If I had to pick a single statement that relies on more Overcoming Bias content I've written than any other, that statement would be:

Any Future not shaped by a goal system with detailed reliable inheritance from human morals and metamorals, will contain almost nothing of worth.

"Well," says the one, "maybe according to your provincial human values, you wouldn't like it.  But I can easily imagine a galactic civilization full of agents who are nothing like you, yet find great value and interest in their own goals.  And that's fine by me.  I'm not so bigoted as you are.  Let the Future go its own way, without trying to bind it forever to the laughably primitive prejudices of a pack of four-limbed Squishy Things -"

My friend, I have no problem with the thought of a galactic civilization vastly unlike our own... full of strange beings who look nothing like me even in their own imaginations... pursuing pleasures and experiences I can't begin to empathize with... trading in a marketplace of unimaginable goods... allying to pursue incomprehensible objectives... people whose life-stories I could never understand.

That's what the Future looks like if things go right.

If the chain of inheritance from human (meta)morals is broken, the Future does not look like this.  It does not end up magically, delightfully incomprehensible.

With very high probability, it ends up looking dull.  Pointless.  Something whose loss you wouldn't mourn.

Seeing this as obvious, is what requires that immense amount of background explanation.

And I'm not going to iterate through all the points and winding pathways of argument here, because that would take us back through 75% of my Overcoming Bias posts.  Except to remark on how many different things must be known to constrain the final answer.

Consider the incredibly important human value of "boredom" - our desire not to do "the same thing" over and over and over again.  You can imagine a mind that contained almost the whole specification of human value, almost all the morals and metamorals, but left out just this one thing -

- and so it spent until the end of time, and until the farthest reaches of its light cone, replaying a single highly optimized experience, over and over and over again.

Or imagine a mind that contained almost the whole specification of which sort of feelings humans most enjoy - but not the idea that those feelings had important external referents.  So that the mind just went around feeling like it had made an important discovery, feeling it had found the perfect lover, feeling it had helped a friend, but not actually doing any of those things - having become its own experience machine.  And if the mind pursued those feelings and their referents, it would be a good future and true; but because this one dimension of value was left out, the future became something dull.  Boring and repetitive, because although this mind felt that it was encountering experiences of incredible novelty, this feeling was in no wise true.

Or the converse problem - an agent that contains all the aspects of human value, except the valuation of subjective experience.  So that the result is a nonsentient optimizer that goes around making genuine discoveries, but the discoveries are not savored and enjoyed, because there is no one there to do so.  This, I admit, I don't quite know to be possible.  Consciousness does still confuse me to some extent.  But a universe with no one to bear witness to it, might as well not be.

Value isn't just complicated, it's fragile.  There is more than one dimension of human value, where if just that one thing is lost, the Future becomes null.  A single blow and all value shatters.  Not every single blow will shatter all value - but more than one possible "single blow" will do so.

And then there are the long defenses of this proposition, which relies on 75% of my Overcoming Bias posts, so that it would be more than one day's work to summarize all of it.  Maybe some other week.  There's so many branches I've seen that discussion tree go down.

After all - a mind shouldn't just go around having the same experience over and over and over again.  Surely no superintelligence would be so grossly mistaken about the correct action?

Why would any supermind want something so inherently worthless as the feeling of discovery without any real discoveries?  Even if that were its utility function, wouldn't it just notice that its utility function was wrong, and rewrite it?  It's got free will, right?

Surely, at least boredom has to be a universal value.  It evolved in humans because it's valuable, right?  So any mind that doesn't share our dislike of repetition, will fail to thrive in the universe and be eliminated...

If you are familiar with the difference between instrumental values and terminal values, and familiar with the stupidity of natural selection, and you understand how this stupidity manifests in the difference between executing adaptations versus maximizing fitness, and you know this turned instrumental subgoals of reproduction into decontextualized unconditional emotions...

...and you're familiar with how the tradeoff between exploration and exploitation works in Artificial Intelligence...

...then you might be able to see that the human form of boredom that demands a steady trickle of novelty for its own sake, isn't a grand universal, but just a particular algorithm that evolution coughed out into us.  And you might be able to see how the vast majority of possible expected utility maximizers, would only engage in just so much efficient exploration, and spend most of its time exploiting the best alternative found so far, over and over and over.

That's a lot of background knowledge, though.

And so on and so on and so on through 75% of my posts on Overcoming Bias, and many chains of fallacy and counter-explanation.  Some week I may try to write up the whole diagram.  But for now I'm going to assume that you've read the arguments, and just deliver the conclusion:

We can't relax our grip on the future - let go of the steering wheel - and still end up with anything of value.

And those who think we can -

- they're trying to be cosmopolitan.  I understand that.  I read those same science fiction books as a kid:  The provincial villains who enslave aliens for the crime of not looking just like humans.  The provincial villains who enslave helpless AIs in durance vile on the assumption that silicon can't be sentient.  And the cosmopolitan heroes who understand that minds don't have to be just like us to be embraced as valuable -

I read those books.  I once believed them.  But the beauty that jumps out of one box, is not jumping out of all boxes.  (This being the moral of the sequence on Lawful Creativity.)  If you leave behind all order, what is left is not the perfect answer, what is left is perfect noise.  Sometimes you have to abandon an old design rule to build a better mousetrap, but that's not the same as giving up all design rules and collecting wood shavings into a heap, with every pattern of wood as good as any other.  The old rule is always abandoned at the behest of some higher rule, some higher criterion of value that governs.

If you loose the grip of human morals and metamorals - the result is not mysterious and alien and beautiful by the standards of human value.  It is moral noise, a universe tiled with paperclips.  To change away from human morals in the direction of improvement rather than entropy, requires a criterion of improvement; and that criterion would be physically represented in our brains, and our brains alone.

Relax the grip of human value upon the universe, and it will end up seriously valueless.  Not, strange and alien and wonderful, shocking and terrifying and beautiful beyond all human imagination.  Just, tiled with paperclips.

It's only some humans, you see, who have this idea of embracing manifold varieties of mind - of wanting the Future to be something greater than the past - of being not bound to our past selves - of trying to change and move forward.

A paperclip maximizer just chooses whichever action leads to the greatest number of paperclips.

