Conceptual structure and practical implementation are two different concepts.
Let us say that I have a spreadsheet in which there are a thousand cells. You may quite
logically tell me that it would be most efficient if used a thousand processors to update my
spreadsheet. Seems logical that if I use one CPU per cell, that the spreadsheet might update
quite quickly. I may totally agree with you in theory. After I have agreed with you, I will still
use Excel for my spreadsheet and I will use only one CPU to simulate your thousand CPUs.
You may be correct in a structural way. If you are, then I will use my one CPU to simulate
your thousand CPUs operating in parallel. Then I will stop and ask, “Which is the more cost
effective implementation of the solution to my problem, 1000 CPUs or 1 CPU?”
What if you asked me to solve the same problem using rat brain cells in a Petrie dish?
http://www.sciencedaily.com/releases/2004/10/041022104658.htm
If that was the problem, then I would probably say to myself, “These brain cells are so slow that
I had better implement my solution using one thousand parallel mini-brains, one for each
spreadsheet cell.
Conclusion:
We should divide the problem into two parts. Part one is “what is the most efficient structure
of a brain?” Part two is “How should we implement our design?”
I have 1 CPU and it is fast enough to simulate the 10 thousand neurons in a flee's brain, then I
may well be fully capable of simulating a flee's brain with a single CPU. The structure may be
best implemented in cells using 10,000 neurons but may be totally simulated as an imitation
of ten thousand cells operating in parallel.
Second fallacy in the thinking:
If I have a PC and replace it with another PC whose CPU is a "Core Duo", is my PC still a PC?
This is an important question because I intend to use mathematical induction to prove my
point.
http://en.wikipedia.org/wiki/Mathematical_induction
To prove mathematical induction, I must assert that adding one more processor to my PC
doesn't stop it from still being defined as a PC. Since this is true (absolutely nobody is
suggesting that if a PC has two processors, it is “no longer a PC") By induction, I may assume
that PCs containing 8 processors will still be referred to as PCs. The Sony Play Station 3 has 8
processors in the CPU package. If there are 8 processors in a PC by the end of 2007, nobody
will be upset or terribly surprised.
If this trend continues the number of processors may well go up by a factor of 4 each year.
This would mean that in 10 years, I may be able to purchase a PC with 1 million processors. (If
it only doubles each year, then it will take 20 years to have 1 million processors) If each
processor operates at 200 billion instructions per second, then each processor would be about
a billion times faster than a neuron operating at 200 pulses per second. I think that it is safe
to say that a computer instruction, such as add two 64 bit numbers together is objectively
"faster" than a nerve which is simply squirting some seratonin at another four dozen nerves.
This would mean that my PC in ten years could simulate at least 1 million processors times 1
billion times the speed of a neuron. This multiplies out to the equivalent of 1 trillion neurons
being simulated real time. If the human brain has 100 billion neurons, then it will be
possible for a PC to simulate ten human brains by simulating each neuron in one of its 1
million processors.
Suggestion:
Stop complaining and start designing the best way to interconnect the million CPUs and start
writing the code to simulate the neurons in the human brain.
If my job is to simulate an insect with a thousand brain cells, I would very likely use one CPU
to simulate the action of a thousand cells acting as a parallel processing insect brain. The
insect brain would be similar to the spreadsheet in that regard. Since my one CPU goes fast
enough to simulate a thousand slow insect brain cells, why would I implement the solution
any other way?
Equivalence is an important word in this discussion.
Let us say that it is 1972 and the 4004 microprocessor chip set has recently been invented by
Intel and we have 1000 of these 4004 chip sets in front of us. Here we are at the forefront of
microprocessor technology and we decide to make a machine which automatically
recalculates spreadsheets for people. The 4004 is so slow that we figure that we will support
spreadsheets with up to one thousand cells and use one processor for each cell. Our logic is
sound because we know that parallel processing is OBVIOUSLY the best way to calculate a
spreadsheet in which each cells contents depends on the contents of the other cells. It sure
pays to be smart and recognize that parallel jobs should be implemented with parallel
hardware. Once we have the spreadsheet computer working we start selling it for $20,000 per
copy.
