The Brin/Kelly challenge

From: Eliezer S. Yudkowsky (sentience@pobox.com)
Date: Wed Apr 05 2000 - 16:46:31 MDT


David Brin wrote:
>
> ABSTRACT: What daring 21st century concepts or projects would you most
> like to see pursued, if money were no object?

Kevin Kelly wrote:
>
> So here are the rules. You have a billion dollars in cash and you have to
> spend it all in five years. You want maximum good. What do you do?

Originality is the key to effectiveness. Giving ten million dollars to
a well-funded, popular, obvious Cause is a relatively small contribution
to a vast network of forces, all pushing and pulling and
counterbalanced. In concrete terms, if you try to send food parcels to
some starving country, the packages will be confiscated by the squalid,
dictatorial governments that created the famine in the first place. The
effort doesn't change the rules of the game. It treats the symptoms
rather than the problem. It's mundane, and therefore doomed to failure.
It is an application of brute force rather than intelligence, throwing
money at the problem. And above all, someone has already tried it.

There's a place for community action and brute force, of course. It's
usually better than nothing, and sometimes it works. But it's also
important to spend something on originality and intelligence. To make a
major impact, to make a global change to that vast network of forces,
you need to push in an entirely new and non-counterbalanced direction,
or change the structure of the network, or find a first cause that can
be changed with small amounts of force.

The other rule of effectiveness is to work with information rather than
physical objects. To distribute food packages to a billion people, you
would need a million people handing out packages (even if the packages
aren't confiscated). But one research team in one laboratory can create
an innovation used by a billion people. That's how you change a
hundred-trillion-dollar planet without spending a hundred trillion dollars.

For one project to make a real impact on the world, where others have
tried and failed, you must wield either substantially more money or
substantially more intelligence than ever tried before. If someone else
has tried it before with more money *and* more intelligence, then
there's no prospect of a big win, no prospect of permanently solving the
problem, no prospect of giving the world such a giant smack that it
wobbles in its orbit.

There's a place for together-we-make-a-difference, but there's also a
place for originality, especially in Silicon Valley. Silicon Valley was
not founded on the communal application of huge amounts of brute force,
but on individuals and brilliance and changing the rules. If Silicon
Valley doesn't fund originality, who will?

And sometimes, brute force just doesn't work. You can't make hunger go
away by giving people food. You can't make poverty go away by giving
people money. It's been tried. People are still trying it. Looking
over the problem, one gets the terrible sinking feeling that the
philanthropic efforts are not merely linear but saturated. It's not
just that huge efforts are required to produce huge results; even huge
efforts will produce medium-sized or small results, because everything
that can be done is being done.

On some level, humans instinctively approach philanthropy the way our
hunter-gatherer ancestors would have distributed a surplus of coconuts.
People push huge amounts of resources around and create small ripples
that eventually die out, creating nothing but the illusion of
accomplishment. This is the pessimistic view, of course, but it's true
often enough to be depressing.

It isn't that technological research or untried ideas are innately good;
they are, however, innately powerful. There is no way to change the
world, for good or evil, without them.

Of course, genuine originality and creativity aren't easy. To create a
major impact, it's necessary to step outside not just the specific
attempts, but the formulas. The category of "selling some random
object, but OVER THE INTERNET!" has been exhausted by too much venture
capital chasing too few ideas; even if nobody has ever previously
attempted to sell neon-green swimwear to left-handed grandmothers,
NGSLHG.com won't change the world. And yet the NGSLHG.coms of the world
- and the NGSLHG.orgs - still get most of the funding, are even touted
as revolutionary.

Why?

There is no formula for Jumping Out Of The System - JOOTSing, as
Hofstadter puts it - unless you count human intelligence as a "formula".
People who are smart or even brilliant often have only a handful of
truly original, workable ideas. (Not the interesting and vaguely
plausible ideas that any modern-day Renaissance Man (or Woman) can turn
out by the dozen, but ideas original enough to occupy uncharted
territory in "company space" or "nonprofit space", ideas workable enough
to be turned into startups.) The incredibly few people who come up with
dozens of truly original workable ideas can do so because they've
invented a Meta-Idea, a truly original and useful way of seeing - at
most, a handful of original and useful ways of seeing.

Still, with a population of six billion, intrinsic scarcity can't
explain the problem. But people rich enough to be philanthropists (or
venture capitalists) can't listen to six billion people. They are
surrounded by ten thousand people eager to be their very best friend.
Most of the time, rich people can hobnob only with equally rich people -
who are hopefully free of the desperate drive for next week's lunch - or
with people they already know. That closed society is a tremendous
inconvenience from the perspective of a non-rich guy with a brilliant
new idea - but from the perspective of the rich, it satisfies the need
for simple human contact.

