Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

Ghatanathoah comments on In Praise of Boredom - Less Wrong

22 Post author: Eliezer_Yudkowsky 18 January 2009 09:03AM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (100)

Sort By: Old

You are viewing a single comment's thread. Show more comments above.

Comment author: Ghatanathoah 15 June 2012 12:37:04AM *  6 points [-]

Brains are hedonmic maximisers. They're only about 2% of human body mass, though. There are plenty of other optimisation processes to consider as well - machines, corporation, stock markets also maximise. The picture of civilization as a bunch of human brains is deeply mistaken.

All those things are controlled by brains. They execute the brains' commands, which is optimizing the world for fun. They are extensions of the human brains. Now, they might increase entropy or something as a side effect, but everything they do they do because a brain commanded it.

Hedonism is a means to an end. Pleasure is there for a reason.

Life doesn't give us reason and purpose. We give life reason and purpose. Speculating on what sort of metaphorical "purposes" life and nature might have might be a fun intellectual exercise, but ultimately it's just a game. Our purposes come from the desires of our brains, not from some mindless abstract trend. Your tendency to think otherwise is the major intellectual error that keeps you from grokking Eliezer's arguments.

The reason is that it helps organisms reproduce, and organisms reproduce - ultimately - because that's the best way to maximise entropy - according to the deep principle of the MEP.

Here's a question for you: Suppose some super-advanced aliens show up that offer to detonate a star for you. That will generate huge amounts of entropy, far more than you ever could by yourself. All you have to do in return is torture some children to death for the aliens' amusement. They'll make sure the police and your friends never find out you did it.

Would you torture those children? No, of course you wouldn't. Because you care about being moral and doing good and don't give a crap about entropy. You just think you do because you have a tendency to confuse real human goals with metaphorical, fake "goals" that abstract natural trends have.

Think you can more accurately characterise nature's maximand?

Why would I need to do that? My main point is that human civilization doesn't and shouldn't give a crap about nature's worthless maximand. When you post comments on Less Wrong a lot of time you seem to act like maximizing IGF and entropy are good things that organisms ought to do. You get upset at Eliezer for suggesting we should do something better with our lives. This is because you're deeply mistaken about the nature of goodness, progress, and values.

But just for fun, I'll take up your challenge. Nature doesn't have a maximand. It isn't sentient. And even if Nature was sentient and did have a maximand, the proper response for the human race would be to ignore Nature and obey their own desires instead of its stupid, evil commands.

That being said, even you instead asked me to answer the more reasonable question "What trends in evolution sort of vaguely resemble the maximand of an intelligent creature?" I still wouldn't say entropy maximization. The idea that evolution tends to do that is an illusion created by the Second Law of Thermodynamics. Because of the way that 2LTD works, doing anything for any reason tends to increase entropy. So obviously if an evolved organism does anything at all, it will end up increasing entropy. This creates an illusion that organisms are trying to maximize entropy. Carl Shulman is right, calling entropy nature's maximand is absurd, you might as well say "being attracted by gravity" or "being made of matter" are what nature commands.

A better (metaphorical) maximand might actually be local entropy minimization. It's obviously impossible to minimize total entropy, but life has a tendency to decrease the entropy in its local area. Life tends to use energy to remove entropy from its local area by building complex cellular structures. It's sort of an entropy pump, if you will. So if we metaphorically pretended it evolution had a purpose, it would actually be the reverse of what you claim.

But again, that's not my main point. My main point is that while you have a lot of good sources for your biology references, you don't have nearly as good a grasp of basic psychology and philosophy. This causes you to make huge errors when discussing what good, positive ways for life to develop in the future are.

Comment author: timtyler 15 June 2012 12:43:05AM -1 points [-]

Brains are hedonmic maximisers. They're only about 2% of human body mass, though. There are plenty of other optimisation processes to consider as well - machines, corporation, stock markets also maximise. The picture of civilization as a bunch of human brains is deeply mistaken.

All those things are controlled by brains. They execute the brains' commands, which is optimizing the world for fun. They are extensions of the human brains. Now, they might increase entropy or something as a side effect, but everything they do they do because a brain commanded it.

Nope. For instance, look at Kevin Kelly's book "Out of control". Or look into memetics. Human brains are an important force, but there are other maximisation processes going on, in culture, with genes and inside machines.

Comment author: Ghatanathoah 19 June 2012 12:37:57AM -1 points [-]

"Out of Control" appears to be primarily about decentralized decision making processes like democracy and capitalism. I never said that brains controlled the artifacts of civilization in a centralized fashion, I just said that they control them. Obviously human beings use all sorts of decentralized methods to help coordinate with each other.

That being said, while systems are not controlled in a centralized manner, they are restricted in a centralized manner. For instance, capitalism only works properly if people are prevented from killing and stealing. Even if there is no need to centrally control everything to get positive results, there is a need to centrally control some things.

