The final book of rationality
This is part 6 of 6 in my series of summaries. See this post for an introduction.
Part VI
Part VI
Becoming Stronger
T
|
his part asks: how can individuals and
communities put all this into practice? We look at an autobiographical account
of Yudkowsky’s own biggest philosophical blunders, as well as ways to develop
evidence-based applied rationality curricula and institutions.
This final part is less a
conclusion than a call to action and a jumping-off point for further
investigation. For readers who want a fuller understanding of normative
rationality in terms of Bayesianism, books like “Thinking and Deciding” (by
Baron) and the “Oxford Handbook of Thinking and Reasoning” deal with cognitive
science and heuristics and biases. Regarding decision theory and philosophy of
science, “Good and Real” (by Gary Drescher) reaches conclusions similar to
those of Yudkowsky. The Stanford
Encyclopedia of Philosophy has entries on Bayesian epistemology and
naturalized epistemology.
Note that Bayesian and
“frequentist” data analysis methods can both be useful when used correctly, and
training in statistics can improve reasoning skills outside the classroom. But
the Art is still in its infancy, so we have to ask: what’s missing? What should
be in the next generation of rationality primers? Whatever comes next, there is
certainly no shortage of global challenges, and the art of applied rationality
is a new and half-formed thing. There are not many rationalists, and there are
many things left undone.
24
|
Yudkowsky’s Coming of Age
|
This
chapter provides an in-depth illustration of the dynamics of irrational
belief, by spotlighting Yudkowsky’s own intellectual history (over a rough
time period of 2000 to 2003), with advice on how he thinks others might do
better.
|
Young Eliezer (around 1996) had thought that
intelligence was more important than anything else, and thought that it was
essential for ethics and wisdom – contrary to his parents’ insistence that he
just needed life experience. So he thought that superintelligence implied
“supermorality”. He wanted to spend his life creating the Singularity, out of a
sense of duty to give IQ points to everyone. This was a happy death spiral, and
he started to believe that even the light-speed limit would be no barrier to
superintelligence. In some ways he fell into the trap of thinking reversed
stupidity is intelligence.
When Eliezer went into his death spiral around
intelligence, he wound up making a lot of mistakes that later became very
useful. Young Eliezer (between 1996 and 1999) refused to formally define
“intelligence”, because he saw it abused so often in the field of artificial
intelligence. This was his worst mistake (because you cannot fully trust
informal reasoning), but also his best mistake, because it led him to study
lots of cognitive sciences which helped him recover from his mistakes. This
lesson is that what you actually end up doing screens off the clever reason why
you’re doing it.
One major childhood influence on Yudkowsky was
Jerry Pournelle’s “A Step Farther Out”, which portrayed scientists and
engineers as the Good Guys. Eliezer grew up as a technophile, allergic to
people who said “technology has benefits and risks” and who advocated
regulation… It took him a very long time to get to the point where he was
capable of considering that the dangers of technology might outweigh the
benefits. This happened when he realized that molecular nanotechnology would
pose an existential threat, unless we developed AI before it.
In 1996, Yudkowsky encountered a transhumanist
mailing list where someone said that “no one should develop an AI without a
control system that watches it and makes sure it can’t do anything bad.” Young
Eliezer was really good at refuting others’ arguments against his intuition
that a superintelligence would know better than we what is right. His skills at
defeating other people’s ideas led him to believe that his own (mistaken) ideas
must have been correct. But it’s easier to find flaws in someone’s argument
than to get the fact of the matter right.
In 1997, Yudkowsky set out to argue inescapably
that creating superintelligence was the right thing to do. Young Eliezer took a
mysterious view of morality, and so he had lax standards of rigor in defining
“morality” or “intelligence”. This was his big mistake. But Nature doesn’t care
about righteous excuses; if you don’t meet the standard, you fail. You can’t
manipulate a confusing gap in your understanding. In the absence of precision,
you might as well be putting your weight down on a landmine. No matter how
clever your justification, it will blow your foot off just the same.