This particular phenomenon, which acts to restrict the pool of
idea-generators, is a special case of the general problem: Modern-day
philanthropy and venture capital are not optimized for finding original
ideas. Or rather, and this is the real killer, the system is optimized
for a great many things *other* than originality - as one would expect,
when large sums of money are involved. Business cases are optimized for
how plausible they sound to venture capitalists. Startup teams are
optimized for the impressiveness of their resumes.

Committees make compromises. Blame is concentrated while success is
shared. Nobody gets fired for buying IBM. Contradicting anyone is
impolite and contradicting the boss is unthinkable. "Yes" is a career
liability while "No" is safe, and it only takes one "No".

Anyone who reads _Dilbert_ or _Wired_ knows about those problems, and a
great many Silicon Valley managers are doing their best to solve them.
But there are deeper, stronger causes, visible only to the dark arts of
cognitive science. Charisma is valued highly, but does persuasion use
the same abilities, the same modules of the brain, that underlie
rational analysis? "Energy" and "enthusiasm" appear on every job
description, but do these positive emotions - the emotions experienced
when things are going *right* - invoke originality, self-awareness,
finding logical flaws, jumping out of the system... or do they invoke
minor tweaks, skill formation, creeping featurism, doing things the same
way with minor modifications?

There is a famous story (1) about an 18-month-old female monkey named
Imo, undoubtedly the smartest monkey on her island. She discovered how
to wash potatoes, and later, how to separate grain from rubble. She
taught her mother, and her playmates, and her playmates taught their
mothers, and within a few years most monkeys on the island were washing
potatoes (and later, separating grain), at great benefit to the tribe.

But did Imo gain any evolutionary benefit, any *differential*
reproductive success, from her discovery? All the other monkeys were
also washing potatoes; why would Imo have been more successful than
anyone else? The dominant selection pressure did not favor inventing
new things, but imitating others' successes. That's what we've evolved
to do. Thus most venture capital consists of jumping on bandwagon after
overloaded bandwagon, most startups receiving funding are slightly
different permutations of comfortably fashionable ideas - but so long as
the illusion of accomplishment exists, why would anyone do things differently?

Kevin Kelly wrote:
>
> Obviously, these people have problems we could only wish for. But rather
> than a joke, this manager convinced me that handling a billion dollars was
> actually a difficult problem. Even giving away a billion dollars, he said,
> was extremely difficult -- the main problem being that you couldn't give it
> away responsibly fast enough not to have the remainder keep piling up. If
> you give it away too slowly (at less than 10% of capital per year), the
> principal accumulates gains faster than you spend it. On the other hand
> to give it all away at once would require buying such an elaborate
> infrastructure to handle it that most of the money would go towards the
> infrastructure and not the problem. The question we posed was, if you had
> to give a billion dollars away fast and effectively, what would you do?

In a sense, Kevin Kelly and the manager are talking about two different
problems. The manager is pointing out that even if you can find a
hundred ideas worth $10M apiece, you still need to find a hundred
Chairs, a hundred Executive Directors, and several hundred Board
members. Perhaps the infrastructure problem could be solved by an
Idealab-like "nonprofit factory", which I don't think would take "most
of" a billion dollars... but I'm getting off the subject. The issue
Kevin Kelly has raised is finding something *useful* to do with a
billion dollars - something original, practical, and above all *effective*.

Here are eight for-profit and non-profit samples:

1) Institutes of Verification.

In the current school system, the institutions responsible for teaching
are responsible for certifying that knowledge has been acquired; this
verification usually takes the form of a few text-based tests plus the
fact that a warm body was in the classroom. This leads to a short
circuit, the phenomenon where people go to college not to *learn*
anything but simply to get a degree so that they can get a job. It
results in warm bodies showing up in the classroom and last-minute
cramming for exams, educations optimized for getting a sheepskin instead
of learning. From elementary school to college, it creates an uncaring,
bureaucratic attitude that saps the will of students and teachers.

If separate, specialized Institutes of Verification existed to certify
knowledge acquisition, it would break the short circuit. Starting with
computer programming and sciences, an Institute could administer an
extensive multi-week examination that would include real-time
programming tasks or laboratory work. The resulting credentials would
provide superior proof of skill acquisition; it would - by taking the
verificational function away from colleges - require that students pay
attention and acquire useful skills rather than simply showing up to
class; and would eventually give rise to alternatives to assembly-line schooling.

2) Silicon Valley "Don't Tread On Us" Political Action Committee.

