Dear Rick,

Originally Posted by Rick Hirst

> I don't expect much (although I do have a hold placed at my
> library.) His last book, The Age Of Spiritual Machines,
> wasn't particularly enlightening.

I have a big, major disappointment with Ray Kurzweil that's colored
my entire attitude toward him.

Ray Kurzweil posted a paper in the late 1990s in which he found
empirical evidence that computer power has been growing slightly
faster than exponential, something like O(e^x^1.01) rather than
O(e^x). But unfortunately I didn't download and save that paper, and
I'd love to find it because I found a partial proof of exponential
growth for technology, and my proof implies that the rate is a little
faster than exponential.

But now I can't find the paper on his web site anywhere, and he's
changed his tune - he claims that since the mid-90s, technology
trends are exhibiting "accelerated growth," at O(e^x^2).
http://www.kurzweilai.net/articles/a...ml?printable=1

This value is nonsense and without justification. I believe his
motivation for this formula is the same kind of self-delusion,
delirium and denial that we see in financial analysts, journalists and
politicians who would rather claim the earth is flat than admit that
the stock market is more than 100% overpriced, that a crash is in the
offing, and that we're entering a new 1930s style Great Depression,
which is what standard exponential growth clearly implies.

Incidentally, the article I referenced above is called the "Law of
Accelerating Returns" is his seminal article on the subject, and I
suspect that his book is just an expansion of that article. The
problem is that he uses this so-called "Law" to derive many of his
other results, and I'll bet his new book does the same thing.

Originally Posted by Rick Hirst

> His observations also have some interesting implications for
> Generational Theory -- more on that later...

I'm interested. Could you explain this more?

Sincerely,

John

John J. Xenakis
E-mail: john@GenerationalDynamics.com
Web site: http://www.GenerationalDynamics.com

John, I think you ought to read Kurzweil's new book anyway. Some of his ideas seem to support your own views, especially his projection of "The Six Epochs of Evolution," which you may or may not think is like your Singulary 1 & 2, etc. But you criticise him for not mathematically justifying his argument about exponential growth rates of computer performance (r = e^x^y).

Originally Posted by John J. Xenakis

This value is nonsense and without justification. I believe his motivation for this formula is the same kind of self-delusion, delirium and denial that we see in financial analysts, journalists and politicians who would rather claim the earth is flat than admit that the stock market is more than 100% overpriced, that a crash is in the offing, and that we're entering a new 1930s style Great Depression, which is what standard exponential growth clearly implies.

If you recall, I could not find much value in your "statistical" assumptions about universally ubiquitous life and intelligence. And, if you recall, I was not favorably impressed by your reasoniong. Nevertheless, I did not throw your baby out with the bathwater. I only raised the corner of one loose tile in your kitchen of credible concoctions. I still like your general argument. And I still think Kurzweil's argument is good in a more general way.

But I did find something in The Singurality Is Near that troubles me more than just a little. Kurzweil makes a big fuss about the coming age GNR (genes, nanotechnology, robotics), and yet I see a huge mistake on his part about the role of genes in biological evolution. Something he said on p. 16 seems to reverberate throughout his book:

Originally Posted by Ray Kurzweil

In the second epoch, starting several billion years ago, carbon-based compounds became more and more intricate until complex aggregations of molecules formed self-replicating mechanisms, and life originated. Ultimately, biological systems evolved a precise digital mechanism (DNA) to store nformation decribing a larger society of molecules. This molecule and its supporting machinery of codons and ribosomes enabled a record to be kept of the evolutionary experiments of the second epoch.

What? Biological systems came before the genes? Good grief! How did that happen? What exactly were those "biological systems" is they had no genes? There are no credible hypotheses or evidenece that I know of that specifies life before genes. "Self-replicating mechanisms" without genes? Ah, like what, exactly?

Thus I think Kurzweil makes an evolutionary biological mistake. All right, so he's not an evolutionary biologist. I can still extract value from other aspects of his theory. I intend to take this up with Kurzweil (ray@singularity.com), and to ask him where he got this peculiar notion about biogenesis. Maybe you could do the same about his exponentiality-of-growth assumption.

--Croakmore

Originally Posted by Croakmore

Originally Posted by Rick Hirst

If you want more scientifically grounded speculation, try Radical Evolution (from the author of Nine Nations), or On Intelligence. For fiction, try Darwin's Radio.

I'm inclined to agree with you about Kurzweil's last book. But the other books you mention don't appeal quite as much to me for various reasons. The main reason is consistent with Bill Joy's 2000 WIRED article "Why the future doesn't need us" -- because we seem to be doing what Kurzweil says we're doing: transcending biology. From my POV I think this Singularity business speaks more to the point, that instead of enhancing human bodies with nanochips and robotics we will be (eventually) leaving the squishy parts behind for a friendlier cyber existence where immortality is a matter of choice.