Nature doesn’t care about your clever
justifications; what matters is your choice to work the problem out in detail,
such that you may accumulate experience. This is what Eliezer did in 2000 when
pondering how to inscribe a fallback morality into AI. He started to dig himself
out of his philosophical hole when he noticed a tiny inconsistency: even in the
case that life is meaningless, maybe some people would prefer an AI to do
particular things, such as not kill them. Slowly, over succeeding years,
Eliezer started to think inside the
black box of morality.
Only when Yudkowsky permitted himself a line of
retreat (i.e. Friendly AI as contingency plan) was he able to reconsider his
positions in his metaethics, and move gradually towards better ideas. Young
Eliezer (in 2001) abandoned the idea that AI can’t be dangerous, but he still
wanted to charge in, guns blazing, with coding. Yet what he needed to do was
declare “halt, melt, catch fire” and scream “oops!” – instead of slow little
shifts in opinion. In the art of rationality, it is far more efficient to admit
one huge mistake than to admit lots of little mistakes.
Eliezer awoke when he understood intelligence
as an optimization process, squeezing the future into a constrained region by
exerting thermodynamic work. This broke him from human-ish mind designs, and he
comprehended the true risk when he looked back and realized his mistakes.
“Smart” was no longer a property, but an engine, which could pump reality toward
any outcome. He finally admitted to himself that his old AI goal system design
would have wiped out the human species by converting its future light cone into
generic tools like stored energy without a use, or computers without programs
to run.
There are people who have acquired more mastery
over various fields than Eliezer has over his. Eliezer considers the Bayesian
probability theorist E.T. Jaynes and the mathematician John Conway to be above
his level in terms of mastery and perhaps brilliance, but he aspires to that
level. Modest demeanors and humble admissions of doubt are cheap. Eliezer still
thinks he can do important things in his chosen field. Yet he is humble enough
to have invested specific efforts into the possibility that some younger mind
reads his blog and zips off right past him.
It seems to be an uncomfortable truth that the
elite of the upper echelons in business, science and so on (like CEOs and
hedge-fund traders) really are more intelligent, competent and happy than everyone else. However, most
people who want to work on Artificial General Intelligence don’t speak fluent
Bayesian and aren’t even on the level of Peter Norvig or John McCarthy; they
are mere
above-average
scientists, who aren’t formidable enough as individuals to synthesize true AGI.
They aren’t really all that exceptional, and this is a problem which most
people don’t seem to see.
Eliezer considers his training
as a rationalist to have started the day he realized just how awfully he had
screwed up. Eliezer in late 2002 realized that he needed to actually update and
admit that he didn’t know how to do AGI yet, despite the loss of status; and he
realized that Nature was still allowed to kill you even if you had clever
arguments for taking a risk. You could do everything that you were supposed to
do, and Nature was still allowed to kill you. The Future is not indestructible.
Yet other wannabe AGI researchers still care more about being first than about safety.
After Eliezer realized that optimism had misled
him, he invented a thought experiment. Compare the world in which there is a
God, who will intervene at some threshold, against a world in which everything
happens as a result of physical laws. You could simulate a universe in which
the mathematical result is a world with conscious life, where beings suffered
from diseases unfairly. Whatever physics says will happen, will happen, good or
bad. Which universe looks more like our own? We live in a world beyond the
reach of God; thus the Future is vulnerable. People believe that some things
are not allowed, or that “things have to make sense”. If you want to be happy,
meditating on the fragility of life and the unprotectedness of your existence
probably won’t help; but what if you have something to protect? Nature is utterly
neutral, not fair, and won’t prohibit horrible things from happening. Nor can
you trust in technology, democracy, or positive-sum games. Injustice is allowed
to happen. That is what Eliezer hopes to fix – but what does a child need to do
to solve an adult problem?
