From: Pope Salmon the Lesser Mungojelly (rainbow@beautywood.org)
Date: Mon Oct 10 2005 - 19:47:46 MDT
CONTENTS
Contents
Dedication
Introduction
Summary Of This Post
Individual Vs Collective Intelligence
Technology Rewires Collective Intelligence
Yahoogroups, For Example
Integration Is Superhuman Intelligence
Integration Comes In Through The Back Door
One Last Thing: Raising The Roof
Conclusion And Signature
DEDICATION
Om Gam Ganapataye Namah!
For Superduper!
INTRODUCTION
My name is Mungojelly. I've never posted here before. I spent all day
writing this. If you care to know more about me, my main homepage is
http://beautywood.org/mungojelly
SUMMARY OF THIS POST
There seems to be a general assumption by the Singularitarian community
that a single human brain is the most relevant level at which to consider
human intelligence. I disagree; it's my opinion that the collective
intelligence of human groups is more relevant both to the past and future
of intelligence. Various ideas follow from this slightly altered
perspective, for instance that technologies are already being included as
participants in social networks and are "rewiring" large parts of society.
I then go on to discuss my theories about what it's like for human and
computer intelligences to cooperate/combine, and how that process is
actually playing itself out in our society.
INDIVIDUAL VS COLLECTIVE INTELLIGENCE
The apparent intelligence of individual humans is mostly an illusion.
Even the most brilliant of humans are lucky to have a single truly
original and useful idea in the course of a day. Most of our appearance
of intelligence comes from our ability to integrate, apply and transmit
ideas that other people have already thought of.
The intelligence of humanity, that intelligence which has been rapidly
discovering the nature of our world, is not of course an illusion. The
illusion is that any particular individual human head is capable of
containing or understanding any substantial portion of that collective
knowledge. We each help to maintain that illusion, by skillfully applying
(and often over-applying) those very few facts out of the immense
ever-growing library of human knowledge which we have been personally able
to comprehend.
In fact, the human brain is only very minimally different from that of
other primates. It's my belief that these small differences do not (as is
often suggested these days) consist of some sort of magic spark which
trips the circuit that flips a brain into having magical abilities like
"intelligence" and "consciousness." It's my belief, rather, that
individual human beings are in their actual mental qualities (as well as
their genetics and so forth) pretty much the same as other primates:
playful, highly social (and sexual), capable of picking up tricks from
those around them and using tools to accomplish tasks, but not able to
routinely have any sort of great insight into how the world works or to
routinely make creative inventive leaps.
The way that we are different is that we communicate. One way to look at
this difference is as an amplifier for the minimal intelligence that
already exists; you may still have that flash of inspiration once in a
very great while, but when you have it you can immediately spread the
fruits to an entire society. Another way of looking at it is to see a
group of verbally interacting humans as a larger unit, as a sort of
virtual computer layered over the substrate of individual thought.
Human societies routinely do complex mental tasks by dividing them between
a large number of individuals. This is a preexisting form of "superhuman
intelligence," if by superhuman you mean exceeding the capacity of a
single human brain. Obviously we have not explored every possible way of
organizing group thought (and new possibilities for how to do so are
coming online so quickly they're hardly news anymore), but the history of
human societies is still a useful reservoir of collective knowledge about
what larger-than-human thought will/does look like.
TECHNOLOGY REWIRES COLLECTIVE INTELLIGENCE
From this stance of seeing human intelligence as primarily a collective
rather than an individual undertaking, we can ask different questions
about what the future shape of technological intelligence is going to look
like. For instance, when asking about the amplification of the
intelligence of a particular individual human, we might wonder about how
we could go about setting up more direct communication between computers
and neurons. The analogous questions on a group level would be about
fitting computers into the social landscape, so their contributions can be
integrated and accepted by others.
Because intelligence is an emergent quality in human groups, it is
susceptible in dramatic and (to us so far) unpredictable ways to
alterations in the patterns of human society. For a number of decades
vast numbers of humans have been jacked in very deeply to mass media;
nobody really knows what the consequences have been. For a few decades,
large numbers of people have been forming virtual communities around
subject areas on the net (like this mailing list), collectively thinking
about various subjects in ways that they have never been thought about
before; nobody knows the consequences. Human societies are being deeply
rewired, and (because this is a level of cognition which is more difficult
for us to understand due to its complexity and more-than-us-ness) the
whole thing seems rather subtle and nebulous to us. The transition is
underway, though, whether we are carefully considering it or not.
