Origin > How to Build a Brain > How can a small number of genes build a complex mental machine?
Permanent link to this article: http://www.kurzweilai.net/meme/frame.html?main=/articles/art0387.html

Printable Version
    How can a small number of genes build a complex mental machine?
by   Gary F. Marcus

We have only around 35,000 genes, but tens of billions of neurons. How does a relatively small set of genes combine to build a complex brain? Gary F. Marcus responds to Edge publisher/editor John Brockman's request to futurists to pose "hard-edge" questions that "render visible the deeper meanings of our lives, redefine who and what we are."


Originally published January 2002 at Edge. Published on KurzweilAI.net January 21, 2002. Read Ray Kurzweil's Edge question here.

John McCarthy and I are from different generations (in the semester before McCarthy invented Lisp, he taught my dad FORTRAN, using punch cards on an old IBM) but our questions are nearly the same. McCarthy asks "how are behaviors encoded in DNA"?

Until recently, we were not in a position to answer this question. Few people would have even had the nerve to ask it. Many thought that most of the brain's basic organization arose in response to the environment. But we know that the mind of a newborn is far from a blank slate. As soon as they are born, babies can imitate facial gestures, connect what they hear with what they see, tell the difference between Dutch and Japanese, and distinguish between a picture of a scrambled face and a picture of a normal face. Nativists like Steven Pinker and Stanislas Dehaene suggest that infants are born with a language instinct and a "number sense". Since the function of our minds comes from the structure of our brains, these findings suggest that the microcircuitry of the brain is innate, largely wired up before birth. The plan for that wiring must come in part from the genes.

The DNA does not, however, provide a literal blueprint of a newborn's mind. We have only around 35,000 genes, but tens of billions of neurons. How does a relatively small set of genes combine to build a complex brain? As Richard Dawkins has put it, the DNA is more like a recipe than a blueprint. The genome doesn't provide a picture of a finished product, instead it provides a set of instructions for assembling an embryo. Those instructions govern basic developmental processes such as cell division and cell migration; it has long been known that such processes are essential to building bodies, and it now is becoming increasingly clear that the same processes shape our brains and minds as well.

There is, however, no master chef. In place of a central executive, the body relies on communication between cells, and communication between genes. Although the power of any one gene working on its own is small, the power of sets of genes working together is enormous. To take one example, Swiss biologist Walter Gehring has shown that the gene pax-6 controls eye development in a wide range of animals, from fruit flies to mice. Pax-6 is like any other gene in that it gives instructions for building one protein, but unlike the genes for building structural proteins like keratin and collagen because the protein that pax-6 builds serves as a signal to other genes, which in turn build proteins that serve as signals to still other genes. Pax-6 is thus a "master control gene" that launches an enormous cascade, a cascade of 2,500 genes working together to build an eye. Humans that lack it lack irises, flies that lack it lack eyes altogether. The cascade launched by pax-6 is so potent that when Gehring triggered it artificially on a fruit fly's antenna, the fly grew an extra eye, right there on its antenna. As scientists begin to work out the cascades of genes that build the brain, we will finally come to understand the role of the genes in shaping the mind.

Copyright © 2002 by Edge Foundation, Inc.



www.edge.org

 Join the discussion about this article on Mind·X!  
 

   [Post New Comment]
   
Mind·X Discussion About This Article:

article by G.Marcus, Brains from Genes
posted on 01/23/2002 3:49 PM by kwa@prospero.ch

[Top]
[Mind·X]
[Reply to this post]

The question makes sense, the proposed answer neither.
The article may be too short for going into details. The article, however, addresses the question in the wrong way.
(So far I do not know either, but probably soon)

Well, what is going wrong with the way G.Marcus approaches this issue?

Basically, hoping for the geneticists distracts the physical realm and the realm of meaning.

David Marr pointed it out as early as 1981 (shortly before he died, unfortunately). There is a level of physical realization (genes), algorithmic representation (mechanisms) and computational framework (capability for meaning). Arguing on the level of genes and hoping for a solution is like being an amoeba thinking about complexity.

The brain is a semiosic entity (see C.Peirce), capable for meaning. What is meaning? Take a look to Wittgenstein's Philosophy of Psychology. They moved towards a more useful direction than the naive reductionism does.