I picked those two books for a reason. On Intelligence, from Jeff Hawkins (inventor of the PalmPilot), examines why current approaches to computer hardware and software development are unlikely to achieve the consciousness transfer you envision -- not that it's impossible, just that we're approaching it the wrong way.

Radical Evolution is a collection of discussions with the leading thinkers of the day on this topic, ranging from what Garreau titles the "Heaven" scenario as promoted by Kurzweil, to the "Hell" scenarios promoted by the aforementioned Joy. Also included is Francis Fukuyama, whose most recent book I found equally uninspiring. I'm more interested in the third, in-between scenario entitled "Prevail", and analysts such as John Mauldin (I mentioned him upthread), who describes the next decade as the "Muddle-Through Decade".

Originally Posted by Croakmore

I'm beginning to see biological humans as larvae for the future version of us -- cyber-sainted butterflies. I can visualize a time when nearly everything a biological human does involves preparation for his or her ultra-tricked-out computerized future.

I tend to see it the same way, and I even tend to agree with Kurzweil's time frame -- circa 2050 for the first brain dumps. But what happens between now and then? And what will that "preparation" be? That's the far more interesting question to me. I'm not so interested in the destination as much as the journey... then again, I am a Nomad... :wink:



P.S. I just checked out Darwin's Children, the sequel to Darwin's Radio from the library. I'll get to it as soon as I can, but first I have to finish Singularity Sky. :wink:

Addressed transitional scenarios. For example, nanomachines that would interface with individual neurons to bring new capability to our organic brains. If this could be done before uploading is perfected, we may have a Singularity before The Singularity.

Kurtzweil commented that we likely can no more comprehend a post-human entity than a chimpanzee can understand us.

Let's just hope said cybots (Cylons?) don't have a single glowing red eye that sweeps back and forth. :wink:

Wall Street Journal review of
Ray Kurzweil's The Singularity is Near

.... or ... Lee Gomes is an idiot



The Wall Street Journal

October 5, 2005

PORTALS
By LEE GOMES


A Back-Cover Brush
With a High-Tech Seer
And Some of His Pals

October 5, 2005; Page B1

While I recall once hearing something about not judging a book by its
cover, I couldn't help but be bothered by some of the suppliers of
blurbs on the back of the new Ray Kurzweil book, "The Singularity Is
Near."

Mr. Kurzweil is well-known in technology circles as a flag waver for
the "strong" school of artificial intelligence, or AI. This is the
belief that future computers won't just become increasingly powerful
but also increasingly like people.

The theme has been an enduring one for Mr. Kurzweil, who claims street
cred on the topic because of his prior work as a computer scientist
and inventor. An earlier Kurzweil book talked about "spiritual"
machines. This new one promises fully human-like computers in 20 or so
years.

But that's just the warm-up, the amuse bouche. By 2045, predicts Mr.
Kurzweil, human and computer intelligence will merge -- the
"singularity" of the title -- and become some grand new form of life,
one that will then venture out into the universe.

It will be like the end of "2001," I suppose, though without the
Richard Strauss soundtrack.

The first of the off-putting book blurbs comes from Marvin Minsky, the
godfather of strong AI. As a consultant during the filming of "2001"
during the 1960s, Prof. Minsky made predictions about how common
HAL-like machines would be in 30 years. You know how this story ends:
Computers can still barely open a printer port, much less the pod bay
doors.

Minsky's blurb was a reminder of how we have been reading predictions
for ascending computer intelligence for decades. I can't help but ask:
"Where is the ox?" (That, by the way, is "Where's the beef?"
translated -- first into French and then back into English -- by the
artificially intelligent computers at Babel Fish.)

Mr. Kurzweil, it must be said, admits that strong AI hasn't yet
delivered the goods. But he maintains that's because it has gone about
it the wrong way. The new approach will be to first "reverse engineer"
the brain, using all the tools of modern neuroscience.

But that's a variation on the same dodge that strong AI types have
been using for ages. First, they concede that things haven't quite
worked out as planned. But then they say not to worry, because some
new research direction will bring home the still-missing bacon. The
notion that the entire strong AI project may have a flimsy
intellectual foundation is never seriously addressed.

For his reverse-engineering approach to work, Mr. Kurzweil needs to be
right twice. First, the brain must indeed give up its secrets --
including some big ones, like consciousness -- to scanners and probes.
Second, all of those secrets must then be duplicated (and not just
simulated or modeled) inside a computer.

That second step may seem trivial, but I fret that we have hopelessly
confused a computer simulation of something with the duplication of
it. We have, for example, increasingly powerful computer models of the
weather. But you can run one of them in your backyard until the cows
come home and you're not going to make any rain.