No free lunch.  You want a wonderful and mysterious universe?  That's your value.  You work to create that value.  Let that value exert its force through you who represents it, let it make decisions in you to shape the future.  And maybe you shall indeed obtain a wonderful and mysterious universe.

No free lunch.  Valuable things appear because a goal system that values them takes action to create them.  Paperclips don't materialize from nowhere for a paperclip maximizer.  And a wonderfully alien and mysterious Future will not materialize from nowhere for us humans, if our values that prefer it are physically obliterated - or even disturbed in the wrong dimension.  Then there is nothing left in the universe that works to make the universe valuable.

You do have values, even when you're trying to be "cosmopolitan", trying to display a properly virtuous appreciation of alien minds.  Your values are then faded further into the invisible background - they are less obviously human.  Your brain probably won't even generate an alternative so awful that it would wake you up, make you say "No!  Something went wrong!" even at your most cosmopolitan.  E.g. "a nonsentient optimizer absorbs all matter in its future light cone and tiles the universe with paperclips".  You'll just imagine strange alien worlds to appreciate.

Trying to be "cosmopolitan" - to be a citizen of the cosmos - just strips off a surface veneer of goals that seem obviously "human".

But if you wouldn't like the Future tiled over with paperclips, and you would prefer a civilization of...

...sentient beings...

...with enjoyable experiences...

...that aren't the same experience over and over again...

...and are bound to something besides just being a sequence of internal pleasurable feelings...

...learning, discovering, freely choosing...

...well, I've just been through the posts on Fun Theory that went into some of the hidden details on those short English words.

Values that you might praise as cosmopolitan or universal or fundamental or obvious common sense, are represented in your brain just as much as those values that you might dismiss as merely human.  Those values come of the long history of humanity, and the morally miraculous stupidity of evolution that created us.  (And once I finally came to that realization, I felt less ashamed of values that seemed 'provincial' - but that's another matter.)

These values do not emerge in all possible minds.  They will not appear from nowhere to rebuke and revoke the utility function of an expected paperclip maximizer.

Touch too hard in the wrong dimension, and the physical representation of those values will shatter - and not come back, for there will be nothing left to want to bring it back.

And the referent of those values - a worthwhile universe - would no longer have any physical reason to come into being.

Let go of the steering wheel, and the Future crashes.

New Comment
108 comments, sorted by Click to highlight new comments since:
Some comments are truncated due to high volume. (⌘F to expand all)Change truncation settings

"Except to remark on how many different things must be known to constrain the final answer."

What would you estimate the probability of each thing being correct is?

What is human morals and metamorals?

What about "near-human" morals, like, say, Kzinti: Where the best of all possible words contains hierarchies, duels to the death, and subsentient females; along with exploration, technology, and other human-like activities. Though I find their morality repugnant for humans, I can see that they have the moral "right" to it. Is human morality, then, in some deep sense better than those?

7Roko
No. The point of meta-ethics as outlined on LW is that there is no "deeper sense", no outside perspective from which to judge moral views against one another.
-2[anonymous]
What would "deeper sense" even mean? Human morality is better (or h-better, whichever terminology you prefer), that's all there is to it.

It is better in the sense that it is ours. It is an inescapable quality of life as an agent with values embedded in a much greater universe that might contain other agents with other values, that ultimately the only thing that makes one particular set of values matter more to that agent is that those are the values that belong to that agent.

We happen to have as one of our values, to respect others' values. But this particular value happens to be self-contradictory when taken to its natural conclusion. To take it to its conclusion would be to say that nothing matters in the end, not even what we ourselves care about. Consider the case of an alien being whose values include disrespecting others' values. Is the human value placed on respecting others' values in some deep sense better than this being's?

At some point you have to stop and say, "Sorry, my own values take precedence over yours when they are incompatible to this degree. I cannot respect this value of yours." And what gives you the justification to do this? Because it is your choice, your values. Ultimately, we must be chauvinists on some level if we are to have any values at all. Otherwise, what's wrong with a sociopath who murders for joy? How can we say that their values are wrong, except to say that their values contradict our own?

[-]Roko150

I think Eliezer is due for congratulation here. This series is nothing short of a mammoth intellectual achievement, integrating modern academic thought about ethics, evolutionary psychology and biases with the provocative questions of the transhumanist movement. I've learned a staggering amount from reading this OB series, especially about human values and my own biases and mental blank spots.

I hope we can all build on this. Really. There's a lot left to do, especially for transhumanists and those who hope for a significantly better future than the best available in today's world. For those who have more pedestrian ambitions for the future (i.e. most of the world), this series provides a stark warning as to how the well intentioned may destroy everything.

Bravo!

[crosspost from h+ goodness]

Pearson, it's not that kind of chaining. More like trying to explain to someone why their randomly chosen lottery ticket won't win (big space, small target, poor aim) when their brain manufactures argument after argument after different argument for why they'll soon be rich.

The core problem is simple. The targeting information disappears, so does the good outcome. Knowing enough to refute every fallacious remanufacturing of the value-information from nowhere, is the hard part.

What are the odds that every proof of God's existence is wrong, when there are so many proofs? Pretty high. A selective search for plausible-sounding excuses won't change reality itself. But knowing the specific refutations - being able to pinpoint the flaws in every supposed proof - that might take some study.

I have read and considered all of Eliezer's posts, and still disagree with him on this his grand conclusion. Eliezer, do you think the universe was terribly unlikely and therefore terribly lucky to have coughed up human-like values, rather than some other values? Or is it only in the stage after ours where such rare good values were unlikely to exist?

I imagine a distant future with just a smattering of paper clip maximizers -- having risen in different galaxies with slightly different notions of what a paperclip is -- might actually be quite interesting. But even so, so what? Screw the paperclips, even if they turn out to be more elegant and interesting than us!

Robin, I discussed this in The Gift We Give To Tomorrow as a "moral miracle" that of course isn't really a miracle at all. We're judging the winding path that evolution took to human value, and judging it as fortuitous using our human values. (See also, "Where Recursive Justification Hits Bottom", "The Ultimate Source", "Created Already In Motion", etcetera.)

