Brain Power

[Cris]

The human brain contains about 10^11 neurons. Each neuron has about 5*10^3 synapses, and signals are transmitted along these synapses at an average frequency of about 10^2Hz. Each signal contains, say, 5 bits. This equals 10^17 ops.

Neurons = 100,000,000,000
Synapses per neuron = 5000
Signal transmission frequency = 100Hz
Signal size = 5 bits
Total analogous computer clock speed = 100,000,000 GHz

So we had 1GHz chips in 2000, we are up to around 1.7GHz in the first half of 2001, so we should reach 2GHz by end of 2001. This doubling effect known as Moore’s law has held up very well since the late 1940s. It shows that computer power doubles approximately every 18 months but recently has been increasing to every 12 months. If we assume this trend continues then the table below shows the number of years it will take before a computer has the same operational speed of the human brain.

Or to put it in a different perspective: If your current desktop PC or Laptop currently runs at 1GHz, then 100 million of those machines combined into one machine will be the equivalent power of a human brain. And we should achieve that by 2030.

Year, Clock Speed in GHz
2000 1
2001 2
2002 4
2003 8
2004 16
2005 32
2006 64
2007 128
2008 256
2009 512
2010 1024
2011 2048
2012 4096
2013 8192
2014 16384
2015 32768
2016 65536
2017 131072
2018 262144
2019 524288
2020 1048576
2021 2097152
2022 4194304
2023 8388608
2024 16777216
2025 33554432
2026 67108864
2027 134217728

Neat huh? Or is that simply mind-boggling?

Cris

 

[KneD]

oh yeah, it certainly is really neat...cool!

But I think the doubling effect will slow down, so it will take longer then you concluded.
But we never in know in our recent technology.

 

[Cris]

Kned,

I agree, in this case seeing is believing. But I don’t think the lack of power in an individual chip will change the end result – equivalent human power by 2030. The brain is essentially a massively parallel system. So even if we only achieve 1/100th of the ultimate chip power I stated then coordinating the results of 100 chips to form a brain should be quite achievable. Heck, my company has been building parallel systems with over 100 processors for the past 30 years, and we have systems in the field with over 1000 processors.

There are some technical barriers to continually increasing clock speed while making the chips smaller. Even Intel with their Itanium family has introduced significant parallelism and the clock speeds are not as high as the Pentium IV, but the effective processing power is much higher.

But consider what happens beyond 2030, assuming that we have achieved equivalent human brain power. Do we then say, “that’s it, we’ve done enough?” I don’t think so. The implication in my table is that the following year we would have double brain power, etc etc. And then we are simply in a realm that is unimaginable.

The future will certainly be very interesting.

Cris

 

[AB_saratov]

When I was searching for something interesting in web about AI, I first found archives of this forum and viewed few topics “from the beginning”.
I think this one of them should be picked up.

First of all there is no 32GHz computers now... very pity. And there will be no 64Ghz in 2006.

But at the second - in my opinion – the math calculating up there is for the size of memory database of “brain” but not for processor speed, computer power etc.

 

[Light]

When I was searching for something interesting in web about AI, I first found archives of this forum and viewed few topics “from the beginning”.
I think this one of them should be picked up.

First of all there is no 32GHz computers now... very pity. And there will be no 64Ghz in 2006.

But at the second - in my opinion – the math calculating up there is for the size of memory database of “brain” but not for processor speed, computer power etc.

Agreed. Since no one can even estimate what percentage of the brain is used exclusively for storage and how much is dedicated to processing (and there appears to be a major overlap between the two), the numbers presented here are rather meaningless.

 

[JoeTheMan]

Exponential expansion is a humbling thing.

You know our parents and technology? Think how much more in the dust even the tech-savvy of us are gonna be in fifty or sixty years.

But even if we make a computer which is strictly as powerful as the human brain, we haven't even begun to touch how to make a computer have a conscious experience. If our chips are fast enough, and processing in parallel, we've got the ingredients for an incredibly powerful imitation of a conscious human. Would that be the same as duplicating self-awareness? --or at that point, does the fact that the 'hardware' is different (brains and computers) matter any longer, since both are 'running' a mind?

 

[Light]

Exponential expansion is a humbling thing.

