What is the Singularity Summit?

post by Liron · 2009-09-16T07:18:06.675Z · LW · GW · Legacy · 17 comments

Contents

17 comments

As you know, the Singularity Summit 2009 is on the weekend of Oct 3 - Oct 4. What is it, you ask? I'll start from the beginning...


 

An interesting collection of molecules occupied a certain tide pool 3.5 to 4.5 billion years ago, interesting because the molecule collection built copies of itself out of surrounding molecules, and the resulting molecule collections also replicated while accumulating beneficial mutations. Those molecule collections satisfied a high-level functional criterion called "genetic fitness", and it happened by pure chance.

If you think about all the possible arrangements of atoms that can occupy a 1-millimeter by 1-millimeter by 1-millimeter cube of space, most of them are going to suck at causing the future universe to contain copies of themselves. Genetic fitness is a vanishingly small target in configuration-space.

And if you studied the universe 5 billion years ago, you would not see a process capable of hitting such a small target. No physical process could create low-entropy collections of atoms satisfying high-level functional criteria. The second law of thermodynamics thus ensured that mice, as well as mousetraps, were physically impossible.

Then a mutating replicator randomly emerged, and suddenly Earth was home to something special: the process of Natural Selection. Natural Selection optimizes for genetic fitness. It squeezes the space of possible futures into a tiny subspace -- the space of universes that contain self-replicators which are very good at self-replicating. And it remained a flickering candle of optimization in a dark, random universe for three billion years.

An interesting product of Natural Selection occupied a certain region of savannah 100 thousand to 2 million years ago, interesting because it could form internal representations of the world around it and predict the consequences of its own actions. By pure chance, Natural Selection had created its successor.

Thought is a more powerful process than Natural Selection. Thought can optimize atom configurations much faster than Natural Selection can. It takes much less time to think of a big design improvement for an organism, than to breed it for as many generations as it takes for a specimen to manifest one.

Now remember, Natural Selection emerged by coincidence -- not by Natural Selection. Processes that optimize for genetic fitness were previously not to be found in the universe. And remember, Thought was evolved by coincidence -- not by Thought. Organs that represent the world around them and make predictions were previously not to be found among the optimized organisms of Earth.

It is still early in the age of optimization processes. Brains are not very good equipment for doing optimization -- Natural Selection just hacked them together out of cells. Yet, Thought is much more powerful than Natural Selection. So what happens when Thought designs an optimization process more powerful than Thought?

What happens when that optimization process designs an optimization process that is more powerful still?

The Singularity Summit is about the critical transition into the third era of optimization processes, the successor to human Thought. To say we need to be careful about initial conditions is to make the understatement of our own entire era.

17 comments

Comments sorted by top scores.

comment by Liron · 2009-09-16T11:05:39.684Z · LW(p) · GW(p)

The goal was to make a Singularity Summit post you can show to smart people who haven't heard of the intelligence explosion.

Replies from: dclayh
comment by dclayh · 2009-09-16T18:30:28.804Z · LW(p) · GW(p)

Very nice. A distinct Douglas Adams flavor to it.

comment by Bongo · 2009-09-17T15:36:57.466Z · LW(p) · GW(p)

Is this historical story of optimization predictively applicable or just a story?

Replies from: arundelo
comment by arundelo · 2009-09-17T17:07:29.013Z · LW(p) · GW(p)

The story points out the two previous times that a new and much better type of optimization process appeared. There's no particular reason to think that there will never be another one, and it seems plausible that a sufficiently good optimization process that works on itself (1) could be the third and (2) could run on a computer.

comment by Vladimir_Nesov · 2009-09-16T20:34:25.303Z · LW(p) · GW(p)

Could we please NOT have this discussion on the forum?

Replies from: Eliezer_Yudkowsky
comment by Eliezer Yudkowsky (Eliezer_Yudkowsky) · 2009-09-16T21:00:27.674Z · LW(p) · GW(p)

Agreed. I'll leave the original main comment, but after this, creationism (called ID or otherwise) is cause for comment removal.

Replies from: CannibalSmith, thomblake, DS3618
comment by CannibalSmith · 2009-09-17T15:18:01.777Z · LW(p) · GW(p)

Disagreed. -16 vote is sufficient to inform readers of how worthy the content is of their attention. If Vladimir doesn't heed the warning, it's his problem.

Replies from: Vladimir_Nesov
comment by Vladimir_Nesov · 2009-09-17T17:12:01.321Z · LW(p) · GW(p)

What's the downside? The problem is that you have to vote it down first, and some people won't stop posting, thus adding more and more noise to sort out, furthermore some people will answer, etc. It's more robust this way, in uncontroversial cases.