[-]Roko21

RH: "I have read and considered all of Eliezer's posts, and still disagree with him on this his grand conclusion. Eliezer, do you think the universe was terribly unlikely and therefore terribly lucky to have coughed up human-like values, rather than some other values?"

  • yes, it almost certainly was because of the way we evolved. There are two distinct events here:

  • A species evolves to intelligence with the particular values we have.

  • Given that a species evolves to intelligence with some particular values, it decides that it likes those values.

... (read more)

Evolution (as an algorithm) doesn't work on the indestructible. Therefore all naturally-evolved beings must be fragile to some extent, and must have evolved to value protecting their fragility.

Yes, a designed life form can have paper clip values, but I don't think we'll encounter any naturally occurring beings like this. So our provincial little values may not be so provincial after all, but common on many planets.

Ian C.: [i]"Yes, a designed life form can have paper clip values, but I don't think we'll encounter any naturally occurring beings like this. So our provincial little values may not be so provincial after all, but common on many planets."[i] Almost all life forms (especially simpler ones) are sort of paperclip maximizers, they just make copies of themselves ad infinitum. If life could leave this planet and use materials more efficiently, it would consume everything. Good for us evolution couldn't optimize them to such an extent.

Ian: some individual values of other naturally-evolved beings may be recognizable, but that doesn't mean that the value system as a whole will.

I'd expect that carnivores, or herbivores, or non-social creatures, or hermaphrodites, or creatures with a different set of senses - would probably have some quite different values.

And there can be different brain architectures, different social/political organisation, different transwhateverism technology, etc.

Roko:

Not so fast. We like some of our evolved values at the expense of others. Ingroup-outgroup dynamics, the way we're most motivated only when we have someone to fear and hate: this too is an evolved value, and most of the people here would prefer to do away with it if we can.

The interesting part of moral progress is that the values etched into us by evolution don't really need to be consistent with each other, so as we become more reflective and our environment changes to force new situations upon us, we realize that they conflict with one another. The analysis of which values have been winning and which have been losing (in different times and places) is another fascinating one...

"Ingroup-outgroup dynamics, the way we're most motivated only when we have someone to fear and hate: this too is an evolved value, and most of the people here would prefer to do away with it if we can."

So you would want to eliminate your special care for family, friends, and lovers? Or are you really just saying that your degree of ingroup-outgroup concern is less than average and you wish everyone was as cosmopolitan as you? Or, because ingroup-concern is indexical, it results in different values for different ingroups, so you wish every shared ... (read more)

I suspect it gets worse. Eliezer seems to lean heavily on the psychological unity of humankind, but there's a lot of room for variance within that human dot. My morality is a human morality, but that doesn't mean I'd agree with a weighted sum across all possible extrapolated human moralities. So even if you preserve human morals and metamorals, you could still end up with a future we'd find horrifying (albeit better than a paperclip galaxy). It might be said that that's only a Weirdtopia, that's you're horrified at first, but then you see that it's actuall... (read more)

I'll be horrified for as long as I damn well please.

Well, okay, but the Weirdtopia thesis under consideration makes the empirical falsifiable prediction that "as long as you damn well please" isn't actually a very long time. Also, I call scope neglect: your puny human brain can model some aspects of your local environment, which is a tiny fraction of this Earth, but you're simply not competent to judge the entire future, which is much larger.

I would like to point out that you're probably replying to your past self. This gives me significant amusement.

This post seems almost totally wrong to me. For one thing, its central claim - that without human values the future would, with high probability be dull is not even properly defined.

To be a little clearer, one would need to say something like: if you consider a specified enumeration over the space of possibile utility functions, a random small sample from that space would be "dull" (it might help to say a bit more about what dullness means too, but that is a side issue for now).

That claim might well be true for typical "shortest-first"... (read more)

2akshatrathi
Making a DNA sequence will count as (an extremely low level activity) [http://lesswrong.com/lw/xr/in_praise_of_boredom/] which is necessary to support non-boring activities. It is a very simple argument that these are the very activity we stop thinking about and concentrate on novel activities.
6Roko
Yes, but there would be no persons. There would be no scientists, no joy of discovery, no feeling of curiosity. There would just be a "process" that, from the outside, would look like an avalanche of expanding machinery, and on the inside would have no subjective experience. It would contain a complex intelligence, but there would be no-one to marvel at the complex intelligence, not even itself, because there would be "no-one home" in all likelihood. For me, what proved decisive in coming to a low estimate of the value of such a system was the realization that the reason that I liked science, technology, etc, was because of my subjective experiences of finding out the answer. Interestingness is in the eye of the beholder, but this piece argues that the beholder would have no eye; that there would be an optimizing process that lacked the ability to experience joy over any of its discoveries.
1Ben
While I think you may very plausibly be correct, there is (I think) some reasonable grounds for uncertainty. I can imagine that an advanced algorithm that performs the role of making scientific discoveries to aid in the development of technologies for the great paperclip fleet might indeed have "some one home". It maybe that this is beneficial to its effectiveness, or might be close to essential. I can't make any strong claims about why this would be needed, only that human beings (at least me) do have "some one home", but if we didn't know about human beings are we were speculating on what organisms evolution might produce we might find ourselves postulating complex, social creatures who solve complicated tasks, but have "no one home", and we would obviously be wrong.
4benelliott
But you don't need very many, and you're free to enslave them while they work then kill them once they're done. They might not need to be conscious, and they certainly don't need to enjoy their work. Probably, they will just be minor subroutines of the original AI, deleted and replaced once they learn everything necessary, which won't take long for a smart AI.
3AgentME
We don't particularly value copying DNA sequences for its own sake either though. Imagine a future where an unthinking strain of bacteria functioned like grey goo and replicated itself using all matter in its light cone, and it was impervious to mutations. I wouldn't rate that future as any more valuable than a future where all life went extinct. The goals of evolution aren't necessarily our goals.

Carl:

I don't think that automatic fear, suspicion and hatred of outsiders is a necessary prerequisite to a special consideration for close friends, family, etc. Also, yes, outgroup hatred makes cooperation on large-scale Prisoner's Dilemmas even harder than it generally is for humans.

