
Consider unrestrained nanobot replication. Nanobot technology requires
billions or trillions of such intelligent devices to be useful.
The most cost-effective way to scale up to such levels is through
self-replication, essentially the same approach used in the biological
world. And in the same way that biological self-replication gone
awry (i.e., cancer) results in biological destruction, a defect
in the mechanism curtailing nanobot self-replication would endanger
all physical entities, biological or otherwise.
Other salient concerns include “who is controlling the nanobots?”
and “who are the nanobots talking to?” Organizations (e.g.,
governments, extremist groups) or just a clever individual could
put trillions of undetectable nanobots in the water or food supply
of an individual or of an entire population. These “spy”
nanobots could then monitor, influence, and even control our thoughts
and actions. In addition to introducing physical spy nanobots, existing
nanobots could be influenced through software viruses and other
software “hacking” techniques.
My own expectation is that the creative and constructive applications
of this technology will dominate, as I believe they do today. But
there will be a valuable (and increasingly vocal) role for a concerned
and constructive Luddite movement (i.e., anti-technologists inspired
by early nineteenth century weavers who destroyed labor-saving machinery
Once brain porting technology has been refined and fully developed,
will this enable us to live forever? The answer depends on what
we mean by living and dying. Consider what we do today with our
personal computer files. When we change from one personal computer
to a less obsolete model, we don’t throw all our files away;
rather we copy them over to the new hardware. Although our software
files do not necessary continue their existence forever, the longevity
of our personal computer software is completely separate and disconnected
from the hardware that it runs on. When it comes to our personal
mind file, however, when our human hardware crashes, the software
of our lives dies with it. However, this will not continue to be
the case when we have the means to store and restore the thousands
of trillions of bytes of information stored and represented in our
The longevity of one’s mind file will not be dependent, therefore,
on the continued viability of any particular hardware medium. Ultimately
software-based humans, albeit vastly extended beyond the severe
limitations of humans as we know them today, will live out on the
web, projecting bodies whenever they need or want them, including
virtual bodies in diverse realms of virtual reality, holographically
projected bodies, and physical bodies comprised of nanobot swarms,
and other forms of nanotechnology.
A software-based human will be free, therefore, from the constraints
of any particular thinking medium. Today, we are each confined to
a mere hundred trillion connections, but humans at the end of the
twenty-first century can grow their thinking and thoughts without
limit. We may regard this as a form of immortality, although it
is worth pointing out that data and information do not necessarily
last forever. Although not dependent on the viability of the hardware
it runs on, the longevity of information depends on its relevance,
utility, and accessibility. If you’ve ever tried to retrieve
information from an obsolete form of data storage in an old obscure
format (e.g., a reel of magnetic tape from a 1970 minicomputer),
you will understand the challenges in keeping software viable. However,
if we are diligent in maintaining our mind file, keeping current
backups, and porting to current formats and mediums, then a form
of immortality can be attained, at least for software-based humans.
Our mind file—our personality, skills, memories—all of
that is lost today when our biological hardware crashes. When we
can access, store, and restore that information, then its longevity
will no longer be tied to hardware permanence.
Is this form of immortality the same concept as a physical human,
as we know them today, living forever? In one sense it is, because
as I pointed out earlier, we are not a constant collection of matter.
Only our pattern of matter and energy persists, and even that gradually
changes. Similarly, it will be the pattern of a software human that
persists and develops and changes gradually.
But is that person based on my mind file, who migrates across many
computational substrates, and who outlives any particular thinking
medium, really me? We come back to the same questions of consciousness
and identity, issues that have been debated since the Platonic dialogues.
As we go through the twenty-first century, these will not remain
polite philosophical debates, but will be confronted as vital and
A related question is, “Is death desirable?” A great
deal of our effort goes into avoiding it. We make extraordinary
efforts to delay it, and indeed often consider its intrusion a tragic
event. Yet we might find it hard to live without it. We consider
death as giving meaning to our lives. It gives importance and value
to time. Time could become meaningless if there were too much of
But I regard the freeing of the human mind from its severe physical
limitations of scope and duration as the necessary next step in
evolution. Evolution, in my view, represents the purpose of life.