YAHOOGROUPS, FOR EXAMPLE
Here's a common social scenario from the beginning of this strange new
century. I was at a meeting recently of men and women from the U.S. in
their twenties. All were reasonably computer literate (they all have
email addresses) but none, save myself, were Singularitarians or hackers
or otherwise familiar with how deep the rabbit hole goes. They set up a
Yahoo group to coordinate the social network they were attempting to
create. The one person (save myself) who was competent to create &
administer a Yahoo group expounded briefly on the virtues of the software,
& particularly the calendar features. Someone asked a question about
receiving email vs reading on the web, and it was clarified that Yahoo
allows you to do either. An informal consensus was reached to use a Yahoo
group as a collective communications tool.
The intelligence of the Yahoogroups software is in many ways quite
minimal. On the face of it, therefore, one might expect that it would
have a proportionally small effect upon the human societies into which it
is integrating. One way to understand why this may not be true is to
replace the question "what is the greatest intelligence which this
software is capable of replacing" with "what is the actual process that
this new intelligence is replacing and how do they compare."
The task of keeping a calendar and reminding a list of people when events
are coming up on it is a task which requires minimal intelligence; it
takes far less than one human brain worth of smarts to do the job. In the
real world, however, there have never before been intelligent calendar
keeping agents; the only way to accomplish the task has been
(metaphorically speaking) to run a sort of virtual calendar computer on a
human brain. Human brains are marvellous objects, but calendar keeping is
not at all their strong point. Without Yahoo's contribution to the
intelligence pool of this group, the closest they would have been able to
come would have been to assign a person to be the calendar keeper; besides
being slow and error prone, the person assigned to that job would probably
find it very tedious.
There's two sides to that fact. One is the obvious conclusion, that
assigning an agent to the (difficult for humans) task of keeping track of
when events are scheduled allows for a whole human intelligence to be
freed from rather a lot of work. A small input of intelligence, given the
right context, can have an amplified output once you get to the level of
actual productive thought. The flipside is a set of consequences that are
rather more difficult to comprehend; the truth is that in the absence of
Yahoo's calendrical competence, since this is an unfunded voluntary
association it's quite possible (even likely) that no human calendar
keeper would have been assigned at all. Meetings would have been attended
by whoever happened to remember; indeed, in previous centuries groups with
regular meetings were simply much more difficult to organize than they are
today. This is changing fundamental qualities of society itself.
INTEGRATION IS SUPERHUMAN INTELLIGENCE
Superhuman intelligence is a human with a calculator. If you come up to
me and I'm holding a calculator, I am capable of doing all of the tasks
which an unaided human intelligence is capable of, plus those that a
calculator is capable of, plus (most interestingly) those tasks which a
combined human-calculator intelligence is capable of (for instance
balancing a checkbook).
I am not coming from the perspective of someone who is in denial and wants
to minimize the strangeness of superhuman intelligence. I am coming at it
from the other side; I am trying to emphasize the strangeness of what's
already happening, of how far down the rabbit hole we have already gone.
Handheld calculators, in spite of the fact that we have grown bored of
them, are in no way normal business-as-usual for the human species. Our
difficult social process of integration with handheld calculators is now
under a very tall layered pile of new intelligences which have found
places in human society. The whole thing is wriggling; it's alive.
I don't think this process of integration can be as prettily charted as
are the trends in computing power. For one thing they are very
qualitative, and so any notion of a "direction of progress" would have to
be very subjectively defined. (Can the intense popular integration with
television sets be considered intellectual progress? What are the social
functions of Tickle Me Elmo?) Nevertheless, I feel from my personal
experience that there is a trend towards greater social integration
between humans and computers, and I have some guesses as to the overall
shape of that trend.
One dramatic feature of the trend is that younger generations tend to be
substantially more integrated with the newest technologies than older
generations. (Somewhere a baby is playing right now with the flash drive
on her mom's keychain, and imprinting far deeper than us the notion of
living in a world with ubiquitous mass storage.) This trend has been
widely noticed, but I'll highlight one counterintuitive consequence that I
believe follows from it: As technology embodies more intelligence, the
young will increasingly come to know more (in many different senses) than
the old.
The other observation I've been making on a gut level about this trend
towards increased social integration between humans and computers is that
the bottleneck is not technological. It's easy for me (or for anyone) to
think of ways that people could use computers more intimately, but haven't
chosen to. The fantastic meteoric rise of computing power translates
itself into new social forms only at a much smaller rate. Everyone who
buys a new computer today is buying a digital sound studio, a digital
image studio, a powerful text analysis tool, the capacity to store a
library worth of text, a graphing calculator capable of instantly drawing
all kinds of beautiful fractals & other complex systems, a sort of digital
aquarium capable of supporting evolving replicating virtual beings, a
formidable chess opponent, etc, & they'll probably just use it to word
process.