Another problematic blurb supplier was Bill Joy, the former Sun
Microsystems engineer who has lately gotten much attention for making
predictions similar to Mr. Kurzweil's, albeit with a negative spin.
Mr. Joy is worried about what these super-smart machines will do to
us, their human makers.

Many people find writing like Mr. Joy's, coming, as it does, from
someone in the computer field, to be humble and cautionary. But I find
it enormously self-aggrandizing. After all, how many people get to say
that what they do is so important that it may rend the very fabric of
life on Earth? And this over-the-top dystopia diverts attention from
what may be the real downsides of computers, which are more prosaic:
alienation, short attention spans and the like.

My modest proposal is for the whole Kurzweil-Minksy-Joy lot to be
locked up somewhere -- under humane conditions, of course, with plenty
of sunlight and good Internet access -- and forced to actually build
the computers that they keep insisting are just around the corner.
Until that happens, no more books or articles or hectoring.

In a phrase: "Put to the top or closes it," which is Babel Fish
franglais for "Put up or shut up."

My own prediction is that in 20 years, computers will be about as
different from today's machines as today's are from those of two
decades ago. Which is to say, not a whole lot. More powerful,
obviously; putatively more useful; more insinuated into our lives, no
doubt. But still not your roommate or spouse.

Mr. Kurzweil objects to this as "linear thinking," and says that the
pace of change in coming years will be vastly faster than what we have
heretofore experienced. But beware of folks who are forced by the
facts of their case to play down the extent to which the past is a
guide to the future. A smart computer would never make that mistake.

• Send your comments to lee.gomes@wsj.com 1.

URL for this article:
http://online.wsj.com/article/SB112847119821360122.html

Hyperlinks in this Article:
(1) mailto:lee.gomes@wsj.com
(2) http://online.wsj.com/article/0,,portals,00.html
(3) http://online.wsj.com/article/0,,por...change,00.html
(4) http://online.wsj.com/search/aggregate?article-doc-type={Boom+Town}%20{Boom+Town+Guest}%20&KEYWORDS=% 2A&FIELD=field=vdkvgwkey%2Bmain%2Dhed%2Barticle%2D doc%2Ddate%2Barticle%2Ddoc%2Dpublication%2Barticle %2Ddoc%2Did%2Barticle%2Ddoc%2Dtype%2Bs%2Dmain%2Dhe d&SORT_STRING=sort%2Dstring%3Darticle%2Ddoc%2Ddate %2Bdesc%2Bmain%2Dhed%2Basc&SEARCH_SECTION=article-body&COLLECTION_AGG=collection=wsjie/archive&HEADER_TEXT=BOOMTOWN
(5) mailto:http://online.wsj.com/article/0,,portals,00.html
Copyright 2005 Dow Jones & Company, Inc. All Rights Reserved

John, that's a lot fluff from the WSJ. Gomes should read more than just the back cover to prepare such a high-profile book review. But who cares what he thinks? I'd like to know what Hofstadter thinks, among others, and we'll probably find out. Besides, there's always the Turing test to measure that crucial milestone.

Kurzweil predicts that around 2030 computers will have passed the Turing test. Soon afterwards the great "uploading" will begin. Millies will be in mid-life (and both Boomers and Silents will be, well, dead). I can visualize see some Xer's grandchild saying "When are you going to get uploaded, Grandpa? I want you to virtually attended my college graduation."

Well I dreamed I saw the silver
space ships flying
In the yellow haze of the sun.
There were children crying
And colors flying
All around the chosen ones
All in a dream All in a dream
The loading had begun.
They were flying Mother Nature's
silver seed to a new home in
the sun.

Neil Young (After The Gold Rush)

--Croak

Kurzweil brings up a good point that weighs heavily on our true ability to predict the outcome of evolutionary processes. He discusses Edward Fredkin’s “new theory of physics,” wherein the universe is assumed to be ultimately composed of software. Fredkin differentiates between the analytical approach using traditional mathematics from the computational approach using algorithms. The analytical approach would predict a future state of an evolving system without considering any of the intermediate states the system will occupy between “now” from “then.” The computational approach requires effective consideration of every state along the way from “now” to “then.”

This why I do not believe that evolution, especially macroevolution, is predictable. Evolution appears to me to operate like cellular automata with programmed rules (as discussed in The Singularity Is Near, see pp. 85-94), where one state of occupation determines the next state of occupation. Here is a beautiful example of 2-D patterns in the shell of a cone snail, produced by cellular automata:


[photo by Eddie Hardy, posted for discussion purposes only]

Here is another species with slightly different programmed rules:


[photo by Eddie Hardy, posted for discussion purposes only]

Thus one cannot build an analytic model of the specific pattern appearing on the shell of a cone snail (other than the broadest features that different one cone snail species from another.) Only a computational model will do.

