Transcript for Eliezer Yudkowsky: A Theory of Fun

Today on ChangeSurfer Radio, we're listening to

a speech by Eliezer Yudkowsky, that he

gave at the Transvision Conference on

June 28th at Yale University.

His speech was about his theory of fun.

Which I think is actually kind of a fun topic.

Eliezer is a computer scientist who does research

at the Singularity Institute for Artificial Intelligence.

After that you'll listen to

a short essay that I wrote recently

in response to the Canadian decision to

legalize gay marriage and a bioethicist

in Canada who has been arguing against

gay marriage. I argue for it and for

some additional reforms in marriage law.

So stay tuned to both of those.

Coming up next, Eliezer Yudkowsky.

By the way some of the bumps and noises

that you hear in the background of the talk

at the beginning, are film crews setting up

to tape his talk. They'll go away after a while.

I once heard a friend of mine,

by the name of Liono,

talking to a newcomer to transhumanism.

And this person had remarked

that he wasn't afraid of dying.

And Liono said: "Well, you hear a lot about

overcoming your fear of death,

but how many people have tried to

overcome their fear of living forever?"

As Transhumanists I expect that

most of us at some point have had to

deal with the rationalizations

that people develop

to suppress their fear of death.

The idea that "death gives meaning to life",

and other comforting lies.

Personally I do believe that the fear of death

should be counted as a personal flaw.

Choosing to live does not require that I be

afraid to die, it requires only that I value life.

But resolving that philosophical question

is not an urgent matter. We can

temporarily agree to disagree upon it;

the important thing is making sure we have

at least a billion years in which to argue about it.

Now, living for a billion years is a scary thought.

It's not as scary as the thought of living forever

(which may or may not be permitted

by physical law) but even a mere

thousand years is a scary thought if you

really think about what that would mean for you as a person.

Even if you keep all the neurons in the brain alive,

the human mind is not designed

to handle a thousand years of memories.

You cannot stay a child

for even so much as a thousand years.

You must in that time at least begin to grow up.

But if life is a good thing, should we get scared

at the prospect of a huge amount of life?

What distinguishes transhumanism as a philosophy

is that it wholeheartedly embraces the goal of success

instead of making excuses for failure.

There are many philosophical explanations for:

Why Life Must Suck in Order to be Meaningful.

Why all the pain and death and catastrophe

and the minor annoyances that drain your life-force

are, somehow, necessary.

All these standard excuses have a strained,

forced quality to them, like

a theory stretched to explain evidence

that just doesn't fit.

Now, transhumanism just says:

This sucks. Let's fix it. That's all.

It doesn't need

to be any more complicated than that.

Life is better than death, health is better than sickness,

happiness is better than pain, knowledge is better than ignorance,

and when you see something wrong, you don't rationalize it;

you fix it. It is all very straightforward,

so what's so special about transhumanism?

Actually there is nothing special about transhumanism.

Presently, most other philosophies have a

special additional clause that causes them

to react oddly to advanced technologies.

As if, for example, anything that involves genes is scary.

There is a possibly apocryphal story

about the person who was warned not to eat a tomato

because it has genes in it.

In transhumanism, this special yuck-reaction

is missing and such technologies are

just an ordinary part of the natural universe.

Since this is the case, if you can use

tissue screening to save four years old Zain Hashmi's life,

as was a recent case in Britain,

or if you can use golden wheat

to prevent Vitamin A deficiency,

why on earth wouldn't you? Why not?

This choice does not require

a special enthusiasm for technology,

only that the usual sense of future-shock

be missing.

Typically people encountering transhumanism

for the first time assume that

"Oh these people are a bit different.

What's different about them - they must love technology."

It's not about loving technology.

It's just that the usual yuck-reaction is missing

and the ordinary philosophy

of good means to good ends takes over.

Similarly ordinary transhumanist philosophies

will say that life extension is good up to age 80 or whatever,

but if you talk about living to the age of 200

so then there's a special reaction of surprise,

future shock, because the philosopher

hasn't considered that possibility before.

Well a transhumanist will just say,

"Sure, life extension to 80 is good,

life extension to 200 is better,

life extension to 1,000 is better yet"

and so on. It does not require a special drive

toward immortality to say this,

it's just self-consistency.

Wanting to live to be 1,000 is not a special

and unusual desire that needs explaining,

it's just ordinary common sense after

the future shock has been stripped out.

So transhumanism is just the philosophy

that says life is good, happiness is good,

knowledge and freedom and intelligence and beauty are good

and this does not change for arbitrarily large

amounts of life and beauty.

Transhumanism is not a philosophy with a special adoration of technology;

it's just the philosophy which says that

technology is a normal way of achieving our aspirations.

This does not change even when you're talking about

arbitrarily advanced technology.

Biotech, nanotech, whatever,

there's nothing exciting about it,

it's just an ordinary part of the natural universe.