The technology sector needs a lot more political influence, right away,
before we turn into prey for every class-action lawyer and a target for
every Congressional busybody. The problem is that the usual method of
gaining political influence - buying it - exacerbates the problem. The
immigration-visa blockade is probably continuing at least in part
because of the enormous windfalls represented by all the lobbying that
goes on every year. And with incumbents already winning 98% of primary
elections, it'd be a bad idea to put more money in the hands of existing officeholders.

So, invert the usual strategy. Create a "Technology Hit List" composed
of any Congressperson who gets in Silicon Valley's way, especially for
really obnoxious behavior like the Communications Decency Act or
encryption export restrictions. If a politico goes on the list, one
thousand donors will send $1K apiece to whichever primary opponent is
the most technophilic or libertarian, do it for the general-election
opponent if the politico survives the primaries, and run thousands of
issue ads. Let the word be passed in Congress: "If you want an easy
election, you don't get in Silicon Valley's way." If SVPAC can get a
good reputation for vindictiveness, the technology sector might wind up
with at least as much influence as, say, the Teamsters.

It bypasses the Danegeld problem - and if everyone does it, there'll be
a lot of new faces in Congress, which would be no bad thing for America.

3) Prevoting sheet / active voting booths.

Computer-assisted democracy seems like a no-brainer, and one wonders
that it hasn't happened already. A week before election day, I want to
sit down at my computer, get a simulated election booklet, read what the
candidates have to say about themselves and each other (as obtained from
some open directory), browse the voting records of any incumbents, make
a decision, and print out the answers. Then when I go down to the
polls, I can just punch in the answers off of my printout, knowing who
everyone is and knowing all the issues at stake, instead of skipping
half the items because I've never heard of the candidates. Incumbents
aren't even marked on the ballots we have now, making it difficult for
me to implement my anti-incumbent voting strategy.

There are existing dot-coms and dot-orgs trying to offer this service,
but they usually stop at the state level, and they don't offer bios for
all the candidates. What we really need are Web browsers in the polling
booths, or at least explanatory booklets offering each candidate a
paragraph to make a statement, but try and get that one past Congress.

Nonetheless, if some well-funded nonprofit were to take on the task of
completely duplicating every ballot in the country in every election, it
would probably be a very, very good thing for democracy in America.

4) Rationality training via neuroimaging.

We have evolved, for millions of years, to deceive ourselves.
Self-deception is an advantage in political debate, and political debate
is one of the major determinants of social dominance in the ancestral
environment. From an evolutionary perspective, it's far more important
to be persuasive than to be right, and to be really persuasive you have
to really believe what you're saying.

But despite all the evolution that goes into keeping ideas looking
plausible internally, the brain would not have evolved to hide
self-deception from fMRIs. fMRIs weren't around in the environment of
evolutionary adaptedness. And self-deception is so important,
evolutionarily speaking, that it's probably got a few specialized neural modules.

If people can train themselves to think without rationalizing away
inconvenient problems, without hiding things in the corners of their
mind - if people can train themselves to recognize the built-in
stupidity modules by watching a real-time fMRI scan of the brain - it
would be the greatest boon to human rationality since the invention of
the scientific method.

It might not work, but it definitely deserves a research project.

5) New corporate structures.

The modern corporation has a hierarchical social structure similar to
feudal government. Democracy works better than feudalism in government,
and maybe the same will hold true of corporate structures. Perhaps a
corporation needs enforceable rules to survive, so that everyone doesn't
just play Quake all day, but maybe the enforcement shouldn't take the
form of a boss for every individual. Maybe a much smaller "police
force", specialized for enforcing company rules and reasonable levels of
productivity, would result in less office politics.

The modern corporation keeps internal data secret to protect it from
competitors. Maybe managing with open-source books would yield greater
benefits - a better understanding of the company by employees, more
confidence in the company, more interested and "bought-in" shareholders.

There are a lot of features of companies that seem like either leftover
features from the industrial revolution, or pitfalls of human
psychology. R&D in the area of corporate structures is a high-risk
venture, especially if you actually try the ideas out - but the
potential benefits are enormous.

6) Ubiquitous venture capital.

The modern stock market is not investment. It's a bubble, and we all
know it. You try to get rich by giving money to someone who owns a
stock, then hope someone else will give you more money for the piece of
paper later on. Dividends and stock buybacks are practically forgotten,
and acquisitions are usually paid for in more stock. All that's left is
the zero-sum gamble. The stock-market bubble doesn't use money to make
money, except in the indirect and relatively inefficient sense that a
high stock price is good for a company.