--Croakmore

Croak,
I believe that Mr. Kurzweil is speaking at Town Hall in Seattle on Monday nite, Oct 10 , if you're interested.

Just read the book. Fantastic. He dates the Singularity as 2045, putting it in a late High or early Awakening. Maybe the Singularity will "catalyze" the Awakening, being the Kennedy assassination of the 21st Century.

Dear Tom,

Originally Posted by Tom Mazanec

< Just read the book. Fantastic. He dates the Singularity as 2045,
< putting it in a late High or early Awakening. Maybe the
< Singularity will "catalyze" the Awakening, being the Kennedy
< assassination of the 21st Century.

The other day I spent three hours in Barnes & Noble reading
Kurzweil's book, and I agree that it's fantastic. It's very well
done, and a thorough presentation.

Having said that, the book has several content flaws. (I'm using the
phrase "content flaws" because I don't know whether they're logic
errors, or whether he was just trying to appeal to the crowd. If the
latter, it doesn't appear to have worked, in that I don't believe
that the book is doing well.)

One is the 2045 date, which is unrealistically distant, in a world
where supercomputers will have the power of the human mind by 2010,
and where the Pentagon will begin to deploy autonomous,
decision-making computer soldiers by 2015. I feel pretty certain
that supercomputers will pass the Turing Test by 2015. (The Turing
Test for a computer is whether you can communicate with the computer
and not be able to tell whether it's a computer or a human.)

The second issue is this merging of human minds with computers. That
makes as little sense as merging a human mind with an ape's mind.
(What do you get? A hairy human that likes bananas?) I don't doubt
that there'll be some experimenting with the merging of human minds
with computers, but I don't see any plausible way that it will go
much beyond that, or that it will make any sense to do so.

Third -- and this criticism is admittedly a little parochial -- he
doesn't integrate his development with generational theory, which he
should at least be aware of. Worse, he ignores a direct consequence
of his presentation of exponential growth: that we haven't recovered
from the 1990s bubble, and that a stock market crash can't be
avoided. He must surely be aware of this, and aware of its
importance, and aware of the fact that he's uniquely able to
publicize that fact, and able to gain some credibility when the crash
comes. Instead, he avoids it, and covers over it with a dubious
presentation of doubly-exponential growth.

Still, I am impressed with the depth of his research and the breadth
of his presentation. This book will be valuable as a reference book
on the Singularity for years to come.

Sincerely,

John

John J. Xenakis
E-mail: john@GenerationalDynamics.com
Web site: http://www.GenerationalDynamics.com

Kurzweil and Joy: Un-publish that genome!

Here's some silliness from Ray Kurzweil and Sun Microsystem's Bill
Joy:


http://www.nytimes.com/2005/10/17/op...gewanted=print


The New York Times
October 17, 2005
Op-Ed Contributors
Recipe for Destruction
By RAY KURZWEIL and BILL JOY

AFTER a decade of painstaking research, federal and university
scientists have reconstructed the 1918 influenza virus that killed 50
million people worldwide. Like the flu viruses now raising alarm bells
in Asia, the 1918 virus was a bird flu that jumped directly to humans,
the scientists reported. To shed light on how the virus evolved, the
United States Department of Health and Human Services published the
full genome of the 1918 influenza virus on the Internet in the GenBank
database.

This is extremely foolish. The genome is essentially the design of a
weapon of mass destruction. No responsible scientist would advocate
publishing precise designs for an atomic bomb, and in two ways
revealing the sequence for the flu virus is even more dangerous.

First, it would be easier to create and release this highly
destructive virus from the genetic data than it would be to build and
detonate an atomic bomb given only its design, as you don't need rare
raw materials like plutonium or enriched uranium. Synthesizing the
virus from scratch would be difficult, but far from impossible. An
easier approach would be to modify a conventional flu virus with the
eight unique and now published genes of the 1918 killer virus.

Second, release of the virus would be far worse than an atomic bomb.
Analyses have shown that the detonation of an atomic bomb in an
American city could kill as many as one million people. Release of a
highly communicable and deadly biological virus could kill tens of
millions, with some estimates in the hundreds of millions.

A Science staff writer, Jocelyn Kaiser, said, "Both the authors and
Science's editors acknowledge concerns that terrorists could, in
theory, use the information to reconstruct the 1918 flu virus." And
yet the journal required that the full genome sequence be made
available on the GenBank database as a condition for publishing the
paper.

Proponents of publishing this data point out that valuable insights
have been gained from the virus's recreation. These insights could
help scientists across the world detect and defend against future
pandemics, including avian flu.