Maybe there should be a more general rule: people have to stop posting (at least for a few days) on a topic if their comments on these topics receive consistent significant negative votes.

Replies from: CannibalSmith
comment by CannibalSmith · 2009-09-17T18:47:38.760Z · LW(p) · GW(p)

Voting down the root comment of the offending comment tree is enough to hide it entirely regardless how big it is.

Mind you, repeatedly posting out of context top level comments to subvert the tree structure would constitute spam which is another matter entirely.

Replies from: Vladimir_Nesov
comment by Vladimir_Nesov · 2009-09-17T18:55:38.419Z · LW(p) · GW(p)

It doesn't work like this in the comment feed.

comment by thomblake · 2009-09-17T18:06:50.180Z · LW(p) · GW(p)

Why no discussing creationism/ID on relevant topics? Do you have a justification for this policy like "politics is the mind-killer" or are there some ideas we just don't want to have to argue? Is it really enough noise to kill the channel?

comment by DS3618 · 2009-09-17T00:31:46.326Z · LW(p) · GW(p)

"Agreed. I'll leave the original main comment, but after this, creationism (called ID or otherwise) is cause for comment removal."

I never argued for ID or creationism (the closest I have come is arguing for a more complete understanding of the topic before bashing it), I have been merely pointing out that Evolution has some serious holes. If you believe it so blindly that you can't see the holes then you haven't done your homework. If Darwin was alive today many speculate he would have never bothered to formulate evolution since the unfortunate discovery that cells are more the blobs of protoplasm. I am actually arguing for a re-working of the entire idea of a naturalistic explanation of the world.

But whatever, I give up... you guys win, your bonehead comments have won... I need not waste more time here...

En effet la règle ignorante ici (how's your french?)

Replies from: CarlShulman
comment by CarlShulman · 2009-09-17T01:53:00.569Z · LW(p) · GW(p)

Before you leave, could you confirm that you did go to CMU for graduate school?

comment by Liron · 2009-09-16T11:04:51.014Z · LW(p) · GW(p)

The goal was to make a singularity Summit post you can show to smart people who haven't heard of the intelligence explosion.

comment by DS3618 · 2009-09-16T07:46:10.378Z · LW(p) · GW(p)

Did you read what you wrote here?

"An interesting collection of molecules occupied a certain tide pool 3.5 to 4.5 billion years ago, interesting because the molecule collection built copies of itself out of surrounding molecules, and the resulting molecule collections also replicated while accumulating beneficial mutations. Those molecule collections satisfied a high-level functional criterion called "genetic fitness", and it happened by pure chance."

So essentially ignoring the rest of your post because it just goes into more "detail" we have some molecules + lots of magic=the world as we know it. Then followed by a big please accept my straw-man argument therefore the Singularity Summit is necessary. Does that about sum it up?

Just an aside: Have you ever considered that naturalistic explanations cannot explain DNA, or the Cambrian explosion. Have you ever looked at chemistry, quantum mechanics,... and even tried to figure out how that could work? What about considering the early earth and the available chemicals which are essentially what comes out of a volcano? If you have then I challenge you to explain how through naturalistic processes you can form even the simplest bacteria (which by the way has ~160 kilobases).

Replies from: Mitchell_Porter, Mitchell_Porter
comment by Mitchell_Porter · 2009-09-16T10:24:12.966Z · LW(p) · GW(p)

A brief manifesto regarding what is mysterious, what is not mysterious, and the nature of reality.

Some things that are mysterious:

  • The existence of anything at all is fundamentally mysterious. Trying to explain why there is something rather than nothing leads to odd and perhaps dubious concepts like a "necessary being", and the alternative is just to say there is no explanation at all.

  • The existence of consciousness, with its manifest properties, is fundamentally mysterious if you suppose a world in the image of today's mathematical physics. Just to be tedious and repeat a fact stated many times, but something as basic as experienced color is just not there, in a universe made of nothing but colorless particles. However, the strongly validated part of physics is the mathematical part. There is considerable room to rethink the nature of the entities which behave according to the equations, and I for one see an opening here to solve the problem of consciousness.

  • That we have "laws of physics" is semi-mysterious and possibly fundamentally mysterious. In a way it follows naturally from the premise that the world contains things which have different "natures" and that their behavior is determined by this nature. A physical law is then just a statement of how they behave. But these basic ontological terms - substance, property, nature - are all potentially mysterious in themselves. When you try to define basic ontological terms, it starts to sound like poetry, or like circular definition. So I think ontology - not ontology in the computer-science sense of sub-sub-categorization, but fundamental ontology, the nature of existence and that which exists - is an area we do not understand at all well, and its mysteries are probably continuous with the first two mysteries I mentioned above.