But finally, I want to point out that we are currently wired so that we can't get as motivated to face a huge problem if there's no villain to focus fear and hatred on. The "fighting" circuitry can spur us to superhuman efforts and successes, but it doesn't seem to... (read more)

@Eliezer: Can you expand on the "less ashamed of provincial values" part?

@Carl Shuman: I don't know about him, but for myself, HELL YES I DO. Family - they're just randomly selected by the birth lottery. Lovers - falling in love is some weird stuff that happens to you regardless of whether you want it, reaching into your brain to change your values: like, dude, ew - I want affection and tenderness and intimacy and most of the old interpersonal fun and much more new interaction, but romantic love can go right out of the window with me. Friends - I... (read more)

Patrick,

Those are instrumental reasons, and could be addressed in other ways. I was trying to point out that giving up big chunks of our personality for instrumental benefits can be a real trade-off.

http://lesswrong.com/lw/gz/policy_debates_should_not_appear_onesided/

Jordan: "I imagine a distant future with just a smattering of paper clip maximizers -- having risen in different galaxies with slightly different notions of what a paperclip is -- might actually be quite interesting."

That's exactly how I imagine the distant future. And I very much like to point to the cyclic cellular automaton (java applet) as a visualization. Actually, I speculate that we live in a small part of the space-time continuum not yet eaten by a paper clip maximizer. Now you may ask: Why don't we see huge blobs of paper clip maximizers... (read more)

Probability of an evolved alien species:

(A) Possessing analogues of pleasure and pain: HIGH. Reinforcement learning is simpler than consequentialism for natural selection to stumble across.

(B) Having a human idiom of boredom that desires a steady trickle of novelty: MEDIUM. This has to do with acclimation and adjustment as a widespread neural idiom, and the way that we try to abstract that as a moral value. It's fragile but not impossible.

(C) Having a sense of humor: LOW.

Probability of an expected paperclip maximizer having analogous properties, if it originated as a self-improving code soup (rather than by natural selection), or if it was programmed over a competence threshold by foolish humans and then exploded:

(A) MEDIUM

(B) LOW

(C) LOW

the vast majority of possible expected utility maximizers, would only engage in just so much efficient exploration, and spend most of its time exploiting the best alternative found so far, over and over and over.

I'm not convinced of that. First, "vast majority" needs to use an appropriate measure, one that is applicable to evolutionary results. If, when two equally probable mutations compete in the same environment, one of those mutations wins, making the other extinct, then the winner needs to be assigned the far greater weight. So, for example,... (read more)

it would seem easier to build (or mutate into) something that keeps going forever than it is to build something that goes for a while then stops.

On reflection, I realize this point might be applied to repetitive drudgery. But I was applying it to the behavior "engage in just so much efficient exploration." My point is that it may be easier to mutate into something that explores and explores and explores, than it would be to mutate into something that explores for a while then stops.

Thanks for the probability assessments. What is missing are supporting arguments. What you think is relatively clear - but why you think it is not.

...and what's the deal with mentioning a "sense of humour"? What has that to do with whether a civilization is complex and interesting? Whether our distant descendants value a sense of humour or not seems like an irrelevance to me. I am more concerned with whether they "make it" or not - factors affecting whether our descendants outlast the exploding sun - or whether the seed of human civilisation is obliterated forever.

@Jordan - agreed.

I think the big difference in expected complexity is between sampling the space of possible singletons' algorithms results and sampling the space of competitive entities. I agree with Eliezer that an imprecisely chosen value function, if relentlessly optimized, is likely to yield a dull universe. To my mind the key is that the ability to relentlessly optimize one function only exists if a singleton gets and keeps an overwhelming advantage over everything else. If this does not happen, we get competing entities with the computationally d... (read more)

What if I want a wonderful and non-mysterious universe? Your current argument seems to be that there's no such thing. I don't follow why this is so. "Fun" (defined as desire for novelty) may be the simplest way to build a strategy of exploration, but it's not obvious that it's the only one, is it?

A series on "theory of motivation" that explores other options besides novelty and fun as prime directors of optimization processes that can improve the universe (in their and maybe even our eyes).

"This talk about "'right' means right" still makes me damn uneasy. I don't have more to show for it than "still feels a little forced" - when I visualize a humane mind (say, a human) and a paperclipper (a sentient, moral one) looking at each other in horror and knowing there is no way they could agree about whether using atoms to feed babies or make paperclips, I feel wrong. I think about the paperclipper in exactly the same way it thinks about me! Sure, that's also what happens when I talk to a creationist, but we're trying to app... (read more)

@Jotaf, "Order > chaos, no?"

Imagine God shows up tomorrow. "Everyone, hey, yeah. So I've got this other creation and they're super moral. Man, moral freaks, let me tell you. Make Mennonites look Shintoist. And, sure, I like them better than you. It's why I'm never around, sorry. Thing is, their planet is about to get eaten by a supernova. So.. I'm giving them the moral green light to invade Earth. It's been real."

I'd be the first to sign up for the resistance. Who cares about moral superiority? Are we more moral than a paperclip maxi... (read more)

Carl:

Those are instrumental reasons, and could be addressed in other ways.

I wouldn't want to modify/delete hatred for instrumental reasons, but on behalf of the values that seem to clash almost constantly with hatred. Among those are the values I meta-value, including rationality and some wider level of altruism.

I was trying to point out that giving up big chunks of our personality for instrumental benefits can be a real trade-off.

I agree with that heuristic in general. I would be very cautious regarding the means of ending hatred-as-we-know-it in human ... (read more)

(E.g. repeating the mantra "Politics is the Mind-Killer" when tempted to characterize the other side as evil)

Uh, I don't mean that literally, though doing up a whole Litany of Politics might be fun.

[-]TGGP4-20

Maybe it's the types I of haunts I've been frequenting lately, but the elimination of all conscious life in the universe doesn't strike me as too terrible at the moment (provided it doesn't shorten my own lifespan).

We can sort the values evolution gave us into the following categories (not necessarily exhaustive). Note that only the first category of values is likely to be preserved without special effort, if Eliezer is right and our future is dominated by singleton FOOM scenarios. But many other values are likely to survive naturally in alternative futures.