Yudkowsky’s mathematical intuitions were always
Bayesian, but reading Tversky and Kahneman (“Judgment Under Uncertainty”), E.T.
Jaynes (“Probability Theory: The Logic of Science”) and Judea Pearl (“Probabilistic
Reasoning in Intelligent Systems”) helped him level up. He realized that
precision, though inconvenient, can save time (because you arrive at the only correct answer). We should hold
ourselves to the standard of mathematical proof. The prospect of saying “Oops”
in the future should make you feel alive, because it means that you’ll acquire
new Jedi powers that your present self doesn’t dream exist.
›
25
|
Challenging the Difficult
|
This
chapter asks what it takes to solve a truly difficult problem – including
demands that go beyond epistemic rationality.
|
In Orthodox Judaism, knowledge derives from the
authority of ancient rabbis, and thus Torah loses knowledge in every
generation, as time passes. Tsuyoku
naritai is Japanese and means “I want to become stronger”. It expresses the
will to transcendence. It is about continuous progress, gaining knowledge from
science rather than authority, and becoming less biased, instead of taking
pride in confessing your ignorance. You should aspire to become stronger, and
study your flaws so as to remove them. The temptation to be satisfied in
confessing your biases can impede progress.
In the ancestral environment, successful
hunters would downplay their accomplishments to avoid envy. Hence there are
evolutionary-psychological factors that encourage us to signal modesty and
mediocrity. However, tsuyoku means
always reaching higher, without shame, even if you pull ahead of the crowd.
Sooner or later, if you aim to do the best you can, you will set your aim above
the average. You should be able to admit to yourself that you’ve done better
than others, without being ashamed of it – it can even be a useful motivator.
Yoda famously said: “No! Try not! Do, or do
not. There is no try.” There’s a difference between “I’m going to flip that
switch” and “I’m going to ‘try to flip’ that switch” – the latter means you’re
going to try to “try to flip the switch”. Trying
to try, and being satisfied with a plan, is too easy. You should actually put in the effort to win. As a
human, if you try to try something,
you will put much less work into it than if you try something. It’s only when
you want to, above all else, actually flip
the switch without consolation prizes just for trying, that you will actually put in the effort to actually maximize the probability. Many
of life’s challenges consist of holding yourself to a high-enough standard.
Instead of asking “what can I do?”, ask “what needs to be done?”.
In the original Star Wars trilogy, Luke
Skywalker comes across as a whiny teen. Imagine the following fictional
exchange between Mark Hamill and George Lucas over the scene in The Empire Strikes Back where Luke
attempts to lift his X-wing with the force: Mark Hamill doesn’t want his
character Luke to give up on raising his X-wing out of the swamp, as he didn’t
find it compelling; but George Lucas replies that the audience will buy it,
because human beings wouldn’t try for five minutes before giving up. As John
McCarthy said, “When there’s a will to fail, obstacles can be found.”
A lot of projects seem “impossible” in the
sense that we don’t immediately see a way to do them. (Of course, confusion
exists in the map, not in the territory.) If something seems impossible, you
won’t try; but important problems only look less intimidating and confusing if
you persevere through difficulties enough that you understand the domain (if
you have the native ability). After working on them for a long time, these
impossible problems will start to look merely extremely difficult. Trying to do
the impossible is definitely not for everyone, and learning when to lose hope
is an important skill in life. But if there is something you can imagine that
is even worse and scarier than wasting
30 years of your life (like unfriendly AI), then you may have cause to attempt
the impossible; in that case, don’t give up instantly at the very first sign of
difficulty, and keep working even though you could be getting higher personal
rewards elsewhere.
It takes an extraordinary amount of rationality
before you stop making stupid mistakes. A “strong” effort usually results in
only mediocre results. Doing better requires making extraordinary efforts. The
level beyond tsuyoku naritai is Isshoukenmei – which means “make a
desperate effort”. This is the gestalt of trying your utmost, as if your life
was at stake. Extraordinary efforts require you to bypass the System. This can
be dangerous, but humanity won’t survive Nature’s challenges unless some of us
think or act outside our comfort zones.