There are other approaches, however, to sharing the scientifically
useful information. Specific insights - for example, that a key
mutation noted in one gene may in part explain the virus's unusual
virulence - could be published without disclosing the complete genetic
recipe. The precise genome could potentially be shared with scientists
with suitable security assurances.

We urgently need international agreements by scientific organizations
to limit such publications and an international dialogue on the best
approach to preventing recipes for weapons of mass destruction from
falling into the wrong hands. Part of that discussion should concern
the appropriate role of governments, scientists and their scientific
societies, and industry.

We also need a new Manhattan Project to develop specific defenses
against new biological viral threats, natural or human made. There are
promising new technologies, like RNA interference, that could be
harnessed. We need to put more stones on the defensive side of the
scale.

We realize that calling for this genome to be "un-published" is a bit
like trying to gather the horses back into the barn. Perhaps we will
be lucky this time, and we will indeed succeed in developing defenses
for these killer flu viruses before they are needed. We should,
however, treat the genetic sequences of pathological biological
viruses with no less care than designs for nuclear weapons.

Ray Kurzweil, an inventor, is the author of "The Singularity is Near:
When Humans Transcend Biology." Bill Joy, founder and former chief
scientist of Sun Microsystems, is a partner at a venture-capital
firm.

Originally Posted by John J. Xenakis

Kurzweil and Joy: Un-publish that genome!

Here's some silliness from Ray Kurzweil and Sun Microsystem's Bill
Joy:

http://www.nytimes.com/2005/10/17/op...gewanted=print

I attended Kurzweil's lecture at Microsoft last week. His delivery was painfully monotone, so I left about halfway through; but when he mentioned that he was going to joint-publish with Joy, I nearly fell out of my chair. I couldn't imagine what the two would have in common, but after a while I realized: both suffer from a fetishization of High Technology, wherein digital information is automatically orders of magnitude more powerful than analog information simply because it is digital.

Did I miss a Really Big Announcement, or is a genomic sequence still far from sufficient to actually produce a living organism?

In the real world, a would-be bioterrorist intent on spreading bird flu would simply acquire a few dozen fowl from Vietnam (where the virus is endemic in the bird population), culture the virus the same way cheese or yeast has been cultured for the last ten thousand years, and inject himself and/or other (un)willing subjects until the necessary mutation occurred.

So why hasn't this already been done, given that the bird flu has been spreading for a couple of years now? Well, perhaps there's a dearth of dedicated, yet scientifically gifted psychopaths in the world. Maybe Ernst Blofeld only exists in the movies.

Or maybe it's actually impossible. Time will tell -- Mother Nature has a way of making the possible... probable -- no terrorist bogeyman required.

Does anyone believe that the events in Charles Stross's "Accelerando" (downloadable for free at www.accelerando.org) are possible or even likable?

Originally Posted by Rick Hirst

In the real world, a would-be bioterrorist intent on spreading bird flu would simply acquire a few dozen fowl from Vietnam (where the virus is endemic in the bird population), culture the virus the same way cheese or yeast has been cultured for the last ten thousand years, and inject himself and/or other (un)willing subjects until the necessary mutation occurred.

So why hasn't this already been done, given that the bird flu has been spreading for a couple of years now? Well, perhaps there's a dearth of dedicated, yet scientifically gifted psychopaths in the world. Maybe Ernst Blofeld only exists in the movies.

IMHO, neither the Mad Scientists of James Bond nor the The Professor of Gilligan's Island exist in the real world. They are embodiments of U.S. society's anxieties and hopes about the power of science.

Thanks to John I'm back on the old The Singularity thread. So, it occurred to me today that this so-called Singularity, popularized by Ray Kurzweil, may become manifestly indistinguishable from that ultimate search engine looming in our future.

What will it mean when some whizbang search engine becomes smarter than the collective intelligence of all human beings?

Another question: What would you do today without your Google? We treat Google like some Wizard of Oz—"Oh, great wizzard Google, how many frogs are dying in the Amazon because of bad forestry practices?" Or: "Oh, great Google, how can I get from Boston to New York by way of the Erie Canal?"

Darwin must be smiling when we now speak of the evolution of search engines, which are nothing more than a lot digits flinging around inside computers and on the telephone lines. We've come to rely so heavily on digits you'd think they came here from another place "to steal your face right off your head," as Panama Red put it. Something strange is evolving in this digital world of search engines. Maybe The Singularity will be some kind of coming-out party.

Here is one site concerned about the evolution of search engines, posted on Webcredible:

Maybe we can't see it coming. Maybe Google is just a front for this optimization monster, creeping up on us from behind. Evolution is a dangerous thing if we don't what's coming around the corner. Look out for natural selection!