  • likely values for all intelligent beings and optimization processes (power, resources)
  • likely values for creatures with roughly human-level brain power (boredom, knowledge)
  • likely values for all creatures under
... (read more)

Wei_Dai2, it looks like you missed Eliezer's main point:

Value isn't just complicated, it's fragile. There is more than one dimension of human value, where if just that one thing is lost, the Future becomes null. A single blow and all value shatters. Not every single blow will shatter all value - but more than one possible "single blow" will do so.

It doesn't matter that "many" values survive, if Eliezer's "value is fragile" thesis is correct, because we could lose the whole future if we lose just a single critical value. Do we have such critical values? Maybe, maybe not, but you didn't address that issue.

[-]ata210

I like the idea of replying to past selves and think it should be encouraged.

[-]Giles140

The added bonus is they can't answer back.

8DSimon
"Yeah, past me is terrible, but don't even get me started on future me, sheesh!"
1Luke_A_Somers
Quite. I never expected LW to resemble classic scenes from Homestuck... except, you know, way more functional.
  • likely values for all intelligent beings and optimization processes (power, resources)

Agree.

  • likely values for creatures with roughly human-level brain power (boredom, knowledge)

Disagree. Maybe we don't mean the same thing by boredom?

  • likely values for all creatures under evolutionary competition (reproduction, survival, family/clan/tribe)

Mostly agree. Depends somewhat on definition of evolution. Some evolved organisms pursue only 1 or 2 of these but all pursue at least one.

  • likely values for creatures under evolutionary competition who cannot
... (read more)
I agree with Eliezer that an imprecisely chosen value function, if relentlessly optimized, is likely to yield a dull universe.

So: you think a "paperclip maximiser" would be "dull"?

How is that remotely defensible? Do you think a "paperclip maximiser" will master molecular nanotechnology, artificial intelligence, space travel, fusion, the art of dismantling planets and stellar farming?

If so, how could that possibly be "dull"? If not, what reason do you have for thinking that those technologies would not help with t... (read more)

Maybe we don't mean the same thing by boredom?

I'm using Eliezer's definition: a desire not to do the same thing over and over again. For a creature with roughly human-level brain power, doing the same thing over and over again likely means it's stuck in a local optimum of some sort.

Genome equivalents which don't generate terminally valued individual identity in the minds they descrive should outperform those that do.

I don't understand this. Please elaborate.

Why not just direct expected utility? Pain and pleasure are easy to find but don't work nearly as we... (read more)

@Jotaf: No, you misunderstood - guess I got double-transparent-deluded. I'm saying this:

  • Probability is subjectively objective
  • Probability is about something external and real (called truth)
  • Therefore you can take a belief and call it "true" or "false" without comparing it to another belief
  • If you don't match truth well enough (if your beliefs are too wrong), you die
  • So if you're still alive, you're not too stupid - you were born with a smart prior, so justified in having it
  • So I'm happy with probability being subjectively objective, a

... (read more)
Another way of saying this is that human beings are not expected utility maximizers, not as individuals and certainly not as societies.

They are not perfect expected utility maximizers. However, no expected utility maximizer is perfect. Humans approach the ideal at least as well as other organisms. Fitness maximization is the central explanatory principle in biology - and the underlying idea is the same. The economic framework associated with utilitarianism is general, of broad applicability, and deserves considerable respect.

But there is no principled way to derive an utility function from something that is not an expected utility maximizer!

You can model any agent as in expected utility maximizer - with a few caveats about things such as uncomputability and infinitely complex functions.

You really can reverse-engineer their utility functions too - by considering them as Input-Transform-Output black boxes - and asking what expected utility maximizer would produce the observed transformation.

A utility function is like a program in a Turing-complete language. If the behaviour can be computed at all, it can be computed by a utility function.

A utility function is like a program in a Turing-complete language. If the behaviour can be computed at all, it can be computed by a utility function.

Tim, I've seen you state this before, but it's simply wrong. A utility function is not like a Turing-complete language. It imposes rather strong constraints on possible behavior.

Consider a program which when given the choices (A,B) outputs A. If you reset it and give it choices (B,C) it outputs B. If you reset it again and give it choices (C,A) it outputs C. The behavior of this program cannot be reproduced b... (read more)

Wei Dai: Consider a program which when given the choices (A,B) outputs A. If you reset it and give it choices (B,C) it outputs B. If you reset it again and give it choices (C,A) it outputs C. The behavior of this program cannot be reproduced by a utility function.

I don't know the proper rational-choice-theory terminology, but wouldn't modeling this program just be a matter of describing the "space" of choices correctly? That is, rather than making the space of choices {A, B, C}, make it the set containing

(1) = taking A when offered A and B, (2) ... (read more)

Consider a program which when given the choices (A,B) outputs A. If you reset it and give it choices (B,C) it outputs B. If you reset it again and give it choices (C,A) it outputs C. The behavior of this program cannot be reproduced by a utility function.

That is silly - the associated utility function is the one you have just explicitly given. To rephrase:

if (senses contain (A,B)) selecting A has high utility; else if (senses contain (B,C)) selecting B has high utility; else if (senses contain (C,A)) selecting C has high utility;