Any answer about the brain will be accessible only after answering the following questions

- What is novelty?
- What is a function?
- What is a process?
- What is similarity?
- What is meaning?

only to give a small number of relevant questions. The issue of Pax-6 will never be relevant for questions about meaning, consciousness and intelligent self-organised problem-solving.

regards
Klaus Wassermann
23-1-02






















Re: How can a small number of genes build a complex mental machine?
posted on 06/15/2002 5:33 PM by mindidea@hotmail.com

[Top]
[Mind·X]
[Reply to this post]

It is never disscused about the devoloupment of mind, that we are a creature born of the womb. It would be be a terrible waste of reasources to let the time spent there not be involved in copying a great deal of the mind. The structure of language and the obvious skills of communication displayed by both prenates and newborns accompany skills of perception. Babies begin formal language training in the womb. An early discovery using acoustic spectrography revealed that the initial cry of a 900 gram baby already contained intonations, rhythms, and other speech features of the baby,s mother. By about 26 weeks of gestation, this baby had acquired certain features of its "mother tongue". We do not know how much of the mind is copied directly from our mothers. The genetic code either has a blueprint for building a intellegent concious mind or the address to find this information. This is the only avenue we have to realize strong AI in our lifetime. The approches I have seen so far, are much to simplistic to realize this in a thousand years. How can we get this kind of information from the code? Is anyone working on this avenue? Where do we start?

Re: How can a small number of genes build a complex mental machine?
posted on 05/18/2004 2:49 PM by greiner

[Top]
[Mind·X]
[Reply to this post]

I have a question. What about genetic memory. I watched a program about birds building nests. Different species built different styles. Now it was shown that they don't learn this behavior, but instead know it from birth. Seems pretty intriguing that a bird can build a complex structure without ever being shown. So that leads me to wonder then, would it be possible to genetically alter DNA so that an infant from birth could perform basic math, and or speak multiple languages? I understandstand that there would be physiological issues of a newborn, but you get the general idea.

Re: How can a small number of genes build a complex mental machine?
posted on 05/18/2004 3:04 PM by _CriX_

[Top]
[Mind·X]
[Reply to this post]

That seems to logically follow.

That's an amazing thought. We may be engineered to know certain things at birth. Wow, that is scary.

Hopefully there will be laws at that point that only mechanical routines (like performing arithmatic, standard warp core repair :) can be genetically input. I would hope they would leave religion and personal values out of it.

Re: How can a small number of genes build a complex mental machine?
posted on 07/18/2004 3:30 PM by JPryne

[Top]
[Mind·X]
[Reply to this post]

To me, innate nest-building skills seem rather "fuzzy" as opposed to higher concepts such as mathematics & spoken languages.

These various species of bird have instincts for scavenging material & weaving it together in concave clusters.

Math & language have complex sets of rules & esoteric concepts. To implement them explicitly would require the biological equivalent of a ROM-like database, different than our algorithmic genes. It would be like trying to breed an orchid that had the lyrics from "Stairway to Heaven" printed on it's petals.

I do suppose, however, that DNA could be optimized to focus more resources upon the development of structures & mechanisms most supportive of linguistic or logical talent. For example, a subject might be tailored to posses a greatly increased portion of brainspace to rote memory, or even the perception of thoughts & feelings in others (& one's own self) as language is so much more than memorization. It is eye contact, body language, tone of voice, innuendo, and so on. The faculties dealing with these subtle arts might well be strengthened.

Such an optimized subject may easily absorb a dozen languages as simply as we learn a single native tongue. But, of course, I conjecture...

Re: How can a small number of genes build a complex mental machine?
posted on 05/19/2004 10:49 AM by zurk@arbornet.org

[Top]
[Mind·X]
[Reply to this post]

huh ? small number of genes ?
even with our present technology decoding a large part of gene information shows us that there is at least 800-1000 Megabytes of information stored in genes.
Thats more than a hundred times the size of the kernel in a modern operating system on a modern computer. and genes encode far more information relating to complex chemical interactions and everything else.