Will The Singularity be friendly to us when it arrives?

Dear Richard,

Originally Posted by Croakmore

> What will it mean when some whizbang search engine becomes smarter
> than the collective intelligence of all human beings?

> Another question: What would you do today without your Google? We
> treat Google like some Wizard of Oz—"Oh, great wizzard Google, how
> many frogs are dying in the Amazon because of bad forestry
> practices?" Or: "Oh, great Google, how can I get from Boston to
> New York by way of the Erie Canal?"

> Darwin must be smiling when we now speak of the evolution of
> search engines, which are nothing more than a lot digits flinging
> around inside computers and on the telephone lines. We've come to
> rely so heavily on digits you'd think they came here from another
> place "to steal your face right off your head," as Panama Red put
> it. Something strange is evolving in this digital world of search
> engines. Maybe The Singularity will be some kind of coming-out
> party.

> Here is one site concerned about the evolution of search engines,
> posted on Webcredible:

> Maybe we can't see it coming. Maybe Google is just a front for
> this optimization monster, creeping up on us from behind.
> Evolution is a dangerous thing if we don't what's coming around
> the corner. Look out for natural selection!

> Will The Singularity be friendly to us when it arrives?

You may recall that a couple of years ago I posted the design of
software that would turn computers into super-intelligent computers,
once they were powerful enough.
http://www.fourthturning.com/forum/s...&postcount=268

Since I don't know how the human brain works, the objective was not
to design software that would work exactly the same way that the
human brain works. Instead, the objective was to develop
intelligence and learning and goal-setting and creativity so the
computer would act as an intelligent human being would (and, in fact,
would soon far surpass a human being in intelligence).

Even though the software didn't have to work exactly the same way
that the human brain works, the process of designing the software
made it clear to me why the human being works in certain way.

For example, I really began to understand why sleeping is important.

Like a human being, a super-intelligent computer has to continually
process huge amounts of information. It will see things and hear
things constantly, and these things are stored in the computer's
memory (just as new things have to be stored in a human being's
memory).

But you very quickly run into the same problem that Google faces: How
do you keep track of all that information, and how do you retrieve a
particular piece of information when it's needed?

The super-intelligent computer's database has to be set up in an
associative manner, so that, for example, if you ask the computer
about Madonna, then it can retrieve all the information it has about
Madonna, including old information from biographies and new
information from television.

Building that kind of associative database very quickly runs into
severe performance problems. The number of new "facts" that you have
to learn every day may grow linearly, but the number of associative
links between those facts has combinatorial growth (faster than
exponential growth).

In the human brain, the purpose of sleeping is to provide "down time"
so that the brain can rearrange its collection of facts, so that only
the most important ones are in active memory, and the less important
ones get moved, somehow, to the brain's version of "secondary
storage." If you want to retrieve one of those facts, then you have
to think a while. ("It's right on the tip of my tongue ... I can
almost remember it ... aha!")

In the super-intelligent computer software, you have to do exactly
the same thing -- provide a way for less important portions of "active
memory" to be moved into secondary storage. Super-intelligent
computers probably won't have to "sleep" to do that, since there are
ways to add additional processors to do that kind of work in the
background.

But it's still very difficult to do, and the sophisticated algorithms
being developed by Google and other search engine companies will
undoubtedly be very useful.

Sincerely,

John

John J. Xenakis
E-mail: john@GenerationalDynamics.com
Web site: http://www.GenerationalDynamics.com

http://apnews.myway.com/article/20070715/D8QD61V80.html

BALAD AIR BASE, Iraq (AP) - The airplane is the size of a jet fighter, powered by a turboprop engine, able to fly at 300 mph and reach 50,000 feet. It's outfitted with infrared, laser and radar targeting, and with a ton and a half of guided bombs and missiles.

The Reaper is loaded, but there's no one on board. Its pilot, as it bombs targets in Iraq, will sit at a video console 7,000 miles away in Nevada.

Originally Posted by John J. Xenakis

In the super-intelligent computer software, you have to do exactly
the same thing -- provide a way for less important portions of "active
memory" to be moved into secondary storage. Super-intelligent
computers probably won't have to "sleep" to do that, since there are
ways to add additional processors to do that kind of work in the
background.

Of course it's unnecessary to sleep completely. Dolphins faced the requirement of active attitude control: they can't afford to go to sleep because they would roll over and drown. The answer was to sleep one hemisphere at a time. If an associative AI also faces the need for constant uptime, it will handle the problem in the same way.