Here's another example:
... (read more)
0tut
No it isn't. It is a list of preferences. The corresponding utility function would be a function U(X) from {A,B,C} to the real numbers such that 1) U(A)>U(B) 2) U(B)>U(C) and 3) U(C)>U(A) But only some lists of preferences can be described by utility functions, and this one can't, because 1) and 2) imply that U(A)>U(C), which contradicts 3).
0[anonymous]
Err, that got ugly. How do you make beutiful quotes on this site?
0arundelo
There's a help link under the box you type in. (Use > for quotes, as in email.) See also the Markdown documentation.
0tut
Thank you.
-1timtyler
I doubt the premise. Where are you getting that from? It wasn't in the specification of the problem.
0tut
From the definition of utility function.
-2timtyler
That seems like a ridiculous reply - it says nothing about the issue there.
2Sniffnoy
Tim, that's what the term means. This other thing that you have called a "utility function", is not in fact a utility function, because that's not what the term means. It's already been pointed out that not every list of preferences can be derived from a utility function. If you want to define or use a generalization of the notion of utility function, you should do so explicitly.
0timtyler
I have no argument with the definition of the term "utility function". It is a function that maps outcomes to utilities - usually real numbers. The function I described did just that. If you don't understand that, then you should explain what aspects of the function's map from outcomes to utilities you don't understand - since it seemed to be a pretty simple one to me. I don't think that all preferences can be expressed as a utility function. For example, some preferences are uncomputable. Note that Tyrrell_McAllister2's reply makes exactly the same point as I am making.
1Sniffnoy
See, this would have been a lot clearer if you had specified initially that your objection was to the domain.
2timtyler
Sorry if there was any confusion. Here are all the possible outcomes - and their associated (real valued) utilities - laboriously spelled out in a table: Remembers being presented with (A,B) and chooses A - utility 1.0. Remembers being presented with (A,B) and chooses B - utility 0.0. Remembers being presented with (B,C) and chooses B - utility 1.0. Remembers being presented with (B,C) and chooses C - utility 0.0. Remembers being presented with (C,A) and chooses C - utility 1.0. Remembers being presented with (C,A) and chooses A - utility 0.0. Other action - utility 0.0.
The core problem is simple. The targeting information disappears, so does the good outcome. Knowing enough to refute every fallacious remanufacturing of the value-information from nowhere, is the hard part.

The utility function of Deep Blue has 8,000 parts - and contained a lot of information. Throw all that information away, and all you really need to reconstruct Deep Blue is the knowledge that it's aim is to win games of chess. The exact details of the information in the original utility function are not recovered - but the eventual functional outcome... (read more)

The "targeting information" is actually a bunch of implementation details that can be effectively recreated from the goal - if that should prove to be necessary.

It is not precious information that must be preserved. If anything, attempts to preserve the 8,000 parts of Deep Blue's utility function while improving it would actually have a crippling negative effect on its future development. Similarly with human values: those are a bunch of implementation details - not the real target.

If Deep Blue had emotions and desires that were attached to the 8,000 parts of its utility function, if it drew great satisfaction, meaning, and joy from executing those 8,000 parts regardless of whether doing so resulted in winning a chess game, then yes, those 8,000 parts would be precious information that needed to be preserved. It would be a horrible disaster if they were lost. They wouldn't be the programmer's real target, but why in the world would Emotional Deep Blue care about what it's programmer wanted? It wouldn't want to win at chess, it would want to implement those 8,000 parts! That's what its real target is!

For humans, our real target is all those complex values that evolut... (read more)

0timtyler
The problem with self-improving Deep Blue preserving its 8,000 heuristics is that it might cause it to lose games of chess, to a player with a better representation of its target. If that happens, its 8,000 heuristics will probably turn out to assign very low values to the resulting lost games. Of course, that means that the values weren't very effectively maximized in the first place. Just so - that's one of the problems with working from a dud set of heuristics that poorly encode your target. We potentially face a similar issue. Plenty of folks would love to live in a world where their every desire is satisfied - and they live in continual ecstasy. However, pursuing such goals in the short-term could easily lead humanity towards long-term extinction. We face much the same problem with our values that self-improving Deep Blue faces with its heuristics. This issue doesn't have anything particularly to do with the difference between psychological and genetic optimization targets. Both genes and minds value dying out very negatively. They agree on the relevant values. There's a proposed solution to this problem: pursue universal instrumental values until you have conquered the universe, and then switch to pursuing your "real" values. However it's a controversial proposal. When will you be confident of not facing a stronger opponent with different values? How much does lugging those "true values" around for billions of years actually cost? My position is that you'll probably never know that you are safe, and that the cost isn't that great - but that any such expense is an intolerable squandering of resources.
0Ghatanathoah
Minds value not dying out because dying out would mean that they can no longer pursue "true values," not because not dying out is an end in itself. Imagine we were given a choice between: A) The human race dies out. B) The human race survives forever, but every human being alive and who will ever live will be tortured 24/7 by a sadistic AI. Any sane person would choose A. That's because in scenario B the human race, even though it survives, is unable to pursue any of its values, and is forced to pursue one of its major disvalues. There is no point in the human race surviving if it can't pursue its values. I personally think the solution for the species is the same as it is for an individual, mix pursuit of terminal and instrumental values. I do this every day and I assume you do as well. I spend lots of time and effort making sure that I will survive and exist in the future. But I also take minor risks, such as driving a car, in order to lead a more fun and interesting life. Carl's proposal sounds pretty good to me. Yes, it has dangers, as you correctly pointed out. But some level of danger has to be accepted in order to live a worthwhile life.
4timtyler
It's likely to not be a binary decision. We may well be able to trade preserving values against a better chance of surviving at all. The more we deviate from universal instrumental values, the greater our chances of being wiped out by accidents or aliens. The more we adhere to universal instrumental values, the more of our own values get lost. Since I see our values heavily overlapping with universal instrumental values, adopting them doesn't seem too bad to me - while all our descendants being wiped out seems pretty negative - although also rather unlikely. How to deal with this tradeoff is a controversial issue. However, it certainly isn't obvious that we should struggle to preserve our human values - and resist adopting universal instrumental values. That runs a fairly clear risk of screwing up the future for all our descendants.
0Ghatanathoah
If that's the case I don't think we disagree about anything substantial. We probably just disagree about what percentage of resources should go to UIV and what should go to terminal values. You might be right to some extent. Human beings tend to place great terminal value on big, impressive achievements, and quickly colonizing the universe would certainly involve doing that.
0timtyler
It's a tricky and controversial issue. The cost of preserving our values looks fairly small - but any such expense diverts resources away from the task of surviving - and increases the risk of eternal oblivion. Those who are wedded to the idea of preserving their values will need to do some careful accounting on this issue, if they want the world to run such risks. While the phrase "universal instrumental values" has the word "instrumental" in it, that's just one way of thinking about them. You could also call them "nature's values" or "god's values". You can contrast them with human values - but it isn't really an "instrumental vs terminal" issue.

Tim and Tyrrell, do you know the axiomatic derivation of expected utility theory? If you haven't read http://cepa.newschool.edu/het/essays/uncert/vnmaxioms.htm or something equivalent, please read it first.