The arguments being used regarding human brain simulation could be used to argue that:
There will never be a single CPU computer which will be capable of doing spreadsheets,
because the structure of a single CPU machine is “all wrong” for the problem being solved.
We figure that large corporations might buy them. Our If I have a single CPU and it can
simulate a thousand older CPUs, then it may well perform the job that a thousand old CPUs
did. Why would I use a thousand old CPUs in parallel when I could use one new CPU?
Let us say that I have a spreadsheet in which there are a thousand cells. You may quite
logically tell me that it would be most efficient if used a thousand processors to update my
spreadsheet. You may be correct in a structural way, but I may simply tell you that I would
rather use one processor to simulate the action of the thousand processors that you
recommend that I use. If my simulation cost less than your parallel implementation, then in
what way was your suggestion "more efficient"?
If my job is to simulate an insect with a thousand brain cells, I would very likely use one CPU
to simulate the action of a thousand cells acting as a parallel processing insect brain. The
insect brain would be similar to the spreadsheet in that regard. Since my one CPU goes fast
enough to simulate a thousand slow insect brain cells, why would I implement the solution
any other way?
Structure being simulated: parallel spreasheet or parallel insect brain.
Implementation: Today it will be serial simply to reduce cost.
Final submission to http://dangerousintersection.org/?p=178
Let us say that it is 1972 and the 4004 microprocessor chip set has recently been invented by
Intel and we have 1000 of these 4004 chip sets in front of us. Here we are at the forefront of
microprocessor technology and we decide to make a machine which automatically
recalculates spreadsheets for people. The 4004 is slow but we figure that we will support
spreadsheets with up to one thousand cells and use one processor for each cell. Our logic is
sound because we know that parallel processing is OBVIOUSLY the best way to calculate a
spreadsheet in which each cell's contents depends on the contents of the other cells. It sure
pays to be smart and recognize that parallel jobs should be implemented with parallel
hardware. Once we have the spreadsheet computer working we start selling it for $120,000
per copy. We obtain a patent on our technique of using parallel processors to solve
spreadsheet problems. We mortgage our homes and put all of the money into our company
because we absolutely know that we have the very best way to calculate spreadsheets, and it is
patented. We have a lock on the market because we KNOW that parallel jobs are most
efficiently implemented with parallel hardware. If venture capitalists doubt our plan, we tell
them to talk to Gerald M. Edelman, who knows the difference between jobs which are
obviously serial and jobs which are obviously parallel.
The arguments being used regarding human brain simulation with a PC could be used to
argue that:
There will never be a PC computer which will be capable of doing spreadsheets, because the
structure of a PC is “all wrong” for the problem being solved.
Anybody want to buy my parallel processor spreadsheet computer? Marked down to $10,000,
which is a bargain considering that it has 1000 processors and weighs a ton.
don
In the original argument above, the following was stated:
"Paul Churchland also notes that we are horrible at logic and other types of systematic
thinking. How many years do we study math, but look how we still struggle as adults!” The
Engine of Reason, the Seat of the Soul, Paul M. Churchland (1995)."
This is precisely what we are dealing with here. We have people who cannot see through the
bad logic and bad systematic thinking shown in the original argument which claims that
human brains are not like computers. Humans fall for arguments like "humans are better
because our minds are more robust" simply because, as Paul Churchland notes, "we are
horrible at logic and other types of systematic thinking."
Even the person who notes that humans are bad at logic goes on to say, “If the brain were a
computer, this would not be the case.” Here, Paul Churchland is demonstrating that he is a
human and how difficult it is for humans to think logically and systematically. What he
should have said is "If the brain had evolved in an environment where logic was more
important and "hunting, gathering, eating and reproducing" were less important, then
humans would be as logical or more logical than computers are. Then he could have noted
that computers evolved simply because once humans had their eating and reproducing needs
satisfied, they were ready to move on to other tasks requiring more mathematics, logic and
systematic thinking. It is in this latter environment that computers evolved. Life is a solution
to an opportunity in an environment.