Re: How can a small number of genes build a complex mental machine?
posted on 05/19/2004 11:56 AM by /:setAI

[Top]
[Mind·X]
[Reply to this post]

huh ? small number of genes ?
even with our present technology decoding a large part of gene information shows us that there is at least 800-1000 Megabytes of information stored in genes.


38 thousand genes and "other" sequences encoded into less than a Gb IS a tiny amount of information to build the most complex system in the known universe- a system that couldn't be reduced to any less than exabytes- perhaps zettabytes of information to even roughly describe just the crucial molecular/ cellular/ autopoietic relationships and relationships of relationships in the entire biochemical hierarchy of the human body and mind-

___________________________

/:set\AI transmedia laboratories

http://setai-transmedia.com

Re: How can a small number of genes build a complex mental machine?
posted on 05/19/2004 3:01 PM by zurk@arbornet.org

[Top]
[Mind·X]
[Reply to this post]

rubbish. a gig is a huge amount of information for what is essentially a blueprint.
a blueprint of a modern jet aircraft or a car can be reduced down to a set of numbers which equates to less than 5 megabytes.
the exabytes youre getting is after the finished product is built and starts functioning. genes are the blueprint to build the machine (human or animal) and bootstrap it..they do nothing else.
1 gig of complex non reducible information for a machine is a LOT of information....far more than any other device we have built today.

Re: How can a small number of genes build a complex mental machine?
posted on 05/19/2004 4:50 PM by /:setAI

[Top]
[Mind·X]
[Reply to this post]

rubbish? now you are merely being confrontational based on lack of understanding of information and biology- you won't find a computer scientist or biologist alive that was not SHOCKED that the human genome only contained 38 thousand genes- while an individual gene represents a lot of data PER SIZE- the data itself is simply very small- this is a very well publisized and universally agreed upon observation- a Gb might seem like a lot to someone only familiar with PCs and office documents- but in complex fields 10^9 bytes is a miniscule amount of information to result in systems which represent at least 10^18- 10^21 bytes at any given time

you simply don't understand how TINY a gigabyte of information is when considering the quinillion-fold IRRIDUCIBLE non-linear complexity of biological organisms with trillions of cells arranged in vastly entagled hierarchies of function- it's the equivalent of building a supercomputer and all it's software-from an instruction set less than a dozen bits long-

___________________________

/:set\AI transmedia laboratories

http://setai-transmedia.com

Re: How can a small number of genes build a complex mental machine?
posted on 05/19/2004 5:26 PM by zurk@arbornet.org

[Top]
[Mind·X]
[Reply to this post]

err..do you actually know how much code there is to run a supercomputer ?
UNICOS which runs on crays has a 10 megabyte kernel image and the computer hardware has a 300K bootstrap loader.
Thats 10.3 megabytes for your supercomputer.
Add another 2-3 megs for the VHDL which defines the vector processing chips and youre in the 15 megabyte range for your supercomputer, hardware and software.
1 gigabyte for a BLUEPRINT is a friggin BOATLOAD of information. its a HUNDRED FOLD increase in complexity to a supercomputer.
the reason you get exabytes of information is because the DNA is copied trillions of times and executed with different jump entry points, depending on the cells used, to unravel and build the machine (human or animal) from the blueprint.

Re: How can a small number of genes build a complex mental machine?
posted on 05/19/2004 7:08 PM by /:setAI

[Top]
[Mind·X]
[Reply to this post]


for the anaology to work- you have to be able to PHYSICALLY build a supercomputer's HARDWARE and SOFTWARE from a pile of silicon gold and plastic using an encoded object of a few bits long- like a single protein-

you just don't understand HOW much work DNA has to do- it is NOT a "blueprint"- it's a complete program that can build something 10 orders of magnitude more complex than itself- no other process we know of can come close to that

do you have any idea how many scientists skoffed at the initial findings of the genome race becasue they thought that what was discovered was just too absurdly small of a data-set?

___________________________

/:set\AI transmedia laboratories

http://setai-transmedia.com

Re: How can a small number of genes build a complex mental machine?
posted on 05/20/2004 12:20 AM by tharsaile

[Top]
[Mind·X]
[Reply to this post]

Okay, so setai says it's a small number of genes and zurky says it's a lot.