Yes, if you change the spaces of states and choices, maybe you can encode every possible agent as an utility function, not just those satisfying certain axioms of "rationality" (which I put in quotes because I don't necessarily agree with them), but that would be to miss the entire point of expected utility theory, which is that it is supposed ... (read more)

0[anonymous]
Utility theory is bigger than the VN axioms. They are just one way of looking at things.

Wei: Most people in most situations would reject the idea that the set of options presented is part of the outcome - would say that (A,B,C) is a better outcome space than the richer one Tyrrell suggested - so expected utility theory is applicable. A set of preferences can never be instrumentally irrational, but it can be unreasonable as judged by another part of your morality.

Specifically, the point of utility theory is the attempt to predict the actions of complex agents by dividing them into two layers:

  1. Simple list of values
  2. Complex machinery for attaining those values

The idea being that if you can't know the details of the machinery, successful prediction might be possible by plugging the values into your own equivalent machinery.

Does this work in real life? In practice it works well for simple agents, or complex agents in simple/narrow contexts. It works well for Deep Blue, or for Kasparov on the chessboard. It doesn't w... (read more)

3timtyler
Counter-example 1: gene-frequency maximization in biology. A tremendously simple principle with enormous explanatory power. Counter-example 2: Entropy maximization. Another tremendously simple principle with enormous explanatory power. Note that both are maximization principles - the very type of principle whose limitations you are arguing for.

To expand on my categorization of values a bit more, it seems clear to me that at least some human value do not deserved to be forever etched into the utility function of a singleton. Those caused by idiosyncratic environmental characteristics like taste for salt and sugar, for example. To me, these are simply accidents of history, and I wouldn't hesitate (too much) to modify them away in myself, perhaps to be replaced by more interesting and exotic tastes.

What about reproduction? It's a value that my genes programmed into me for their own purposes, so why... (read more)

In dealing with your example, I didn't "change the space of states or choices". All I did was specify a utility function. The input states and output states were exactly as you specified them to be. The agent could see what choices were available, and then it picked one of them - according to the maximum value of the utility function I specified.

The corresponding real world example is an agent that prefers Boston to Atlanta, Chicago to Boston, and Atlanta to Chicago. I simply showed how a utility maximiser could represent such preferences. Su... (read more)

The analogy between the theory that humans behave like expected utility maximisers - and the theory that atoms behave like billiard balls could be criticised - but it generally seems quite appropriate to me.

This is a critical post. I disagree with where Eliezer has gone from here; but I'm with him up to and including this point. This post is a good starting point for a dialogue.

I don't know, or maybe I don't understand your point. I would find a quiet and silent, post-human world very beatiful in a way. A world where the only reminders of the great, yet long gone civilisation would be ancient ruins.. Super structures which once were the statues of human prosperity and glory, now standing along with nothing but trees and plants, forever forgotten. Simply sleeping in a never ending serenity and silence...

Don't you too, find such a future very beatiful in an eerie way? Even if there is no sentient being to perceive it at that time, the fact that such a future may exist one day, and that it can now be perceived through art and imagination, is where it's beauty truly lies.

2JoshuaZ
I suspect that you are imagining this world a good because you can't actually separate your imagined observer from the world. The world you are talking about is not just a failure of humanity it is a world where we have failed so much that nothing is alive to witness our failure.
0Randolf
I don't think you can call such a world good or perfect, but I don't think it's all bad either. I quess you could call it neutral. I mean, I don't see that world as a big failure, if a failure at all. No civilization will be there forever*, but the one I mentioned had at least achieved something at it's time: it had once been glorious. While it left it's statues, it still managed to keep the world habitable for life and other species. (note how I mentioned trees and plants growing on the ruins). To put it simple, it was a beatiful civilization that left a beatiful world.. It isn't fair to call it a failure only because it wasn't eternal. *Who am I to say that?
6APMason
I'll only speak for myself, but 'everybody dead' gives an output nowhere near zero on my utility function. Everybody dead is awful. It's not the worst imaginable outcome, but it is really really really low in my preference ordering. I can see why you would think it's neutral - there's nobody to be happy but there's nobody to suffer either. However, if you think that people dying is a bad thing in itself, this outcome really is horrifying.

Value isn't fragile because value isn't a process. Only processes can be fragile or robust.

Winning the lottery is fragile, is a fragile process, because it had to be done all in one go. Contrast that with the process of writing down a 12 digit phone number: if you to try to memoriese the whole number, and then write it down, you are likely to make a mistake, due to Millers law. Writing digits down one at time, as you hear them, is more robust. Being able to ask for corrections, or having errors pointed out to you, is more robust still.

Processes that ar... (read more)

This is the key point on which I disagree with Eliezer. I don't disagree with what he literally says here, but with what he implies and what he concludes. The key context he isn't giving here is that what he says here only applies fully to a hard-takeoff AI scenario. Consider what he says about boredom:

Surely, at least boredom has to be a universal value. It evolved in humans because it's valuable, right? So any mind that doesn't share our dislike of repetition, will fail to thrive in the universe and be eliminated...

If you are familiar with the differ

... (read more)

I think some of the assumptions here have lead you to false conclusions. For one, you seem to assume that because humans share some values, all humans have an identical value system. This is just plain wrong, humans each have their own unique value "signature" more or less like a fingerprint. If there is one thing that you place more value weight on than a person who is otherwise identical, you are different. That being said, does your argument still hold with this, albeit minor in the grand scheme of things, heterogeneity added to human valu... (read more)

1hosford42
Human values are special because we are human. Each of us is at the center of the universe, from our own perspective, regardless of what the rest of the universe thinks of that. It's the only way for anything to have value at all, because there is no other way to choose one set of values over another except that you happen to embody those values. The paperclip maximizer's goals do not have value with respect to our own, and it is only our own that matter to us. A paperclip maximizer could have its values adjusted to want to make staples instead. But what would the paperclip maximizer think of this? Clearly, this would be contrary to its current goal of making paperclips. As a consequence, the paperclip maximizer will not want to permit such a change, since what it would become would be meaningless with respect to its current values. The same principle applies to human beings. I do not want my values to be modified because who I would become would be devalued with respect to my current values. Even if the new me found the universe every bit as rich and meaningful as the old me did, it would be no comfort to me now because the new me's values would not coincide my current values.

