|
Future
Vision: What's a Computer?
by
John SMART
A
Large Excerpt from the "Accelerating Times: Signs
of the Singularity, July-December 2001"
A multidisciplinary,
systems science exploration of the Big Picture of universal
development, local technological acceleration, compassionate
transhumanism, and signs of the coming singularity.
Future Vision: What's a Computer? Aaron Ricadela, Information Week, 9.10.01
Michio
Kaku is one of those rare futurists who understands
that ubiquitous computing will come much earlier than
ubiquitous A.I. In this all-too-short article, he intimates
that 2020 will be an era of very cheap, seamlessly integrating
computational devices, but we won't see serious AI for
perhaps another decade or two after that. This is spot
on with my own assessment of the coming technological
singularity: perhaps two decades more of relatively
stupid AI, then a decade of semi-interesting stuff,
then a decade of real surprises (mostly in the last
year or two of the latter decade). Go Michio! Let's
hope we at least see compelling voice recognition software
by 2010: my fingers are tired of typing!
Accelerating Technology Cycles, "A Crunch of Gears,"
Economist.com, Tech Quarterly, 09.20.01 Here is a brief
introduction to the great economist Joseph Schumpeter's
theories of cyclic technology development. In
my substrate language, Schumpeter is describing cyclic
phases of Variation (Differentiation, Divergence), Interaction
(Competition and Cooperation), and Selection (Extinction,
Convergence) ("VIS") within particular technology
paradigms. Schumpeter charted out the first four of
these economic cycles, ending in the late 1980's. I've
reformatted Schumpeter's first cycle from sixty to seventy
years, to make my case a bit clearer. James Watt took
out the patent for his new, radically more efficient
steam engine in 1769, so we can start the first era
in 1770, instead of 1780 as Schumpeter did. I've also
added the last three eras and the Post-Singularity Scenario,
extending the obvious, for your consideration. Enjoy.
Schumpeter observed:
A 70 year cycle, from 1770-1840.
Steam Power, Iron making, Cotton Spinning.
A 60 year cycle, from 1840-1900.
Steel making, Railway Era.
A 50 year cycle, from 1900-1950.
Internal Combustion Power, Electricity Grid.
A 40 year cycle,
from 1950-1990. Computing Power, Electronics, Petrochemicals,
Aerospace, Early Mol. Bio.
I further
propose:
A 30 year cycle, from 1990-2020. Building Out the Internet
Grid (the so-called "Stupid Network"), Early
Intelligence Amplification (IA), Commercial Biotechnology,
Ongoing Miniaturization, Weak Nanotech (Academic Evolvable
Hardware), 2nd Gen Robots, Early Evolutionary Computing.
Major National and International Sociopolitical Convergence
Reforms Begin.
A 20 year cycle, from 2020-2040. The Modularly Intelligent,
Distributed, Semi-Ubiquitous Network, Commercial Intelligence
Amplification, Powerful Biotech (Isolated "Medical
Miracles"), Early Computational Nanotech (True
Evolvable Hardware), 3rd Gen Robots, Commercial Evolutionary
Computing, Ongoing Significant Sociopolitical Reform,
Justice/Equity Crusades Continue, Early Transparent
Society, Accelerating Compassion Begins, Age of Materialism
Begins to Unravel.
A 10 year cycle, from 2040-2050. Semi-Intelligent Machines
and Networks, IA Suddenly Becomes Less Important,
More Powerful Biotech ("Common Miracles"),
Mature Computational Nanotech, 4th Gen Robots, Mature
Evolutionary Computing. Sociopolitical Systems Begin
to See Their Impending Extinction/Transformation, Personal
Social and Spiritual Transformations Become Big Issues,
Materialism Loses Further Ground. By mid-2040's, most
are still unaware, but hundreds of millions across the
globe sense an approaching "hurricane" of
change.
2050: Technological Singularity. The AI (and shortly
thereafter, AI's) Claim Self-Awareness. Autonomous Intelligence
Emerges on Earth.
Post
2050: Strong ("Drexlerian") nanotech arrives.
Technology begins to seem "magic" from the
perspective of unaugmented biology. Human machine integration
becomes ever more seamless at all scales. Voluntary
biological "uplifting" into the machine substrate
(willfully, but not statistically, reversible) begins,
slowly at first. We chart an accelerating course (100
years? 200?), inexorably taking us to a local Developmental
Singularity, Near Black-Hole Entity (NBE), and Universal
Transcension. Are you ready for the Future? It's ready
for you!