The question then becomes: what beautiful

and worthwhile occupations can you find

to fill the next million years?

This question of course is the province of

fun theory. Fun theory is the branch of science,

or rather, wild speculation,

that we use to answer questions such as:

How much fun is there in the universe?

Will we ever run out of fun?

Are we having fun yet?

And could we be having more fun?

Is fun scalable? Does it require

an exponentially greater amount of computation

to support a linear increase in fun?

I'm not going to present a rigorous theory of fun.

This presentation is too short

to present a rigorous theory of fun.

The only give you 20 minutes, very unfair.

Also I don't have a rigorous theory of fun.

So instead, let's look at some things that are fun.

For example, why is being in love fun?

This question was recently answered

when scientists discovered that being in love

stimulates the same brain centers

that are stimulated by eating chocolate.

So this could be one possible answer

to the question of what is fun?.

We could say that fun involves a certain kind of brain chemistry.

I do not think this is a good answer

Or if you select this as an answer,

then you are leaving out something important,

whether or not you call it fun. Why?

Well, suppose we ran a little wire into

one of the brain's pleasure centers

and hooked the wire up to a button

and gave you that button; and you spent

the next million years pressing that button.

Just that, doing nothing else.

I consider this to be

a highly dystopian scenario, a sterile dead end.

It seems more like counterfeit fun

than the genuine article.

Having your pleasure centers artificially stimulated,

wireheading, as Larry Niven called it,

is not philosophically acceptable fun.

That's what we're looking for, not just fun,

however we end up defining that, but

philosophically acceptable fun.

To have good clean fun,

or even wicked dirty fun,

it seems like you need to be doing something and getting somewhere.

What about eating cookies? Why do we eat cookies?

Certainly not because we're hungry. We eat

cookies because 50,000 years ago, before

the invention of agriculture, sugar and fat

were the limiting resources.

So we evolved to prefer sugar and fat,

and we still prefer sugar and fat today,

in modern times, when people are dying

from too many calories instead of too few,

because evolution has not had time to catch up.

So our fun activities, thus those activities

that your ancestors performed,

the way that men watch football games

because it involves competing tribes throwing things at each other,

is personal fulfillment,

to be found in harmony with your ancestral environment.

It would explain why modern day jobs

such as accounting have the annoying property

of draining people's life-force.

Of course this will only remain true of humans

who retain their ancestral neurology.

If 21st century civilizations

stayed around for a couple of million years,

humans would eventually evolve to find

personal fulfillment in doing their income tax forms.

This also strikes me as dystopian.

Although it is less dystopian than eating

Pringles and watching football.

What about the Rubik's cube?

Why is solving the Rubik's cube fun?

It's not an ancestral activity,

it does not directly mess with your brain chemistry,

it doesn't deliver any kind of

sensual reward when you finish solving it.

Why do we find it fun?

It's not a competitive game even. It's just pure math.

It's group theory embodied in a physical cube.

And we, as humans can have funplaying with that,

it's something to be proud of.

But the Rubik's cube contains

only a limited amount of fun.

You use it up, and then it's gone.

At first solving the cube is fun,

and then it's boring. What changes?

Does the cube itself change?

Do they ship Rubik's cubes from the factory

containing only a limited amount of fun,

so you have to buy a new one every 3 months?

No, the cube is a

mathematical concept, and it is eternal.

The cube does not change.

You change. The person who finishes

solving the cube is not the same person who started it.

The person who picked up the cube for the first time

may have known nothing about this class of puzzle;

or have not yet discovered the idea of

transformations that move only a few cubes around

and leave the rest constant.

When you finish the Rubik's cube you have

not only learned something about the cube,

you've learned something about how to solve this kind of problem.

You even have learned something about learning.

Once you have learned what the cube has to teach you,

solving the cube becomes easy.

So easy that it isn't fun anymore.

In having fun with the cube, you outgrow it,

and outgrowing the cube is part of the fun.

What do you do when you've outgrown the cube?

Well you can find online a Java applet

that implements a fourth dimensional version

of the Rubik's cube, a Rubik's tesseract.

I do not understand this puzzle.

I am still trying to figure out how the pieces move.

If I had not already learned the principles

involved in the Rubik's cube, the Rubik's tesseract

would be completely incomprehensible.

So you solve one problem, outgrow it,

and that gives you the abilities

to move on to the next problem.

And when you solved every possible form

the Rubik's cube can take, generalized on a level

where the entire class of problems

becomes uninteresting, then what?

Then you move on to the next class of problems

at a higher level of complexity.

This is a key ingredient of fun.

It's not quite the only ingredient of fun,

but it's a key ingredient, and

it's the only one I have time to cover here:

novel complexity.

Complexity you have not yet encountered.

The smarter you are, the faster you generalize,

and the faster you become bored with any given problem.

Yes, the smarter you are, the faster you get bored;

that is how it works.

But when you get smarter you can perceive

new areas of the problem space that would've been

incomprehensible to you before.