That said, the level of sophistication needed for an associative, learning AI has been continually underestimated for decades. Each time, the cause has been traced to lack of awareness of preprocessing software inbuilt to the human perception toolset that massively reduced the amount of necessary conscious processing. Examples include 'expert AI's, the famous ELIZA case, visual pattern recognition, speech synthesis and recognition, the Cyc database project... on and on and on. Attempts to fake this using massive databases can produce hacks that work -- such as chessplaying routines that can now outthink grand masters -- but they're still hacks with no learning capabilities.

I do think AI will come, but not before we learn a tremendous amount about the human brain, enough to do direct and sophisticated engineering with brains. At that point other routes to improved cognition will appear, such as mind-mind interfaces, mind-machine interfaces, collective distributed intelligence, emergent-behavior intelligences, the old 'brain-in-a-vat' cliche...

This is part of why I don't like the term 'Singularity', preferring the broader term 'Coaelescence' to describe the inclusivity that the process will necessitate. (Other reasons: I know from biology that all exponential growth curves must come to an end, and the limits of 'Singularity' growth have been ill explored; I know from ecology that population explosions often lead to population crashes; I know from genetics that bottlenecks create illusions of 'inevitability' in behavior; I know from mathematics that probability laws can also produce exponential curves, but may vary from them if a low-probability event occurs; I know from economics that past performance is no guarantee of future results; I know from physics that negative entropy has a cost in enthalpy, and heat dissipation is starting to be a problem already.) I think the general process described by singulatarians will happen, but not at the breakneck speed posited.

Finally, I think Accelerando is a great example of an 'if this goes on' story, but the one thing that singulatarians and fourth-turners agree on is that linear trends do not last, and Accelerando depends on the assumption that the information tech and economics fields will remain as they were at the turn of the century for the next half-century. It so is not going to happen.

Originally Posted by catfishncod

> That said, the level of sophistication needed for an associative,
> learning AI has been continually underestimated for decades. Each
> time, the cause has been traced to lack of awareness of
> preprocessing software inbuilt to the human perception toolset
> that massively reduced the amount of necessary conscious
> processing. Examples include 'expert AI's, the famous ELIZA case,
> visual pattern recognition, speech synthesis and recognition, the
> Cyc database project... on and on and on. Attempts to fake this
> using massive databases can produce hacks that work -- such as
> chessplaying routines that can now outthink grand masters -- but
> they're still hacks with no learning capabilities.

> I do think AI will come, but not before we learn a tremendous
> amount about the human brain, enough to do direct and
> sophisticated engineering with brains. At that point other routes
> to improved cognition will appear, such as mind-mind interfaces,
> mind-machine interfaces, collective distributed intelligence,
> emergent-behavior intelligences, the old 'brain-in-a-vat'
> cliche...

> This is part of why I don't like the term 'Singularity',
> preferring the broader term 'Coaelescence' to describe the
> inclusivity that the process will necessitate. (Other reasons: I
> know from biology that all exponential growth curves must come to
> an end, and the limits of 'Singularity' growth have been ill
> explored; I know from ecology that population explosions often
> lead to population crashes; I know from genetics that bottlenecks
> create illusions of 'inevitability' in behavior; I know from
> mathematics that probability laws can also produce exponential
> curves, but may vary from them if a low-probability event occurs;
> I know from economics that past performance is no guarantee of
> future results; I know from physics that negative entropy has a
> cost in enthalpy, and heat dissipation is starting to be a problem
> already.) I think the general process described by singulatarians
> will happen, but not at the breakneck speed posited.

That's quite an amusing list of lessons learned, but I still have to
disagree with your conclusions.

First off, I'm pretty much on the verge of just calling all of
Artificial Intelligence a complete scam. AI has never accomplished
ANYTHING that I'm aware of.

This became overwhelming clear with last week's announcement of a
world champion checker-playing program.
** Researchers solve the game of checkers (draughts)
http://www.generationaldynamics.com/cgi-bin/D.PL?xct=gd.e070723b#e070723b

What an example of AI! What's the algorithm? Play all possible
games of checkers, and keep track of the winning moves. What does AI
have to do with that? I used the same algorithm in the 70s, but then
it was for 3D tic tac toe. There's more AI in a program that
computes corporate income taxes than there was in the checker
program.

I guess that you could say that there's a little bit of AI in the
Deep Fritz program that beat world chess champion Vladimir Kramnik
last year.
** Deep Fritz computer beats world chess champion Kramnik, 4-2.
http://www.generationaldynamics.com/cgi-bin/D.PL?xct=gd.e061206c#e061206c

But what's the AI in that? Just a minimax algorithm that was
invented in the 1950s. Greenblatt's 1967 MacHack chess playing
program could be world champion too if it could look 18 moves ahead.

And that's the point. There's nothing to AI except for two
algorithms: pattern matching (like the checker playing program) and
minimax (like the chess playing program). Other than that, AI is
nothing but thumb-twiddling.