So, when you want to decide the truth of any complex logical argument, do you believe the
horrible-at-logic humans or should you ask your computer?
The human brain is simply a large, slow parallel processing computer.
Get used to it.
From above: "Paul Churchland also notes that we are horrible at logic and other types of
systematic thinking. How many years do we study math, but look how we still struggle as
adults! If the brain were a computer, this would not be the case. The Engine of Reason, the
Seat of the Soul, Paul M. Churchland (1995)."
Note the recursive and reflexive qualities of this argument. Humans cannot think logically
and systematically so here is an illogical and unsystematic argument which humans will
believe simply because they aren't very logical and systematic.
The logical and systematic truth is that life evolves in a particular environment and must
evolve those qualities which allow it to survive in that specific environment. In a jungle, an
organic biped omnivore must evolve abilities of agility, dexterity and thinking for the purposes
of hunting, gathering, eating and reproducing.
On a desktop, a computer finds itself in a totally different environment. The eating and
reproducing are all taken care of by electrical outlets and computer factories. In that
environment, a logical and systematic entity would notice that computers should evolve those
different qualities which are lacking yet needed in its environment. Those specific qualities
are logic and systematic thinking.
Now that human hunting has been converted to gathering by the McDonald's drive-thru,
humans have a need for less hunting and more logic and systematic thinking. The problem is
that organic entities evolve so slowly. So, the need for logic and math is being fulfilled by
computers.
Nevertheless, that doesn’t mean that humans, given enough time to evolve, couldn’t
eventually have brains which could be as logical and as systematic as computers. Maybe
genetic engineers can speed up the process.
Bottom line is that brains are just like computers only they evolved to emphasize different
qualities because they evolved in different environments.
don
Let us say that I have a spreadsheet in which there are a thousand cells. You may quite
logically tell me that it would be most efficient if used a thousand processors to update my
spreadsheet. Seems logical that if I use one CPU per cell, that the spreadsheet might update
quite quickly. I may totally agree with you in theory. After I have agreed with you, I will still
use Excel for my spreadsheet and I will use only one CPU to simulate your thousand CPUs.
You may be correct in a structural way. If you are, then I will use my one CPU to simulate
your thousand CPUs operating in parallel. Then I will stop and ask, “Which is the more cost
effective implementation of the solution to my problem, 1000 CPUs or 1 CPU?”
What if you asked me to solve the same problem using rat brain cells in a Petrie dish?
http://www.sciencedaily.com/releases/2004/10/041022104658.htm
If that was the problem, then I would probably say to myself, “These brain cells are so slow that
I had better implement my solution using one thousand parallel mini-brains, one for each
spreadsheet cell.
Conclusion:
We should divide the problem into two parts. Part one is “what is the most efficient structure
of a brain?” Part two is “How should we implement our design?”
I have 1 CPU and it is fast enough to simulate the 10 thousand neurons in a flee's brain, then I
may well be fully capable of simulating a flee's brain with a single CPU. The structure may be
best implemented in cells using 10,000 neurons but may be totally simulated as an imitation
of ten thousand cells operating in parallel.
Second fallacy in the thinking:
If I have a PC and replace it with another PC whose CPU is a "Core Duo", is my PC still a PC?
This is an important question because I intend to use mathematical induction to prove my
point.
http://en.wikipedia.org/wiki/Mathematical_induction
To prove mathematical induction, I must assert that adding one more processor to my PC
doesn't stop it from still being defined as a PC. Since this is true (absolutely nobody is
suggesting that if a PC has two processors, it is “no longer a PC") By induction, I may assume
that PCs containing 8 processors will still be referred to as PCs. The Sony Play Station 3 has 8
processors in the CPU package. If there are 8 processors in a PC by the end of 2007, nobody
will be upset or terribly surprised.