There seems to be a later section in "Out of Control" where Kelly suggests giving up control to our machines is good in the same way that dictators giving up central control to democracy and capitalism is good. This seems short-sighted, especially in light of things like Bostrom's orthogonality thesis. The reasons democracy and capitalism do so much good is that:

  1. Human minds are an important component of them, and (most) humans care about morality, so the systems tend to be steered towards morally good results.
  2. There are some centralized restrictions on what these decentralized systems are able to do.

Unless you are somehow able to program the machines with moral values (i.e. make an FAI), turning control over to them seems like a bad idea. Creating moral machines isn't impossible, but the main point of Eliezer's writing is that it is much, much harder than it seems. I think he's quite correct.

As for memetics, the idea impressed me when I first came across it, but there doesn't seem to have been much development in the field since then. I am no longer impressed. In any case, the main reason memes "propagate" is that they help a brain fulfill its desires in some way, so really ever-evolving memes are just another one of the human mind's tools in its continuing quest for universal domination.

Comment author: timtyler 19 June 2012 01:03:43AM *  1 point [-]

As for memetics, the idea impressed me when I first came across it, but there doesn't seem to have been much development in the field since then. I am no longer impressed. In any case, the main reason memes "propagate" is that they help a brain fulfill its desires in some way, so really ever-evolving memes are just another one of the human mind's tools in its continuing quest for universal domination.

From a biological perspective, brains are seen as being a way for genes and memes to make more copies of themselves.

That this is a valuable point of view is illustrated by some sea squirts - which digest their own brains to further their own reproductive ends.

In nature, genes are fundamental, while brains are optional and expendable.

Comment author: Ghatanathoah 19 June 2012 03:46:34PM -1 points [-]

From a biological perspective, brains are seen as being a way for genes and memes to make more copies of themselves....

...In nature, genes are fundamental, while brains are optional and expendable.

Genes are biologically fundamental, certainly. You will get no argument from me there (although the fact that brains are biologically expendable does not imply that it is moral to expend them). The evidence that memes are more fundamental than brains, however, is not nearly as strong.

It is quite possible to model memes as "reproducing," by being passed from one brain to another. But most of the time the reason the meme is passed from one brain to another is because they aid the brain in fulfilling its desires in some way. The memes associated with farming, for instance, spread because they helped the brain fulfill its desire to not starve. In instances where brains stopped needing the farming memes to obtain food (such as when the Plains Indians acquired horses and were suddenly able to hunt bison more efficiently) those memes promptly died out.

There are parasitic memes, cult ideologies for instance, that reproduce by exploiting flaws in the brain's cognitive architecture. But the majority of memes "reproduce" by demonstrating their usefulness to the brain carrying them. You could say that a meme's "fitness" is measured by its usefulness to its host.

Comment author: timtyler 20 June 2012 12:15:42AM *  1 point [-]

You could say that a meme's "fitness" is measured by its usefulness to its host.

That wouldn't be terribly accurate, though. Smoking memes, obesity memes, patriotism memes, and lots of advertising and marketing memes are not good for their hosts, but rather benefit those attempting to manipulate them. However, there's usually a human somewhere at the end of the chain today.

That probably won't remain the case, though. After the coming memetic takeover we are likely to have an engineered future - and then it will be memes all the way down.

Comment author: Ghatanathoah 20 June 2012 10:03:38PM -1 points [-]

That probably won't remain the case, though. After the coming memetic takeover we are likely to have an engineered future - and then it will be memes all the way down.

The memetic takeover you describe would just consist of intelligences running on computer-like substrates instead of organic substrates. That isn't morally relevant to me, I don't care if the creatures of the future are made of carbon or silicon. I care about what sort of minds they have, what they value and believe in.

I'm not sure referring to an intelligent creature that is made of computing code instead of carbon as a "meme" is true to the common definitions of the term. I always thought of memes as contagious ideas and concepts, not as a term to describe an entire intellect.

After the memetic takeover there would still be intelligent creatures, they'd just run on a different substrate. Many of them could possibly be brain-like in design or have human-like values. They would continue to exchange memes with each other just as they did before, and those memes would spread or die depending on their usefulness to the intelligent creatures. Just like they do now.

Comment author: timtyler 21 June 2012 11:45:05PM *  0 points [-]

I'm not sure referring to an intelligent creature that is made of computing code instead of carbon as a "meme" is true to the common definitions of the term.

People don't call the works of Shakespeare a "meme" either. Conventionally, such things are made of memes - and meme products.

Comment author: asparisi 15 June 2012 12:46:36AM 3 points [-]

you might as well say "being attracted by gravity" or "being made of matter" are what nature commands.