Always keep improving, make a desperate effort,
and leave your comfort zone. But sometimes even isshokenmei will not be enough. When the problem is “impossible”,
don’t aim to try your best, aim to win! The ultimate level of attacking a
problem is the point at which you simply shut up and solve the impossible
problem. An example of something “impossible” that Yudkowsky has accomplished
is the “AI Box Experiment”, where he pretended to be an AI sealed in a box, and
persuaded a gatekeeper to let him out. Understand the reasons why you can’t succeed, then shut up and
do the impossible! You have to, without doublethink, hold the awful tension of both
views in your mind at once – seeing the full impossibility of the problem, and
really intending to solve it. This should be reserved for very special
occasions. And remember, you can lose,
and it will hurt.
This is the conclusion of the Beisutsukai series. Jeffreyssai says
farewell to Brennan and his fellow students, and tells them that they’re done
for now, but that “the rhythm at the center of everything is missing and
astray.” The way to arrive at mastery is by using to the fullest the techniques
you have already learned until they shatter in your hands. You must be
determined to remake your art in the
midst of a wreckage of a surprise catastrophe. You must avoid the flaws of
motivated skepticism, excess cleverness, and underconfidence. And you must ask
yourself what you really want.
›
26
|
The Craft and the Community
|
This
final chapter on individual and collective self-improvement discusses
rationality groups and group rationality. It raises questions like: can
rationality be learned and taught? What community norms would make this
process of bettering ourselves easier? Can we effectively collaborate on
large-scale problems without sacrificing our freedom of thought and conduct?
|
Religion is harmful, but just a symptom of a
larger problem of a low sanity waterline. Even if all religious content were
deleted tomorrow from all human minds, the larger and more general failures of
social rationality that permit religion would still be present. Getting rid of
the asphyxiated canary in the coalmine doesn’t get rid of the gas. How can we
teach rationality without explicit
mention of religion? What could you teach people that would raise their general epistemic waterline to the point
that religion went underwater? Perhaps we could start by teaching about
evidence, epistemology, updating beliefs, curiosity-stoppers and cached
thoughts, affective death spirals, conformity pressures, reductionism, and so
on. There are religious Nobel laureates (like Robert Aumann, who proved Aumann’s Agreement Theorem) who haven’t
studied these. This suggests that the current sanity waterline is ridiculously
low, even in the highest halls of science.
If it were possible to teach people reliably
how to become exceptional, then it would no longer be exceptional. It is a challenge
to teach things we do that we
can’t easily understand how we do them. Success is hard to duplicate due to
luck, genetic potential, and incommunicable insights or intuitions. Yet by
learning about a domain and about the mind, we might make new skills more
teachable by diminishing the role of luck on future occasions. We can teach how not
to lose. So Yudkowsky asked in his first blog post: why are there schools
of martial arts, but not rationality dojos?
We don’t have a “Martial Art
of Rationality” because rationalists haven’t gathered to systematize and test
their skills and training methods, and hence rationalists don’t seem happier or
more successful; but we must have a sense that more is possible! Most self-proclaimed “rationalists” don’t seem to
get huge amounts of personal mileage out of their craft, because the level of
expertise they strive to develop is not on par with the skills of a
professional mathematician, but more like that of a strong casual amateur. We
should see this as a problem, and develop more systematic training.
An essay by Gillian Russell titled “Epistemic
Viciousness in the Martial Arts” generalizes amazingly to possible and actual
problems with building a community around rationality. There are epistemic
vices in the martial arts, due to data poverty (arising from the difficulty of
testing the skills in the real world), deference to historical masters of the
sacred dojo, emotional investment in old teachings, and trust in teachers
(because the art cannot be learned from a book). These may be transferable to
rationality training – what can be done about it?