So when you talk about databases producing "hacks that work," that's
exactly right. Who cares about "associative, learning AI"? It's
garbage. It's just a way to get funding. It's meaningless. The
ONLY thing that works is brute force -- either minimax or pattern
matching with brute force.

And of course these "hacks" have learning capabilities. As I point
out in my checkers article, Arthur Samuel's 1950s check program had
the ability to learn -- by playing over games and seeing how they
turn out. That's how humans learn, by the way.

A computer with super-human intelligence will not use any of the crap
that AI professors screw around with. It will use brute force
pattern matching for vision, voice recognition, natural language
processing, etc., and minimax for decision-making and goal-reaching.

And learning? Just turn it loose on the internet and have it read
everything. It'll learn, with the right kind of brute force
algorithm.

My estimate of late 2020s for the Singularity is not based on the
belief that AI researchers will come up with some mystical algorithm
that captures the essence of human thought, or some crap like that.
That's simply my estimate for how long it will take for computers to
be powerful enough to implement simple brute-force algorithms that
will do everything necessary.

Sincerely,

John

John J. Xenakis
E-mail: john@GenerationalDynamics.com
Web site: http://www.GenerationalDynamics.com

http://blog.wired.com/defense/2007/0...wnational.html

Robots have been roaming the streets of Iraq, since shortly after the war began. Now, for the first time -- the first time in any warzone -- the machines are carrying guns.

Dear Matt,

Originally Posted by MichaelEaston

> First Armed Robots on Patrol in Iraq (Updated)


http://blog.wired.com/defense/2007/08/httpwwwnational.html

That's a great video accompanying that story.

One thing that I noticed was missing was the any autonomous
capabilities of the robot, or any ability to track a moving human
being.

The South Korean robots that I wrote about apparently are able to
track a human being by themselves, and are allowed to "decide" to
shoot to kill under certain circumstances.
** South Korea to deploy deadly robots on border with North Korea
- This is the scariest video you've ever seen.
http://www.generationaldynamics.com/cgi-bin/D.PL?xct=gd.e070203#e070203

I've also recently read a news story that Israelis are considering
the use of armed robots on the border with Gaza.

Meanwhile, here's a story about a robot that examines women's
breasts:
http://www.newscientist.com/article.ns?id=dn7630

Sincerely,

John

John J. Xenakis
E-mail: john@GenerationalDynamics.com
Web site: http://www.GenerationalDynamics.com

Originally Posted by Tom Mazanec

Just read the book. Fantastic. He dates the Singularity as 2045, putting it in a late High or early Awakening. Maybe the Singularity will "catalyze" the Awakening, being the Kennedy assassination of the 21st Century.

Perhaps, I'd love to see an Awakening, I wish I could have been around during the last two, those were such good times.

Originally Posted by sean '90

Perhaps, I'd love to see an Awakening, I wish I could have been around during the last two, those were such good times.

Not for folks like us they weren't! Sure, the Prophets and Artists probably had tons of fun, with their muckraking and free loving and rioting, but we...we're the Organization Man, the Establishment. Hey hey, LBJ, how many kids you gonna kill today?

I do not relish being a Civic around in the next 2T. Maybe I'll pull a Timothy Leary and preempt my kids' rebellion by being hipper (or should I say, hippier) than they. Then they can be like preseasonal 3T Nomads and turn into Alex P. Keatons.

Sigh, the options are not good.

Originally Posted by 1990

Not for folks like us they weren't! Sure, the Prophets and Artists probably had tons of fun, with their muckraking and free loving and rioting, but we...we're the Organization Man, the Establishment. Hey hey, LBJ, how many kids you gonna kill today?

I do not relish being a Civic around in the next 2T. Maybe I'll pull a Timothy Leary and preempt my kids' rebellion by being hipper (or should I say, hippier) than they. Then they can be like preseasonal 3T Nomads and turn into Alex P. Keatons.

Sigh, the options are not good.

It wasn't so good even for Prophets if you didn't share the mindset of those you found yourself surrounded by. And lots of people didn't.

Muckraking for the sake of muckraking led to Carter's "national malaise", opening the door for someone like Reagan... who in turn set the stage for today's corporate malfeasance.

Free-loving and drug-doing led to blighted courtship rituals for Xers and later-born Boomers... largely by casting the traditional-minded as hopeless wimps.

And the riots were just plain scary... THE proximate cause of urban blight that didn't begin to abate until the mid-90s.

The fun times I remember from the Awakening largely involve events and activities... Sunday drives into the Lake Country... weekends down at the Jersey Shore... neighborhood block parties... playing numerous variants of Tag/Hide-And-Seek (do people even still play H&S???) that in retrospect were holdovers from the High.