That is, the purpose of life—and of our lives—is to evolve.
What does it mean to evolve? Evolution moves towards greater complexity,
greater elegance, greater knowledge, greater intelligence, greater
beauty, greater creativity, greater love. And God has been called
all these things, only without any limitation: infinite knowledge,
infinite intelligence, infinite beauty, infinite creativity, and
infinite love. Evolution does not achieve an infinite level, but
as it explodes exponentially, it certainly moves in that direction.
So evolution moves inexorably towards our conception of God, albeit
never reaching this ideal. Thus the freeing of our thinking from
the severe limitations of its biological form may be regarded as
an essential spiritual quest.
In making this statement, it is important to emphasize that terms
like evolution, destiny, and spiritual quest are observations about
the end result, not justifications for it. I am not saying that
technology will evolve to human levels and beyond simply because
it is our destiny and the satisfaction of a spiritual quest. Rather
my projections result from a methodology based the dynamics underlying
the (double) exponential growth of technological processes. The
primary force driving technology is economic imperative. We are
moving towards machines with human level intelligence (and beyond)
as the result of millions of advances, each with their own economic
justification. To use an example from my own experience at one of
my companies (Kurzweil Applied Intelligence), whenever we came up
with a slightly more intelligent version of speech recognition,
the new version invariably had greater value than the earlier generation
and, as a result, sales increased. It is interesting to note that
in the example of speech recognition software, the three primary
surviving competitors (Kurzweil—now Lernout & Hauspie,
Dragon, and IBM) stayed very close to each other in the intelligence
of their software. A few other companies that failed to do so (e.g.,
Speech Systems) went out of business. At any point in time, we would
be able to sell the version prior to the latest version for perhaps
a quarter of the price of the current version. As for versions of
our technology that were two generations old, we couldn’t even
give those away. This phenomenon is not only true for pattern recognition
and other “AI” software. It’s true of any product
from cars to bread makers. And if the product itself doesn’t
exhibit some level of intelligence, then intelligence in the manufacturing
and marketing methods have a major effect on the success and profitability
of an enterprise.
There is a vital economic imperative to create more intelligent
technology. Intelligent machines have enormous value. That is why
they are being built. There are tens of thousands of projects that
are advancing intelligent machines in many diverse ways. The support
for “high tech” in the business community (mostly software)
has grown enormously. When I started my OCR and speech synthesis
company in 1974, the total U.S. annual venture capital investment
in high tech was around $8 million. Total high tech IPOs for 1974
was about the same figure. Today, high tech IPOs (principally software)
are about $30 million per day, more than a thousand fold increase.
We will continue to build more powerful computational mechanisms
because it creates enormous value. We will reverse engineer the
human brain not because it is our destiny, but because there is
valuable information to be found there that will provide insights
in building more intelligent (and more valuable) machines. We would
have to repeal capitalism and every visage of economic competition
to stop this progression.
By the second half of this twenty-first century, there will be
no clear distinction between human and machine intelligence. On
the one hand, we will have biological brains vastly expanded through
distributed nanobot-based implants. On the other hand, we will have
fully nonbiological brains that are copies of human brains, albeit
also vastly extended. And we will have myriad other varieties of
intimate connection between human thinking and the technology it
Ultimately, nonbiological intelligence will dominate because it
is growing at a double exponential rate, whereas for all practical
purposes biological intelligence is at a standstill. By the end
of the twenty-first century, nonbiological thinking will be trillions
of trillions of times more powerful than that of its biological
progenitors, although still of human origin. It will continue to
be the human-machine civilization taking the next step in evolution.
Before the twenty-first century is over, the Earth’s technology-creating
species will merge with its computational technology. After all,
what is the difference between a human brain enhanced a trillion
fold by nanobot-based implants, and a computer whose design is based
on high resolution scans of the human brain, and then extended a
Most forecasts of the future seem to ignore the revolutionary impact
of the inevitable emergence of computers that match and ultimately
vastly exceed the capabilities of the human brain, a development
that will be no less important than the evolution of human intelligence
itself some thousands of centuries ago.
Copyright ' 2002 by the Discovery
Institute. Used with permission.
| ||Join the discussion about this article on Mind·X!||