Many schools of psychotherapy have proliferated
without experimental evidence, perhaps because they had the right air of
authority. For example, the Rorschach test is still used despite evidence of
ineffectiveness. Branches of “schools” become prestigious through charisma,
good stories, and by inventing their own schools and having students – not by excelling at any visible
performance criterion. But you need testing and statistics to tell how well
your organized practice is doing! An example of good measurements being used is the field of hedonic psychology
(happiness studies).
Rationality groups need methods to verify their
ideas on three levels: the reputational
level (grounded in reality, like success on some real-world problem or
competition), the experimental level (randomized
testing with replicable measurements, like well-validated surveys that can be
run on each of a hundred students), and the organizational
level (avoiding people gaming the test, to preserve the integrity of
organizations and have low-noise measurements). We need these to make
rationality useful! The strength of solution invented at each level will
determine how far the craft of rationality can be taken in the real world.
Pluralistic ignorance, evaporative cooling, and
affective death spirals bind groups together on the Dark Side; but for
rationalists to win we have to
cooperate too, and our culture of disagreement (and dispassion) is a barrier.
The “nonconformist cluster” (consisting of atheists, libertarians, sci-fi fans,
technophiles, programmers and early adopters etc.) seems to be stunningly bad
at coordinating group projects. Our exclusively individualist traditions
sabotage our ability to cooperate. People are reluctant to speak agreement out loud, but this is
dangerously half-rational, because it doesn’t help us obtain more cohesive and
powerful communities. If you tolerate only
disagreement but not agreement, and are only willing to hear some honest
thoughts but not others, then you are not fully rational.
Aspiring rationalists tend to have a
lower-than-usual tolerance for flawed thinking. But in order to work together,
we need to be able to tolerate other people’s tolerance, because otherwise we’d
need to have exactly the same standards of tolerance, which is unlikely.
Punishing non-punishers can be dangerous, so you should tolerate people who are
more tolerant and patient than you are (including those who say nice things
about crackpots), and judge them only for their own mistakes. It’s not
realistic to expect others to dislike everyone you dislike before cooperating
with them.
You get more done by joining a common project, but
how much should you demand that an existing group adjust toward you before you
will adjust toward it? Nonconformists tend to demand way too high a price (in strategic
shifts) for joining; this could be due to our hunter-gatherer instincts
underestimating the inertia of larger and more specialized groups (since we are
tuned to groups of 40 with minimal administrative demands and equal
participation). Join groups more easily. Don’t withhold from a worthwhile group
unless it has an annoying issue you care about but cannot fix.
Currently, the Pope can effectively mobilize
the Catholics for simple and obvious charitable projects, like responding with
food and shelter to a tidal wave in Thailand. But could an average atheist do
more good, without the motivation that comes from the irrational fear of Hell? For
secular humanists to match the per-capita altruistic output of the Catholic
Church, we need to be physically together to motivate ourselves, encourage
caring strongly about something, practice cognitive-behavioral therapy (CBT)
and/or Zen meditation, and target more efficient causes. Until this is the
case, any increase in atheism at the expense of Catholicism will be somewhat of
a hollow victory.
A post-religion era wouldn’t need artificial
churches, but new idioms of community, because that will be the gap left
behind. Since churches aren’t explicitly optimized for the role of providing
community, can we do better if we desire community without church? Offices can
support communities but aren’t optimized for it; we need strong rationalist
task forces built around worthwhile causes. There’s a great deal of work to be
done in the world; rationalist communities could organize themselves around
good causes while explicitly optimizing for community.
Many causes benefit from the spread of
rationality, so don’t expect to capture all rationalists for your own project –
it is better to tell people “this is a cool thing” than “this is the best thing, the only thing with the highest-return in expected utilons per marginal
dollar”. Otherwise people might lose the willpower to help. Causes can mutually
benefit from mutual effort on creating rationalists, as long as they come to
terms with not individually capturing all
the rationalists they create and learn to shut up about disagreements between
them (except in specialized venues clearly marked for such purposes). We’re all
elements of the common project of human progress.