If this trend continues the number of processors may well go up by a factor of 4 each year.
This would mean that in 10 years, I may be able to purchase a PC with 1 million processors. (If
it only doubles each year, then it will take 20 years to have 1 million processors) If each
processor operates at 200 billion instructions per second, then each processor would be about
a billion times faster than a neuron operating at 200 pulses per second. I think that it is safe
to say that a computer instruction, such as add two 64 bit numbers together is objectively
"faster" than a nerve which is simply squirting some seratonin at another four dozen nerves.
This would mean that my PC in ten years could simulate at least 1 million processors times 1
billion times the speed of a neuron. This multiplies out to the equivalent of 1 trillion neurons
being simulated real time. If the human brain has 100 billion neurons, then it will be
possible for a PC to simulate ten human brains by simulating each neuron in one of its 1
million processors.
Suggestion:
Stop complaining and start designing the best way to interconnect the million CPUs and start
writing the code to simulate the neurons in the human brain.
If my job is to simulate an insect with a thousand brain cells, I would very likely use one CPU
to simulate the action of a thousand cells acting as a parallel processing insect brain. The
insect brain would be similar to the spreadsheet in that regard. Since my one CPU goes fast
enough to simulate a thousand slow insect brain cells, why would I implement the solution
any other way?
Equivalence is an important word in this discussion.
Let us say that it is 1972 and the 4004 microprocessor chip set has recently been invented by
Intel and we have 1000 of these 4004 chip sets in front of us. Here we are at the forefront of
microprocessor technology and we decide to make a machine which automatically
recalculates spreadsheets for people. The 4004 is so slow that we figure that we will support
spreadsheets with up to one thousand cells and use one processor for each cell. Our logic is
sound because we know that parallel processing is OBVIOUSLY the best way to calculate a
spreadsheet in which each cells contents depends on the contents of the other cells. It sure
pays to be smart and recognize that parallel jobs should be implemented with parallel
hardware. Once we have the spreadsheet computer working we start selling it for $20,000 per
copy.
The arguments being used regarding human brain simulation could be used to argue that:
There will never be a single CPU computer which will be capable of doing spreadsheets,
because the structure of a single CPU machine is “all wrong” for the problem being solved.
We figure that large corporations might buy them. Our If I have a single CPU and it can
simulate a thousand older CPUs, then it may well perform the job that a thousand old CPUs
did. Why would I use a thousand old CPUs in parallel when I could use one new CPU?
Let us say that I have a spreadsheet in which there are a thousand cells. You may quite
logically tell me that it would be most efficient if used a thousand processors to update my
spreadsheet. You may be correct in a structural way, but I may simply tell you that I would
rather use one processor to simulate the action of the thousand processors that you
recommend that I use. If my simulation cost less than your parallel implementation, then in
what way was your suggestion "more efficient"?
If my job is to simulate an insect with a thousand brain cells, I would very likely use one CPU
to simulate the action of a thousand cells acting as a parallel processing insect brain. The
insect brain would be similar to the spreadsheet in that regard. Since my one CPU goes fast
enough to simulate a thousand slow insect brain cells, why would I implement the solution
any other way?
Structure being simulated: parallel spreasheet or parallel insect brain.
Implementation: Today it will be serial simply to reduce cost.