Humans would get bored with chimpanzee fun extremely fast.

But the space of human fun is enormously larger

than the space of chimpanzee fun.

If you have 10 bits, you have 1,000 possibilities,

if you have 20 bits, you have a million possibilities,

if you have 30 bits you have a billion possibilities.

I don't think fun quite goes exactly like that,

I don't think it's quite that simple,

but I think the basic relation is the same

that the size of fun-space grows roughly exponentially

as the amount of intelligence.

That the more variables you have in your mind,

the larger the problem you solve, as the problem gets larger

the space of possible novel problems goes up exponentially.

Or not, it is only a conjecture.

At any rate,

the idea that we're going to get smart

and then instantly run out of fun is based

I think, first on the stereotype of rich people who get bored;

and secondly, on the idea that you're limited to human fun

and the smarter you are the faster you'll get bored with human fun.

This is true, the smarter you are

the faster you'll get bored with human fun,

but the very act of getting bored with human fun,

that this is no longer novel complexity

means you're ready to move on.

So there's an immense amount of fun

out there to be had.

Even for humans, the size of human fun-space is

so large that no one individual could possibly succeed

in experiencing all the possible kinds of human fun

(though I encourage you to try!).

One person cannot fulfill more than a tiny infinitesimal

fraction of the human-space of possibilities.

Why, because one person's lifespan is too short?

Of course not. I plan to still be alive

after the last star in the milky way is dead.

But even if you live forever,

you'll not be able to explore more than

a tiny fraction of human fun-space

before you outgrow it.

And that of course, is the scary part:

growing up.

Right now, we are each of us growing old.

Not growing up, growing old.

We are actually losing neurons

as we grow old.

Instead of adding new capacity.

We loose vitality, creativity, flexibility,

energy, even personal health as we age.

This is not a feature, this is a bug.

It is a very unnatural thing to loose neurons as you age.

We should be adding more neurons as we grow up.

To hold more and more experiences and

more and more complex skills.

To shrink as we age borders on the perverted.

This is not the way things are supposed to be.

And at some point, we shall have to do a little rearranging!

Once that bug is fixed however, once we are

truly growing up instead of growing old,

there will be only so much time you can remain human.

If everyday you learn something new,

there will come a time, after a sufficiently

large number of days, when you have learned so much

that you can pick up a Rubik's tesseract and

see it as a small trivial thing to be

solved in a few flicks, cast aside.

And if that prospect does not scare you,

you must have a very well integrated personality.

This leads us to the fear of growing up.

The fear of growing up is the third major cause

of the fear of living forever;

after the repressed fear of death,

and the fear of boredom.

In our time the fear of growing up

is actually a minor academic fad

except that it is not called the fear of growing up.

It is called the fear of posthumanity.

The funny thing is that transhumanists

were using the word "posthuman"

long before folks like Francis Fukuyama

tried to make it into a scare word.

We were not particularly scared

by the word "posthuman" even though

we were (way back in the good old days)

we were using the word "posthuman"

to indicate people who had grown up into Jupiter-brains;

that is people whose minds have become so large

they had to run themselves on computers the size of Jupiter.

This, roughly speaking,

is what the old-time transhumanists

meant by "posthuman".

Francis Fukuyama seems to use the term "posthuman"

to indicate someone with

a couple of minor genetic hacks,

which by our standards are such a tiny alteration

as to not be worth noticing.

And yet Fukuyama is scared of his posthumans,

one wonders what he would think of ours,

he would probably catch fire and die.

Not every change is an improvement,

but every improvement is,

necessarily, a change.

To move forward, you must move.

Are you ashamed to be a "post-child"

or a "post-chimpanzee"?

When Fukuyama reinvented the word "posthuman"

he imbued it with an intellectual sleight of hand:

the good old naturalistic fallacy,

that is implies ought.

Human nature is an evolved mix

of light and darkness.

Hitler was not an inhuman monster,

he was a human monster.

The library of Alexandria was built and burnt

by one and the same species.

Only by judging myself, by not being content

to rest where I am, can I move forward.

These rules are not suspended when I judge parts of myself

that are part of universal human nature.

Here again it is transhumanism that is the simpler case.

Transhumanism as a philosophy

contains no special exemption from moral judgment,

for evils that are embedded in human nature.

I am not going to stop debugging myself

when I run across bugs that are part of human nature.

I guess that makes me a posthumanist.

Not every change is an improvement, but

every improvement is necessarily a change.

Not every form of posthumanity is a nice place to live,

but all sufficiently nice places to live

are necessarily posthuman.

Let's make the future a nice place to live!

That was a talk by Eliezer Yudkowsky

the research scientist who works at

Singularity Institute for Artificial Intelligence.

He's also a board member of the World Transhumanist Association.

And he delivered that talk at the Transvision Conference

on Transhumanist Bioethics, which I organized

at Yale University.