Venture capital uses wealth to create new wealth, directly. It has
higher - and more reliable - returns than the stock market. And, right
now, this investment vehicle is only available to extremely wealthy
individuals and large institutional investors. Everyone else is forced
to gamble on a pyramid scheme.

Venture capital needs to democratize. Investors have poured billions,
perhaps trillions, into creating a stock-market bubble, benefiting only
the earlier tiers of the pyramid scheme - money that should have gone
into creating new companies, new wealth. Oh, sure, all that money has
had some positive value for the economy. Some of it was actually
invested by the elite investors permitted to get in on IPOs. It's let
dot-coms survive on stock prices. It's pumped up IPO returns, driving
an increase in venture capital. It's even created a "wealth effect"
that's lifting up the general economy. Still, it would be vastly more
efficient, and vastly safer for the economy, and vastly better for the
small investors, if that money could be invested directly. Venture
capital *must* democratize.

What's needed is not a venture capital company with publicly traded
shares; this is nothing new, since "investing" only sends the money to
the previous shareholder. What's needed is a company where anyone can
invest five dollars, and watch it go directly into the hands of some
startup, or a broad spectrum of startups to reduce risk.

The current surge in venture capital could be just the earliest
beginning of a vast exponential growth that will see a substantial
sector of the entire economy, hundreds of billions of dollars, devoted
to venture capital. Venture capital, once democratized, will continue
to grow until the returns drop to 7% a year. Or, since returns on
venture capital are limited only by the quality of available investment
opportunities, maybe returns will never go below 20% a year. Maybe
banks will offer venture-capital savings accounts. *That'll* send the
economy into hyperdrive.

Venture capital is the best, lowest-risk way to use wealth to make
wealth, with a maximum of investment and a minimum of gambling. The
only reason venture capital is limited to big investors is a silly
pre-Internet tradition.

7) Genesis Seed Fund & Idea Broker.

Venture capital (discussed above) requires high-quality investment
opportunities. An "investment opportunity" consists of a good team and
a workable idea. Oddly enough, the current situation in venture capital
already consists of "too much money chasing too few ideas". You
wouldn't think a planet with six billion intelligent entities could run
out of ideas.

And it hasn't. But right now, it's impossible to get rich off an idea
unless you have the experience, credentials, and unique high-energy
charismatic personality necessary to start your own company. But many
of the most creative and visionary people, perhaps a majority, don't
have the energy required to work 16-hour days. If people could profit
from an idea without actually starting a company, it would unlock a huge
informational resource.

What's needed is a mechanism for identifying ideas, plus an Idealab-like
startup factory. If an individual's idea is accepted, it's added to a
pre-assembled proto-startup containing a CEO and other individuals (who
have applied to be employed by a Genesis Company). Venture capital is
added from (ubiquitous) venture funds that have applied to fund a
Genesis Company. The visionary goes on the Board of Directors and gets
5% of whatever the CEO gets (in salary, stock, and options).

Obviously the key problem is finding people who can recognize good
ideas, and creating a system that lets them do so. Initial versions
might restrict the eligible suggesters to a knowably brilliant
population, although you'd have to be very careful not to exclude
brilliant 16-year-olds. (No, I'm not 16.) Maybe limit it to the
membership of Foresight circa April 2000, or impose a minimum IQ (or
some other elitist thing), during the test period.

I'd have to come in as Genesis Director on this one, since there are
some cognitive issues. It's more complicated than I'd want to explain here.

8) Collaborative filtering.

Take a few million people. Let them rate their favorite Webpages (and
their least favorites). If A has the same pattern of favorable and
unfavorable ratings as B (for those webpages they've both visited), then
suggest A's favorite sites to B and vice versa. That's collaborative filtering.

Unlike the other ideas, collaborative filtering is not my own invention.
There are even companies trying to implement it. However, I think
they're all doing it totally wrong. They're using collaborative
filtering to sell things, meaning that the user base and prediction base
are tiny, and that you need to spend money to find out whether the
prediction is right. To work, collaborative filtering needs millions of
people and billions of ratings. You need a target domain like the Web,
and then you need an inline "rating" button in Mozilla (and preferably a
patch for IE, but you could get by initially on just Mozilla).

Collaborative filtering, done *right*, has the potential to have an
impact even greater than that of the Internet. It could disintermediate
the entire marketing sector of the economy, replacing advertiser push
with customer pull. It could expose millions of people to the great
ideas now contained only in small groups (libertarianism, transhumanism,
your favorite overlooked comic strip or whatever). It could create a
totally new experience where you go on the Web and see nothing but
really good ideas and really interesting pages for hours at a time, and
then go on and do the same thing tomorrow. It could double the
effective IQ of First World communities - and that could be very very
important in the near future.