Final submission to http://dangerousintersection.org/?p=178
Let us say that it is 1972 and the 4004 microprocessor chip set has recently been invented by
Intel and we have 1000 of these 4004 chip sets in front of us. Here we are at the forefront of
microprocessor technology and we decide to make a machine which automatically
recalculates spreadsheets for people. The 4004 is slow but we figure that we will support
spreadsheets with up to one thousand cells and use one processor for each cell. Our logic is
sound because we know that parallel processing is OBVIOUSLY the best way to calculate a
spreadsheet in which each cell's contents depends on the contents of the other cells. It sure
pays to be smart and recognize that parallel jobs should be implemented with parallel
hardware. Once we have the spreadsheet computer working we start selling it for $120,000
per copy. We obtain a patent on our technique of using parallel processors to solve
spreadsheet problems. We mortgage our homes and put all of the money into our company
because we absolutely know that we have the very best way to calculate spreadsheets, and it is
patented. We have a lock on the market because we KNOW that parallel jobs are most
efficiently implemented with parallel hardware. If venture capitalists doubt our plan, we tell
them to talk to Gerald M. Edelman, who knows the difference between jobs which are
obviously serial and jobs which are obviously parallel.
The arguments being used regarding human brain simulation with a PC could be used to
argue that:
There will never be a PC computer which will be capable of doing spreadsheets, because the
structure of a PC is “all wrong” for the problem being solved.
Anybody want to buy my parallel processor spreadsheet computer? Marked down to $10,000,
which is a bargain considering that it has 1000 processors and weighs a ton.
don
In the original argument above, the following was stated:
"Paul Churchland also notes that we are horrible at logic and other types of systematic
thinking. How many years do we study math, but look how we still struggle as adults!” The
Engine of Reason, the Seat of the Soul, Paul M. Churchland (1995)."
This is precisely what we are dealing with here. We have people who cannot see through the
bad logic and bad systematic thinking shown in the original argument which claims that
human brains are not like computers. Humans fall for arguments like "humans are better
because our minds are more robust" simply because, as Paul Churchland notes, "we are
horrible at logic and other types of systematic thinking."
Even the person who notes that humans are bad at logic goes on to say, “If the brain were a
computer, this would not be the case.” Here, Paul Churchland is demonstrating that he is a
human and how difficult it is for humans to think logically and systematically. What he
should have said is "If the brain had evolved in an environment where logic was more
important and "hunting, gathering, eating and reproducing" were less important, then
humans would be as logical or more logical than computers are. Then he could have noted
that computers evolved simply because once humans had their eating and reproducing needs
satisfied, they were ready to move on to other tasks requiring more mathematics, logic and
systematic thinking. It is in this latter environment that computers evolved. Life is a solution
to an opportunity in an environment.
So, when you want to decide the truth of any complex logical argument, do you believe the
horrible-at-logic humans or should you ask your computer?
The human brain is simply a large, slow parallel processing computer.
Get used to it.
From above: "Paul Churchland also notes that we are horrible at logic and other types of
systematic thinking. How many years do we study math, but look how we still struggle as
adults! If the brain were a computer, this would not be the case. The Engine of Reason, the
Seat of the Soul, Paul M. Churchland (1995)."
Note the recursive and reflexive qualities of this argument. Humans cannot think logically
and systematically so here is an illogical and unsystematic argument which humans will
believe simply because they aren't very logical and systematic.
The logical and systematic truth is that life evolves in a particular environment and must
evolve those qualities which allow it to survive in that specific environment. In a jungle, an
organic biped omnivore must evolve abilities of agility, dexterity and thinking for the purposes
of hunting, gathering, eating and reproducing.
On a desktop, a computer finds itself in a totally different environment. The eating and
reproducing are all taken care of by electrical outlets and computer factories. In that
environment, a logical and systematic entity would notice that computers should evolve those
different qualities which are lacking yet needed in its environment. Those specific qualities
are logic and systematic thinking.
Now that human hunting has been converted to gathering by the McDonald's drive-thru,
humans have a need for less hunting and more logic and systematic thinking. The problem is
that organic entities evolve so slowly. So, the need for logic and math is being fulfilled by
computers.
Nevertheless, that doesn’t mean that humans, given enough time to evolve, couldn’t
eventually have brains which could be as logical and as systematic as computers. Maybe
genetic engineers can speed up the process.
Bottom line is that brains are just like computers only they evolved to emphasize different
qualities because they evolved in different environments.
don
 | |  | |  | |  |