Regarding this post and the complexity of value:

Taking a paperclip maximizer as a starting point, the machine can be divided up into two primary components: the value function, which dictates that more paperclips is a good thing, and the optimizer that increases the universe's score with respect to that value function. What we should aim for, in my opinion, is to become the value function to a really badass optimizer. If we build a machine that asks us how happy we are, and then does everything in its power to improve that rating (so long as it doesn't inv... (read more)

I'm not sure if 'fragile' is the right word, removing one component might be devastating, but in my opinion, that more reflects on the importance of each piece, and not so much on the fragility of the actual system. The way I see it, it's something like a tower with 4 large beams for support, if one takes out a single piece, it would be worse than, say if one removed a piece from a tower with 25 smaller beams to support it.

But other than that, thank you very much for the informative article.

Any Future not shaped by a goal system with detailed reliable inheritance from human morals and metamorals, will contain almost nothing of worth.

Did anyone notice that this flatly contradicts Three Worlds Collide? The superhappies and babyeaters don't inherit from human morals at all (let alone detailedly and reliably), but the humans still regard the aliens as moral patients, having meddling preferences for the babyeater children to not be eaten, rather than being as indifferent as they would be to heaps of pebbles being scattered.

(Yes, it was fiction,... (read more)

3Said Achmiz
No, I think this post is right as-is. As you say, Three Worlds Collide was fiction. There is no “but”. It’s fictional evidence, and so it should update us not at all.
3Zack_M_Davis
Sorry, the function of bringing up Three Worlds Collide was to point out the apparent contradiction in the Yudkowskian canon. Forget the story; I agree that fiction didn't happen and therefore isn't evidence. The actual issue is that it seems like worlds shaped by the goal systems of other evolved biological creatures probably don't "contain almost nothing of worth": the lives of octopuses mean much less to me than human lives, but more than tiny molecular paperclips. The theme of "animal-like organisms that feel pleasure and pain" is something that natural selection will tend to reinvent, and the idealized values of those organisms are not a random utility function. (Do you disagree? If so, you at least face a Sorites problem on how fast value drops off as you look at our evolutionary history. Do chimpanzees matter? If not, did Homo erectus?) But if other animals aren't literally-as-valueless-as-paperclips, then some classes of AI architecture might not be, either.
4gjm
Having disagreed with Zack many times in the past, it is a pleasure to say: I think this is absolutely right (except that I think I'd replace "pleasure and pain" with "something pleasure-like and something pain-like"); that bit of "Value is Fragile" is surely wrong, and the intuitions that drove the relevant bits of "Three Worlds Collide" are more reflective of how actual human value systems work. I think I'd want to distinguish two related but separate issues here. (1) Should we expect that (some) other intelligent agents are things whose welfare we value? (Whether they are might depend on whether we think they have internal mechanisms that resemble our mechanisms of pleasure, pain, hope, fear, etc.) (2) Should we expect that (some) other intelligent agents share some of our values? (Whether they do would depend on how far the structure of their thinking has converged with ours.) If there are other intelligent species out there, then whether they're "animal-like organisms that feel pleasure and pain" addresses #1 and whether "the idealized values of those organisms are not a random utility function" addresses #2. (Of course, how much we care about their welfare may depend on how much we think they share our values, for internalized-game-theory-ish reasons. And presumably they're likely to share more of our values if their motivational systems work similarly to ours. So the issues are not only related but interdependent.)
1Said Achmiz
Suppose that (evolved/uplifted/otherwise-advanced-enough-for-sapience) octopuses share some of our values. Now suppose that humans go extinct, and these Octopus sapiens create an advanced civilization, whose products instantiate some values we would recognize, like art, music, science, etc. Does this future contain anything of value? I say it does not, because there are no humans around to value it. There are octopuses, and that’s great for the octopuses, but as far as human values go, this future ended with humanity’s extinction. Whatever happens afterwards is irrelevant. EDIT: Mind you—this is not quite the point Eliezer was making, I don’t think; I am responding to gjm’s comment, here. This comment should not necessarily be taken to constitute part of a defense of the point made in the OP (and quoted by Zack upthread).
7gjm
When I consider this possible universe, I find that I do attach some value to the welfare of these sapient octopuses, and I do consider that it's a universe that contains plenty of value. (It depends somewhat on whether they have, as well as values resembling ours, something I can recognize as welfare; see my last couple of paragraphs above.) If there were a magic switch I could control, where one setting is "humans go extinct, no other advanced civilization ever exists" and the other is "humans go extinct, the sapient octopus civilization arises", I would definitely put it on the second setting, and if sufficiently convinced that the switch would really do what it says then I think I would pay a nonzero amount, or put up with nonzero effort or inconvenience, to put it there. Of course my values are mine and your values are yours, and if we disagree there may be no way for either of us to persuade the other. But I'll at least try to explain why I feel the way I do. (So far as I can; introspection is difficulty and unreliable.) First, consider two possible futures. 1: Humanity continues for millions of years, substantially unchanged from how we are now. (I take it we agree that in this case the future universe contains much of value.) 2: Humanity continues for millions of years, gradually evolving (in the Darwinian sense or otherwise) but always somewhat resembling us, and always retaining something like our values. It seems to me that here, too, the future universe contains much of value. The sapient octopuses, I am taking it, do somewhat resemble us and have something like our values. Perhaps as much so as our descendants in possible future 2. So why should I care much less about them? I can see only one plausible reason: because our descendants are, in fact, our descendants: they are biologically related to us. How plausible is that reason? Possible future 3: at some point in that future history of humanity, our descendants decide to upload themselves into com
4Said Achmiz
One can construct all sorts of hypothetical scenarios, but I am far from convinced of their usefulness in teasing out our “true” values (as contrasted with “confabulating some plausible-sounding, but not reflectively stable, set of values”). That said, it seems to me that how much I value (and should value) any given future depends on the degree of that future’s resemblance to my current values. So, to take the examples: Indeed, we agree. Well, it depends: it seems to me that the further from my current values this future humanity drifts, the less I value this future. Crucially, it seems to me that the degree of difference (at any given future time period) will depend (and how c