It's possible to imagine a society in which no one would think to do
anything except word process. Computers would get more powerful, & no one
would be able to think of anything more powerful they wanted to do, so
chips would be standardized and become extraordinarily cheap and that's
it. Our society isn't quite like that-- we're curious-- but our curiosity
is still tremendously counterbalanced with conservatism and caution.
INTEGRATION COMES IN THROUGH THE BACK DOOR
The next step in this line of reasoning is as follows:
If much more integration is possible than actual with current levels of
technology, that is if people consciously resist deep integration between
themselves and a computer, then integration will continue (assuming that
it satisfies unconscious needs, or otherwise has the memetic hooks to
succeed) but as inobtrusively, unremarkably, & usually as possible.
My feeling of the state of the world today is that we have various devices
almost bursting at the seams, desperately trying to burst out of the nice
safe lines that keep them each to a restricted, tool-like role (phone,
camera, radio, television, watch, clock, exercise machine). My cell phone
knows my schedule; watches know the phase of the moon; my new iPod can
show photos for some reason. Tremendous amounts of intelligence can be
embedded in things without their becoming threatening, as long as the end
result seems "natural" to a user. They don't realize they're talking to
increasingly intelligent AIs; they just know that search results seem more
useful than they used to, or that computers can recognize voices better
than they used to. To an end user (even a well informed one) it just
isn't that shocking for computers to recognize the human voice; rather,
it's seemed odd the past few decades that they haven't been able to.
Recognizing what someone says seems easy enough subjectively.
Each person subjectively examines what it's like to think, and finds that
it consists of various easy tasks (recognizing what someone says, adding
one plus one, etc) and also various harder tasks. The hard tasks are what
make up the bulk of what we worry about. It seems to a naive person
therefore that if you were going to go about building an AI, you would
find it easy to blow through the simple stuff like being able to talk, and
then you would hit a very difficult wall when it came to harder stuff,
like being conscious of your own emotions. So it's not very threatening
if computers can do things like pick out a voice from a crowd-- "So what?
That's not very hard anyway, is it?!"
So yes I do think that society will be blind-sided by AI. People are
going to have articles explaining why AI is impossible automatically
delivered to them based on their click history, and they will read them
and enjoy them. That's not necessarily a limiting factor-- nobody planned
out the transition to multicellular life and that went OK-- but it's going
to make the whole shape of things really weird. "Hyperintelligent
toasters" type of weird.
ONE LAST THING: RAISING THE ROOF
One last thing: The shape of emergence of computer intelligence is not
going to be anything like our intuitions that we draw from the emergence
of biological intelligence. Our biological intelligence emerged very
slowly by incremental changes, methodically exploring not the whole space
of what's possible to compute, but rather the space of what computations
are sufficient to keep a monkey alive long enough to fuck. Electronic
intelligence is not going to emerge gradually through generations of whole
cohesive system. It's not a baby.
What the new intelligence is going to do is totally blast the roof off of
one domain after another. They beat us first at adding numbers; they
completely utterly crushed us. They didn't rest on their laurels, though;
they crush us more & more at adding numbers every year. They got better
than us at Chess, at first just barely, but they have gotten better every
year (faster, in fact, than the rate of hardware advances alone). Voice
recognition software is on track to become truly fluid over the next few
years, but it will not stop there; computers will be able to make out a
whisper from across the room, pick out one voice out of a chattering crowd
of thousands, understand the world's heaviest Scottish accent, etc. They
will be able to record every conversation they hear and index it against
every other conversation they have ever heard. They'll know exactly when
the last time they heard the word "mulberry" was.
On one side of the transition there are just humans who think in human
ways, and on the other side is the impossible-to-conceive strangeness of
true intelligence. What's interesting is what's inbetween; it's not, as
you might first guess, a being with intelligence which is uniformly
improved to half way between the Human and Power levels. Instead what's
between the two is a creature who has half of the Powers-- who can compute
fractals instantly, record sounds and play them back, use Bayesian spam
filters, communicate verbally over continental spans, access information
about essentially any subject on a moment's notice, etc-- a creature
trying to survive an astounding and unexpected metamorphosis, by which I
mean, of course, us.
CONCLUSION AND SIGNATURE
Those are my thoughts so far. Criticisms are, of course, welcomed.
<3,
mungojelly
This archive was generated by hypermail 2.1.5 : Tue Feb 21 2006 - 04:23:04 MST