Some things that are not mysterious:

  • That life - in the sense of complex self-replicating entities - can develop in a world of atoms behaving according to physical law is not mysterious. If it needed any demonstration, the use of evolution made in computing proves that you can generate ingeniously adapted complexities through darwinian procedures. The combinatorial possibilities of organic chemistry and the nature of the physical environment on Earth can together plausibly be regarded as sufficient to explain the phenomenon of life, and in practice chemistry and genetics are showing us that the explanations can be found.

  • That a material system can exhibit properties of intelligence like problem-solving or goal-directed behavior, in a world of atoms behaving according to physical law, is also not mysterious. This is distinct from the problem of consciousness above. The nature of subjectivity is ontologically problematic - hard to characterize even in itself, let alone map onto the physical ontology - but if the question is just, how can a thing made of atoms solve problems or achieve goals, it is not mysterious. Computer science, pattern recognition, algorithms of all sorts show that it can be done.

So returning to your challenge to Liron's worldview, I'd say life is not a problem, consciousness is, but there is nothing to show that consciousness is a challenge to the evolutionary explanation of life or the computational explanation of practical intelligence.

Finally, the nature of reality. It does in fact appear to be a logical possibility that atomistic materialism is radically wrong. For example, one might suppose a universe consisting of one or more fundamentally mind-like entities who are just hallucinating the whole affair, which act upon the material part of the universe 'directly' rather than in a way mediated by many elementary physical interactions, which are the product of a nonevolutionary process of creation, and so forth. However, when it comes to judging the plausibility of such scenarios from the evidence one has, they are basically on a par with the scenario out of The Matrix according to which I'm a body in a pod, or a brain in a vat, being fed images of a simulated world. All these scenarios require that a large part of mundane experience be regarded as false or as basically misleading. Again, it's a logical possibility, but it is one that, if it is to be addressed rationally, has to be addressed on some other plane entirely than the one which one uses to make everyday decisions. If the world is not interpreted as a giant illusion, then the weight of everyday evidence already favors (in a broad sense) physics as the explanation of observable events, evolution as the explanation of life, and computation as the explanation of intelligence.

comment by Mitchell_Porter · 2009-09-16T09:25:34.941Z · LW(p) · GW(p)

DS3618: "Have you ever considered that naturalistic explanations cannot explain DNA, or the Cambrian explosion... I challenge you to explain how through naturalistic processes you can form even the simplest bacteria (which by the way has ~160 kilobases)"

My formula for the origin of life is "RNA world plus micelles".

A micelle is a self-organized sphere of hydrophobic molecules. "RNA world" refers to a stage when you don't have the division of labor between DNA (information) and protein (structure), with RNA instead doing double duty (information from sequence, structure from conformation). Lipids and RNA polymers are capable of forming spontaneously in abiotic circumstances. So the idea is that the lipids mechanically self-organized into populations of micelles, which in turn contained different populations of RNA polymers. RNAs can both reproduce (in that one RNA strand can serve as a template for the formation of a second) and catalytically interact with each other (thus increasing or decreasing the reproduction rate of the other RNA species in a population). Finally one supposes a physical process, such as turbulence, which keeps breaking up these RNA-loaded micelles. Voila, you have a population of protocells subject to natural selection.

This is just a sketch of how biological evolution might get underway, by someone who is not even a biologist. But I don't think it's that hard to understand.

And as for the Cambrian explosion, if you know anything about how tissue differentiation and embyronic development work, I don't see how you could regard it as fundamentally mysterious. Sponges are the primordial multicellular lifeform and they already have the relevant genetic regulatory networks. Once that level of genetic complexity exists, it's not that amazing to see how it could lead to the diversity of multicellular life that we have now. So in a way the question is just, how did we get sponges. I am fond of the theory (promoted by John Mattick and others) that the essential new step involved in the transition to complex multicellularity is the development of a new level of genetic regulation involving transcribed intronic RNAs which never become protein but which do play a regulatory role controlling other genes. The idea is that this extra level of regulatory complexity permitted the transition from homogeneous colonies of single-celled organisms to multicellular organisms with differentiated tissues. But again, the bottom line is that this stuff does not look fundamentally mysterious. There's plenty to discover about how it might have happened and how it did happen, but it's not baffling, let alone naturalistically impossible.