What about the rest of the genome? Would anyone like to comment on last week's (not)junk DNA article?

http://www.washingtonpost.com/ac2/wp-dyn/A9055-200 2Dec4?language=printer

It always seemed kind of premature to dismiss those as so-called junk DNA. Now we know it was premature.

Re: How can a small number of genes build a complex mental machine?
posted on 05/20/2004 8:08 AM by griffman

[Top]
[Mind·X]
[Reply to this post]

for an equal comparison between DNA and a Cray you'd have to look at the hardware architecture blueprint PLUS the software data set. so the 10 meg size kernel is like looking at the activation enzyme.

after a short bout with google, its pretty clear neither Cray or Intel are very vocal on how large their hardware data set is. but considering the number of employees working on any one processor blueprint. I'd bet the farm its larger than 10 meg.

griffman

Re: How can a small number of genes build a complex mental machine?
posted on 05/20/2004 8:09 AM by griffman

[Top]
[Mind·X]
[Reply to this post]

anyone from Intel willing to chime in?



does anyone at intel even pay attention here?

cheers
griffman

Re: How can a small number of genes build a complex mental machine?
posted on 05/20/2004 4:02 PM by zurk@arbornet.org

[Top]
[Mind·X]
[Reply to this post]

the hardware data set is the VHDL that i commented on earlier. for a vector processor like the Cray series is 2-3 megs max.
in extremely simple terms VHDL or verilog code is compiled into the chip mask which is then burned to form your processor chips. 2-3 megs of VHDL code typically produce a chip with 300-400 million transistors (a lot).
http://www.opencores.org has some RISC processors for download (as VHDL/verilog).
the entire supercomputer can be built from silicon and bootstrapped with less than 15 megs of data as i noted earlier.

Re: How can a small number of genes build a complex mental machine?
posted on 05/20/2004 4:30 PM by /:setAI

[Top]
[Mind·X]
[Reply to this post]

what about the power supplies and distribution systems- the cooling systems- the I/O devices- cases and chasis- etc? all of these thing must be built by the code as well- so we have the chips- the software AND the machinery and infrastructure all have to be built from a pile of silicon/plastic/metals by a replicating code that could only be about 10 bits long to accuratley reflect what DNA does- that would be some fancy programming!

___________________________

/:set\AI transmedia laboratories

http://setai-transmedia.com


Re: How can a small number of genes build a complex mental machine?
posted on 05/21/2004 11:08 AM by zurk@arbornet.org

[Top]
[Mind·X]
[Reply to this post]

as i noted the chips and the OS are the most complex parts. blueprints for power supplies and such are so simple theyre almost irrelevant...around 100K will describe the case (which is just a hunk of metal, 6 vectors with the chemical composition of the metal to describe it - less than 100 bytes), the power supply circuitry and all the rest.
thats one reason why your nails (finger/toe) arent well described in DNA -- they're just dead cells, no further description needed.
note that DNA is a blueprint for nanomachines to construct you from scratch, while a regular blueprint describes construction of a machine to an engineer and the VHDL/verilog is compiled to describe the chips to a array of transistors (mask). once we actually have nanotech i believe we will be able to reduce DNA down to around 1/10th of its length since we're considerably smarter than evolution-based engineering systems. a human should only take around 100 megs or so to build...we're not THAT complex as compared to other animals (electric eel, birds, dolphins etc). i'm also betting that sharks have more complex DNA than humans and have more information per DNA strand than humans.

Re: How can a small number of genes build a complex mental machine?
posted on 05/20/2004 5:00 PM by /:setAI

[Top]
[Mind·X]
[Reply to this post]

my feeling is that the reason for such complexity from so few genes is similar to the reason why brains/ autopoietic sytems/ economies/ ecologies/ universes are so complex with just a few fundemental components: modular hierarchy- that is the 38 thousand genes of the human genome aren't simply activated singularly or in linear combinations for each specific task/form- but any GROUP of genes can be activated to result in unique functions- and groups of groups can form/interact- so that DNA acts as a modular matrix of gene adgents/networks that result in very complex functions through entanggled hierarchies and networks of activated genes whose independat action results in complex hierarchical network dynamics-

instead of 38 thousand gene instructions- you would have the potential for trillions of trillions of trillions of possible modular matrix "patches" of active genes each providing specific yet unique functions and enzyme instructions- [and families of types of groups]these kinds of modular hierarchies naturally emerge from selection- the set of kingdoms/species/cell-types/biologics that we have observed in life on earth is only an infantessimal set of modular patches in VERY large matrix of possible/viable patch-programs for DNA- I think that exploring these tranfinite possibilites of living systems will be a major aspect of post-singularity existence-