Some Notes about Calamity and Opportunity, David Brin,
Futurist.com, 10.01
There's been a lot written on the September 11th tragedy.
Brin's is one of the best short pieces I've seen yet.
If you'd like a good glimpse at the future, the inevitabilities,
the likelihoods and some of our best possibilities,
please read it. And consider buying Brin's Transparent
Society, and skimming that too, if you haven't yet.
There are inevitable changes afoot, but the path toward
them is entirely in our hands. Also well worth skimming
(read the headlines and any paragraphs you find unclear)
is Glen Heimstra and Brenda Cooper's "What September
11, 2001 Means for the Future". Take daily action
to make these futures happen. Love your planetary neighbors.
Let's ramp up the immune system, and break the chain
of violence.
"We Remain at the End of History" Francis
Fukuyama, 10.11.01
Excellent developmentalist thinking from Fukuyama, on
the one-month anniversary of 9/11. Though he couches
most of his arguments in cultural, rather than technological
terms, he clearly understands the inevitable appeal
of the freedoms and stability that comes from the construction
of a technological network that supports ever more decentralized,
pluralistic democracy among modern human beings. And
he recognizes that among all those legitimately angered
and offended by Western arrogance and insensitivity
(a problem with our path!), there is no credible alternative
destination being offered (only an alternative path
to democracy), though fundamentalism has been actively
engaged in such a search for "enlightened autocracy."
Perhaps directly due to the tremendous increase in quality
and quantity of information flow among humans, via the
technological infrastructure, we are now in the endgame
for political social structures that involve humans.
That battle has been fought, and won by pluralistic
democracy. As this inevitable attractor ripples across
the remaining autocratic regimes we should expect instabilities
and conflict to continue, but with each passing year
they will be ever more played out within the technological,
not the biological environment.
In the emerging technological substrate, the competition
for scarce computing resources, as intelligent machines
attempt to develop ever better solutions to universal
problems will intensify, and the drama will continue
to unfold. But as these battles will be fought by creatures
with continual lifespan learning plasticity, we should
expect their battles to be far less violent and far
more ethical than our own, and as several insightful
futurists have come to realize, they will occur in a
new niche, one that is not competitive with human beings.
The rate of advancement of biological life forms will
rapidly appear to be frozen in time, plant-like, by
comparison. As "frozen" as we currently
view molecular evolution, for example. (When was the
last time you worried about a some new kind of chemistry
replacing organic chemistry?)
More Prediction for the Coming Century: Human conflict
will of course also continue, until such time as humans
are stolen, via reversible voluntary upgrade/upload,
into the emerging autonomous technological civilization,
the way the youth of the few remaining indigenous tribes
such as the Korubo of Brazil (or the Amish, for that
matter) have all been stolen into modern human culture.
But human conflict will become powerfully constrained
by the developing technological intelligence, with human
rights and entitlements steadily increasing, but becoming
ever more carefully circumscribed with regard to their
negative externalities. We are in the process of creating
Earthpark (Spider Robinson's insightful phrase, implying
that virtually the whole geography and resources of
Earth will be left as a niche unneeded by the developing
nano AI, in its early years), a comfortable preserve
for all those humans who choose to remain biological,
for a final interval before the developmental singularity.
As Arthur Clarke has observed, we are nearing Childhood's
End. What an awesome time to be alive.
Nanotechnology Special Issue, Scientific American, September
2001 This issue has several overview articles with beautiful
graphics, like "The Once and Future Nanomachine"
by George Whitesides. Unfortunately, many of these articles,
such as Smalley's and Whitesides', contain misinformation,
such as the idea that mechanical molecular assemblers
aren't a workable concept (don't forget: our body is
built from them, so they are clearly workable!).
See the excellent and very detailed IMM response to
Whitesides', "Many Future Nanomachines,"
to get an overview of the key issues, such as friction/stiction,
self-replication, power, information storage, atomic
manipulation, etc., and the range of their potential
solutions. There is a lot to be learned by comparing
these two articles.
Bottom line here is that nature figured out how do to
it, and so can we. Nanotechnology appears imminent,
the only question remains which paths will be more fruitful
than others. In brief, we could make a good case that
all paths will be important and beneficial. But in greater
detail, it is my own instinct that the path we'll have
to take to create the strongest variety of assemblers
will necessarily involve self-replicating systems and
recursive evolutionary development, the way nature did
it. A whole pile of evolutionary lessons, "learned"
at the local, not the master level.