9) This Space For Rent

Are you a Silicon Valley philanthropist looking for an original,
thinking-out-of-the-box solution to your (least) favorite problem?
Email me at sentience@pobox.com.

--
There's so *much* that needs doing!  Virtually every component of
modern-day society, from schools to standardized currencies to bank
accounts, is in some way a relic of the industrial revolution or
earlier.  The social and economic forms of the past are uniformly
distorted by lack of computing power.  So many things need to be
upgraded to prepare for the changes ahead...
Still, if you were to hand me $10M, all those projects would need to
wait their turn.  There's a higher priority, a project vastly more
original, more novel, more effective, more "outside the box" than any
innovation in human history.  That project is the creation of
human-equivalent or transhuman artificial intelligence.  Every unsolved
dilemna in human history, every problem we've been unsuccessfully trying
to solve for the last ten thousand years, is a problem that *humans*
have bounced off of.  Will AIs succeed where we have failed?  For a
qualitatively different intelligence - who knows?  For a qualitatively
smarter intelligence - almost certainly.
Because a thousand children fail to solve a problem, you can't conclude
that an adult will fail.  Because a thousand Dark Age theologians
failed to solve a problem, you can't conclude that a modern scientist
will fail.  Because a thousand novices fail to solve a problem, you
can't conclude that an expert will fail.  Because a thousand nongeniuses
fail to solve a problem, you can't conclude that an Einstein will fail. 
And because humanity has failed to solve a problem, you can't conclude
that artificial Minds - not "computers", any more than a human is an
amoeba - will fail.  There is a basic gap, a divide that makes
everything new.
I believe that within the next few decades, it will become possible to
technologically create smarter-than-human intelligence.  I believe that
this is a great good, that it means a wholly new chance to solve any
human problem - perhaps *all* the problems.  And I'm in a hurry, because
I believe that strictly human intelligence won't be able to handle the
destructive applications of other technologies I expect in the near future.
There are others who believe the same.  We call ourselves the
Singularitarians, after Vernor Vinge's use of the term "Singularity" to
describe the fundamental breakdown of the model that occurs with the
rise of smarter-than-human intelligence.
*That* is Jumping Out Of The System.  Not enough original ideas?  Create
a better thinker.  It's thinking outside the *human* box.  The
Singularity taps a force that might be stronger than the entire network
of pushes and pulls, a force stronger than a hundred-trillion-dollar
planetary economy... and it all starts with a computer program.  Now
that's what I call cost-effectiveness.
If you were to hand me a billion dollars, I would spend as much of it as
possible on accelerating the development of an AI, using the novel AI
architecture described in "Coding a Transhuman AI" (353K).  Then, I
would spend as much as possible on the general development of a
Singularity Institute and the other projects described in "The Plan to
Singularity" (404K).  If this still left more money than could usefully
be expended - as seems likely with $1G - there are thirty or forty
for-profit and non-profit projects that could use seed funding, eight of
which are described here.  Getting someone to run each of them would
still be a problem.
"Coding a Transhuman AI"
	http://singinst.org/AI_design.temp.html
"The Plan to Singularity"
	http://singinst.org/singularitarian/PtS/plan.html
--
The reason I'm writing this letter is not to ask for funding for the
Singularity Institute.  I'm not even writing in the hope that someone
will create the first Institute of Verification, or incorporate "Genesis
Idea Broker, Inc.".
As Silicon Valley billionaires turn to philanthropy, I hope to see a
tradition of originality, of going beyond the mundane.  I hope to see
more Internet millionaires living up to the standard for creativity set
by Michael Saylor's free online university, or the standard for leverage
and global impact set by Steven and Michele Kirsch.  I hope to see it
become a rule in Silicon Valley philanthropy:  For every $100M spent on
mundane philanthropy, spend $10M on thinking outside the box.
I hope to see charity for technophiles... technophilanthropy.
Sincerely,
Eliezer S. Yudkowsky.
-- 
       sentience@pobox.com      Eliezer S. Yudkowsky
          http://pobox.com/~sentience/beyond.html
                 Member, Extropy Institute
           Senior Associate, Foresight Institute
--
Footnotes:
(1)  There is a myth called the "hundredth monkey" theory which involves
Imo, to the effect that the potato-washing behavior jumped to other
islands through telepathy.  This story was invented afterwards, and has
been debunked by the _Skeptical Inquirer_.  Imo herself is real, and the
events discussed above are part of the scientific literature.


This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:27:52 MST