___________________________

/:set\AI transmedia laboratories

http://setai-transmedia.com

Re: How can a small number of genes build a complex mental machine?
posted on 03/31/2007 6:29 AM by aaronsloman

[Top]
[Mind·X]
[Reply to this post]

I've only recently noticed this thread and I expect nobody is reading it any more. However an important clue to the answer lies in the fact that six short axioms (Peano's axioms for arithmetic) and some fairly simple logical machinery can together specify an infinite structure of infinite complexity, in the sense that there seem to be infinitely many distinct non-trivial theorems in number theory, including infinitely many finitely statable, unprovable facts (in the standard model).
We still have a lot to learn about the relations between specifications and what they specify.

Regarding how the genes specify a human, there is the further factor that epigenesis is a process in which the environment constantly influences the development of the individual. The amount of variation produced by the environment varies from one species to another and in the case of humans seems to vastly exceed all other cases.

There's more on this in a paper co-authored with Jackie Chappell, to appear in Journal of Unconventional Computation, available here:
Natural and artificial meta-configured altricial information-processing systems:
http://www.cs.bham.ac.uk/research/projects/cosy/pa pers/#tr0609

Aaron
http://www.cs.bham.ac.uk/~axs

Re: How can a small number of genes build a complex mental machine?
posted on 03/31/2007 7:23 AM by doojie

[Top]
[Mind·X]
[Reply to this post]

Just throwing in a bit of old fashioned philosophy here, but Sagan, I believe, pointed out that the brain developed when genes could no longer code necessary information. Since that information was no longer capable of being coded in a programmed fashion, the brain would represent alternatives or "sworls" of options that are limited within the genetic replicative algorithm for each species.

Societies are built to limit the choices created in the development of the brain, and those societies follow sociobiological patterns that limit reproduiction through ritual behavior gearded to territory.

Sopcial behavior adapts to specific territory and responds by limiting its opitions to avoid stresses involved in change.

It would seem that what is generated by the combination of genes is merely a general pattern that is enlarged to enhance reproductive success.

The organism is "informed" however, by a virus or bacteria that invades the organism and causes a response of defense and attack to presefve internal integrity.

Two beneficial results occur:

1. The organism goes through a purification process(diarrhea, vomiting, etc) to rid itself of the invader.
2. In identifying the invader, the immune system is enhanced because it has enlarged its "library" of defenses and created an antibody to eliminate the invader. This means that adaptive "intelligence" in general is also enhanced.

This adaptive process incorporates the viral invader into the organism and alters the reproductive process at the cellular level, but only to the extent necessary for adaptation.


The genes, therefore, interact with a totality of genetic structures provided by the environment, usually in the form of viruses, which are merely genetic bits encased in protein.

The "junk DNA" is similar to viruses in their make-up, and may be part of the organism's adaptive response, each bit of DNA corresponding to bits of viral invading DNA, and reacting to form new reproductive behaviors, which may include speciation to new territory, or even an expansion of the species as its adaptabilioty is enhanced.

After these changes occur, the species may evolve new ritual behavior to separate their reproductive processes from similar species.

That same process of ritualization becomes more complex in humans, but at the core of the various rituals lies a replicative algorithm which follows basic patterns that remain similar in many respects.

Complexity increases, but processes can return to basic desires and instincts to maintain reproductive integrity.

The genes, therefore, only have to produce replicative patterns, and those patterns will form into more complex representations, but the complexity always revolves around efficient replicative strategies that are interchangeable.

The human brain has by-passed the necessity of speciation folloowed by animals, buit has merely evolved the replicative algorithm to non-biological processes, so that interchangeability becomes "homogenized" to the degree that we can explore "foreign" areas of thought and apply it in context to change.

It seems to me that both arguments above are correct, but do not take into account the fact that our genes exist within a sea of environmental genes that interact with our own to create new and complex adaptive strategies.