I presently can see two major ways humans will influence
this process. First, this won't be blind natural selection,
as took place in our ancestry, but artificially guided
natural selection. We use our science and artifice,
all our "top-down" knowledge and intuition
to improve the selection environment. And of course
all the "first generation" nano we get this
way will allow us to ever more accurately manipulate
biological systems in a top-down manner, making small
modifications to them that will have significant incremental
effects. This will of course feed back to give us vastly
greater understanding of biological computational systems,
which leads to our second major platform: advances in
virtual reality.
Second, and perhaps most importantly, we'll learn how
to develop a "bottom-up" virtual selection
environment that moves millions of times faster than
the chemical and biological one from whence we came.
This will happen when our molecular simulation environments
have both 1) encapsulated and 2) accelerated the physical
interactions we see in the natural world. Our current
simulation systems are still too primitive (we don't
know yet how to efficiently collapse the quantum functions
to classicality at the molecular level, even in approximation),
and are certainly too slow (ie, implemented at the software,
not hardware level) for us to do this. But these breakthroughs
may be around the corner.
Keep your eye on Blue Gene, that massively parallel
quadrillion bps IBM system that may bring us a lot closer
to solving, and eventually, real-time simulating, the
protein folding problem when it comes online in 2003.
When we have these kinds of simulation systems, pharmacogenomics
will finally come of age. As an extension of that paradigm,
we'll be able to use these environments to evolve the
solution to the kind of DNA-equivalent strand, and its
gestation environment, necessary to bring faster, smaller,
better-repairing, and smarter nanosystems into the environment.
It seems that our relentless progress in physics, computation
and molecular biology will soon (less than 20 years?)
give us the development environment we need to add this
"bottom-up" plank of a two-part nano strategy.
I could be wrong in my strategic assessment, of course.
We might not need protein folding approximations and
commercially viable pharmacogenomics before we can produce
complex self-assembling nanomachines. But I currently
think we will. Regardless, it seems clear that everything
our present generation of nanotechnologists can learn
about manipulating matter at the atomic scale can be
used to improve our "artificial (top-down) selection"
of these coming systems, and our "artificial manipulation"
of existing biological ones. We need to keep stoking
both our top down and our bottom up dreams of a nano-future.
Hopefully Scientific American will become less dismissive
of Foresight visions. They're going to see them implemented
sooner than they expect.
*
* * * *
Articles
of the Month
Artificial
Intelligence, Today and Tomorrow, Chris Moy, PatentCafe,
4.23.01
A
nice, short AI overview article. Gives some good big
picture overviews of the coming applications, and paints
a picture of a relatively benign, human services based
AI over the next 20 years.
Particuarly
interesting is this paragraph (bold is mine):
Can
true AI ever be accomplished?
With
all that processing power, can we expect computers to
look and respond like humans? Well, most likely not
in our lifetime and perhaps never.
There
are numerous reasons for this. The biggest argument
against developing true machine intelligence is the
argument of evolution. Machines have not
undergone the rigors of survival for millions of years
the way humans have in a dynamic, cultured environment.
The way we interact, think, respond and
adapt are all developmental characteristics that are
critical to our intellectual dominance. This process
took hundreds of millions of years to evolve
and
the failures along the way were critical to our ultimate
intellectual capacity. Ignoring those failures could
be a major hurdle in our efforts to develop
a truly
artificially intelligent machine.
Here Moy does a very nice job of characterizing the
central issue, still missed by most top-down AI designers:
technological evolutionary development is going to have
to recapitulate biological evolutionary development
in the process of building its knowledge base. So it
will only be when technological evolutionary development
can occur millions of times faster than occurred
in the biological substrate that we'll have the conditions
necessary for self-aware AI to emerge, in a mostly subconscious,
massively modular fashion (just like our prized wetware).
In a developmental sense, this requires self-replicating
systems, both in virtual and real worlds. In an evolutionary
sense, this means both high level simulation of "real"
reality, and simulation that can occur millions
of times faster than the reality it encapsulates. As
you may know, simulations that run faster than the real
world processes they emulate are usually done in hardware,
not software. So we may be talking about
evolvable hardware technology as one of the various
limiting factors. I'm looking forward to seeing metrics
for all this fascinating stuff to emerge in
coming years. As computer capability advances, authors
like Moy will come to realize that evolution is being
recapitulated --and massively accelerated-- in the technologic
substrate.
Editor:
John Smart. Available at:
http://www.SingularityWatch.com/pub.html
|