Our grouping instincts are optimized for
50-person hunter-gatherer bands where everyone knows everyone else… In this
light, it seems miraculous that modern-day large institutions survive at all. Most
institutions fail to exist in the first
place, and Science survives not on individual donations (as it’s not a good
emotional fit) but by fastening itself parasitically onto large organizations
like governments, corporations, and large foundations. Modern humanity manages
to put forth very little in the way of coordinated individual effort to serve
our collective individual interests.
Many people prefer to help a good cause by
donating a few unskilled volunteer hours. People don’t like spending money, but
in this world of professional specialization, comparative advantage, gains from
trade and economies of scale, money is the unit of caring: if you want to do
good effectively, pay a full-time specialist instead of volunteering yourself! (Or
directly donate hours of the same specialized capability that you’d ordinarily
trade for money.) These tools are the reason we’re not still in caves. The
reason we have money is to realize the tremendous gains possible from each of
us doing what we do best. This is how one gets things done in the grownup world
when anyone really cares. Frugality
is a virtue, but if you’re never willing to spend any money, you don’t care.
Wealthy philanthropists typically reach a
mediocre final result when they try to purchase warm fuzzy feelings, status
among friends, and actual utilitarian gains simultaneously. This is a mistake
because it’s inefficient. To motivate yourself, you may spend some money to
obtain status and warm fuzzies (e.g. by helping people in person and donating
to something sexy), but do this separately
from purchasing expected utilons! You cannot optimize all three at once. Get
your warm fuzzies by volunteering at a soup kitchen or holding open doors for
little old ladies, and buy nice clothes for status. But spend most of your money purchasing expected
utilons. Altruism requires you to shut up and multiply, through cold-blooded
calculation, without worrying about status or enjoyment.
Bystander
apathy is the phenomenon that
large groups of people are less likely than individuals to act in emergencies.
This might happen due to pluralistic
ignorance (i.e. everyone looks around to see that everyone else appears
calm) and diffusion of responsibility
(i.e. everyone hopes that someone else will be the first to step up). However,
the bystander effect can be countered by telling people about it, and it’s less
likely when people know each other. If you’re ever in need of help, point to
one single bystander and ask them for help.
In the ancestral environment, we didn’t form
task-forces with strangers, which is maybe why today we fail to coordinate
large groups – especially over the internet. This might also account for the
bystander effect. How can we better use the internet to help our causes? There
may be an opportunity here for a startup to deliberately try to avert bystander
apathy in online group coordination. Some ideas may include putting up names
and photos of the first people who helped, giving helpers a video thank-you
from the founders, using referrer link codes, and so on.
Rationality is the spirit of systematized
winning: it’s not about following a certain “reasonable” ritual of cognition
(and whining when you keep losing), but about being flexibly prudent and winning whatever the means. It doesn’t mean that
rationality will make you invincible, but it means that if someone who isn’t
behaving according to your idea of rationality is consistently outcompeting
you, then you should consider that you’re not the one being rational. Perfectly
rational agents can lose; they just can’t know
in advance that they’ll lose. However, what if religious people are
happier?
Only perfect probability theory and decision theory are optimal. An incremental step in the direction of
ideal rationality doesn’t guarantee incrementally
more winning, and sometimes may make you worse off. So if perfection is
unattainable, why dare to try for improvement over a flawed baseline? But
refusing to climb one step up forfeits not just the height of that step, but
the height of the staircase. For some tasks, an unimproved level of performance
isn’t enough. If you care about truth and have something to protect, then with
further steps, things can get even better than before. And once you have
already taken that step forward, you can’t just shut your eyes and deny it to
yourself. In Yudkowsky’s limited experience with specialized applications, huge
improvements are possible – it just takes a lot of progress to get there. Those
first steps can be painful, yet the long road leads out of the valley and
higher than before.