This makes me want to start a religion where the Creator of the Universe gives points to things that behave like a member of the universe. "Thou shalt be made of matter." "Thou shalt be attracted by gravitational force." "Thou shalt increase entropy." etc. Too bad 'Scientology' is taken as a name. Physianity, maybe?

Comment author: Dolores1984 15 June 2012 01:18:25AM *  9 points [-]

In the beginning, there was nothing. The cosmos were void - timeless, and without form. And, lo, God pointed upon the abyss, and said 'LET THERE BE ENERGY' And there was energy. And God pointed to the energy, and said, 'and let you be bound among yourselves that you may wander the void together, proton to neutron, and proton to proton, and let the electrons always seek their opposite number, within the appropriate energy barrier, and let the photons wander where they will.' Lo, and god spoke to the stranger particles, for some time, but what He said was secret. And God saw hydrogen, and saw that it was good.

And God saw the particles moving at all different speeds, away from one another, and saw that it was bad, and God said 'and let the cosmos be bent and cradle the particles, that they may always be brought back together, though they be one billion kilometers apart, within the appropriate energy barrier, of course. And let the curvature of space rise without end with the energy of velocity, that they all be bound by a common yoke.' And god looked upon the spirals of gas, and saw that it was good.

And god took the gas and energy above, and the gas and energy below, and said 'and you shall be matter, and you shall be antimatter, and your charges shall ever be in conflict, and never the twain shall meet, except in very small quantities.' And so there was the matter and the antimatter.

And God saw the cosmos stretching out to a single future, and said 'And let you all be amplitude configurations, that you may not know thyself from the thy neighbor, and that the future may expand without end.' And god saw the multiverse, and saw that it was good.

Comment author: timtyler 15 June 2012 12:50:21AM -2 points [-]

The idea that evolution tends to do that is an illusion created by the Second Law of Thermodynamics. Because of the was the 2LTD works, doing anything for any reason tends to increase entropy. So obviously if an evolved organism does anything at all, it will end up increasing entropy. This creates an illusion that organisms are trying to maximize entropy.

This should be mistunderstanding #1 in the MEP FAQ. MEP is not the same as the second law. It's a whole different idea, which you don't appear to know anything about.

Comment author: pragmatist 15 June 2012 02:15:37AM *  4 points [-]

Extremum rate principles like MEP have proven very useful for describing the behavior of certain systems, but the extrapolation of the principle into a general law of nature remains hugely speculative. In fact, at this point I think the status of MEP can be described as "not even wrong", because we do not yet have a a rigorous notion of thermodynamic entropy that extends unproblematically to nonequilibrium states. The literature on entropy production usually relies on equations for the entropy production rate that are compatible with our usual definition of thermodynamic entropy when we are dealing with quasistatic transformations, but if we use these rate equations as the basis for deriving a non-equilibrium conception of entropy we get absurd results (like ascribing infinite entropy to non-equilibrium states).

Dewar's work, which you link below, is an improvement, in that it operates with a notion of entropy that is clearly defined both in and out of equilibrium, derived from the MaxEnt formalism. But the relationship of this entropy to thermodynamic entropy when we're out of equilibrium is not obvious. Also, Dewar's derivation of MEP relies on applying some very specific and nonstandard constraints to the problem, constraints whose general applicability he does not really justify. If I were permitted to jury-rig the constraints, I could derive all kinds of principles using MaxEnt. But of course, that wouldn't be enough to establish those principles as natural law.

Comment author: timtyler 15 June 2012 10:02:53AM *  0 points [-]

Extremum rate principles like MEP have proven very useful for describing the behavior of certain systems, but the extrapolation of the principle into a general law of nature remains hugely speculative. In fact, at this point I think the status of MEP can be described as "not even wrong", because we do not yet have a a rigorous notion of thermodynamic entropy that extends unproblematically to nonequilibrium states.

Entropy and MEP are statistical phenomena. Thermodynamics is an application This has been understood since Boltzmann's era. Most of the associated "controversy" just looks like ignorance to me.

Entropy maximisation in living systems has been around since Lotka 1922. Universal Darwinism applies it to all CAS. Lots of people don't understand it - but that isn't really much of an argument.

Comment author: timtyler 15 June 2012 12:58:53AM *  0 points [-]

A better (metaphorical) maximand might actually be local entropy minimization. It's obviously impossible to minimize total entropy, but life has a tendency to decrease the entropy in its local area. Life tends to use energy to remove entropy from its local area by building complex cellular structures. It's sort of an entropy pump, if you will. So if we metaphorically pretended it evolution had a purpose, it would actually be the reverse of what you claim.

Prigogine actually came up with a genuine entropy minimization principle once (in contrast to your idea - which has never been formalised as a real entropy minimization principle - AFAIK). He called it the theorem of "minimum entropy production". However, in "Maximum entropy production and the fluctuation theorem" Dewar explained it as a special case of his "MaxEnt" formalism.