If a hypothetical country of Bayesian
rationalists were attacked by savage Evil Barbarians (who know nothing about
probability theory or decision theory, and believe in a heavenly afterlife if
they die in battle), the rationalists should aim to avoid losing the war,
coordinate efficiently (e.g. following orders), and be ready to sacrifice
themselves to defend the community they care about. They should not fall for
the concept of “rationality” which says that the rationalists inevitably lose
because they would all individually prefer to stay out of harm’s way and are
too civilized to fight. Using Yudkowsky’s kind of decision theory, rational
agents will cooperate on the true Prisoner’s Dilemma and coordinate on group
projects whenever the expected probabilistic outcome is better than it would be
without such coordination. And real wars
cannot be won by refined politeness; war is not fun, but losing a war is even
less fun.
Aspiring rationalists tend to vastly
overestimate their ability to optimize other people’s lives. If you discover
the one method out of twenty (e.g. of productivity advice) that actually works
for you, it doesn’t mean that your
confident advice is better than randomly selecting one out of the twenty blog
posts. Beware of trying to optimize other people’s lives (even if they are
friends), because sometimes different things work for different people due to
undiscovered deep laws, and if you don’t take no for an answer, you can scare
people (especially when you have power over them).
“Other-Optimizing” messes around with surface
tricks without understanding the deeper general laws. Practical advice is much
more powerful and useful when backed up by concrete experimental results, true
causal accounts (deep theories), and validly-interpreted math and epistemology.
Thus, one can explain what works in truly general terms. Stripping out the
theories and giving the mere advice alone wouldn’t have nearly the same impact
or even the same message. It seems to be a distinctive style of Less Wrong to translate experiments and
math into practical advice.
When experimental subjects are warned about a
bias, they sometimes overcorrect for
it. If you know you are biased but you’re not sure how much, you may keep
tweaking and overcorrect. There is danger in overcorrecting for overconfidence:
Rationalists should not be underconfident,
because then you pass up opportunities (on which you could have been
successful) and don’t try hard enough – so, test
your abilities to discover your current level, and ask yourself whether a way
of thinking (e.g. unresolved doubts) is making you stronger or weaker! You
should seriously try to win, but aim for challenges you might lose at if you
don’t stretch yourself; sticking to things you always win at is one way smart
people become stupid.
The probability theory and decision theory of
the shared Way is neither masculine nor feminine, but there may be individual
differences in the human practice of rationality, and we have to find our own
paths to the center of the labyrinth and then radio back. The path to
rationality cannot be the same for everyone, although there is still a common
thing we are all trying to find. What Eliezer Yudkowsky is describing is not
just the Way (the thing that lies at
the center of the labyrinth), but also his
Way
(the path he took from wherever he started out). Hence his focus on the arts
required for advanced cognitive reductionism, untangling confused questions, and
writing male characters like Jeffreyssai. Yet there is much left to be
developed, including fighting akrasia, coordinating groups, becoming a proper
experimental science, being happy, and better introductory literature.
Your mileage may vary using
Yudkowsky’s writings on rationality; still, knowing about fake explanations,
the conjunction fallacy, motivated skepticism, affective death spirals etc. may
give you a saving throw; a base to build on. Eliezer has focused on epistemic
rationality more than instrumental rationality or rationality teaching and
verification, so the Art is incomplete – yet discriminating good from bad
systems of thinking saves people from instantaneously going astray. There is a
beginning barrier to surpass before you can start
creating high-quality craft of rationality, and Eliezer hopes that his
writings will serve to surpass this initial barrier; the rest he leaves to you.
Go develop more Art from multiple sources and by confronting challenges beyond
your armchair. And then: “remembering whence you came, radio back to tell
others what you learned.”
›
Comments
Post a Comment