You know our parents and technology? Think how much more in the dust even the tech-savvy of us are gonna be in fifty or sixty years.

But even if we make a computer which is strictly as powerful as the human brain, we haven't even begun to touch how to make a computer have a conscious experience. If our chips are fast enough, and processing in parallel, we've got the ingredients for an incredibly powerful imitation of a conscious human. Would that be the same as duplicating self-awareness? --or at that point, does the fact that the 'hardware' is different (brains and computers) matter any longer, since both are 'running' a mind?

It will still be vastly different. Two things a conputer cannot have are original thoughts and emotions. It is primarily emotions that makes us human.

 

[Cottontop3000]

What is the fastest computer processor created to date? Say one of the government super-computers. Not the ones we have access to in retail.

Also, in 25 years, who knows what else we will have come up with. Maybe we will be able to invent a new type of consciousness. One can only hope and then, perhaps, fear.

 

[AntonK]

It will still be vastly different. Two things a conputer cannot have are original thoughts and emotions. It is primarily emotions that makes us human.

Thats a pretty broad and assertive comment to make without any sort of backup? What makes humans so special that we couldn't duplicate the processes of the brain on a different type of hardware?

-AntonK

 

[Light]

Thats a pretty broad and assertive comment to make without any sort of backup? What makes humans so special that we couldn't duplicate the processes of the brain on a different type of hardware?

-AntonK

No backup needed. Humans are special compared to a machine for the two exact reasons I stated. How about you telling us exactly how you would go about programming true emotions into a computer? And how would you program it to give it the ability to have original thought? It's simply a machine and always will be - it can only know facts you have already told it and that it's allowed to search and find. You cannot teach it to connect unrelated facts as a human does when one of us has a moment of inspiration. If you think so, please tell me and all the AI scientists how you would go about it.

 

[AntonK]

Thats really not a great argument. I have no idea how-to build a computer to mimic emotions. My best guess is that emotions are the filtering of primitive instincts (lower brain) through the consciousness, logic and understanding of our higher brain. I see no reason why this can't be the same for a machine.

The reason I say your argument isn't great is because just because I can't design a computer in no way says it can't be done. What I can show you is that what the human brain is made out of (the hardware) is simply neurons. We can analyze neurons and see exactly what they do. Its not really complicated. They're basically ion transfer cells. We can model what individual neurons do with some clever math. Just take a look at the field of neural networks. Now how does the a collection of these simple machines (neurons) turn into the intelligence that is the human brain? Its all in the massive complexity and in the specific configuration. There are animals with brain sizes comparable to ours yet they do not show the type of intelligence we do. Therefore the answer is obviously in the configuration of these neurons. We can build silicon circuits which can mimic the actions of neurons and implement functions actually more complex. There is no reason then to believe that a certain configuration and complexity of these silicon circuits would not result in intelligence just like humans.

Image for a moment that we took a single cell from your brain and analyzed it, we completely understood its function precisely. We then built a circuit which could exactly mimic that single neuron. So far its really not that far fetched an idea. Then imagine we were able to insert that circuit back where your neuron was. Your brain would function EXACTLY the same. Now why couldn't we simply do this for 2 neurons? 4 neurons? 100 neurons? A million neurons? a billion neurons? The fact is there is nothing magical about our brains. Its vastly complex and very special in that no other species or thing that we know of has the complexity and configuration of ours. But in the end, its just hardware.

-AntonK

 

[funkstar]

No backup needed. Humans are special compared to a machine for the two exact reasons I stated.

What makes you think that those two things are

1) real, and
2) impossible to duplicate on a machine?

What makes you think you're not a machine yourself? I've yet to hear of a single bodily function that must be forever inexplicable by mechanical reasons. Quite the contrary, in fact - the more layers we peel away, the more it seems that we're simply machines ourselves. Complex and extraordinary, but not magical, as you seem to imply.

It's simply a machine and always will be - it can only know facts you have already told it and that it's allowed to search and find. You cannot teach it to connect unrelated facts as a human does when one of us has a moment of inspiration.

You need to qualify this. Merely stating it is not an argument.

 

[JoeTheMan]

Just because we don't see a computer which is incontrovertibly conscious, and just because we can't yet conceive of the proper series of program instructions which could instantiate a consciousness on a computer, *doesn't* by any means demonstrate that such a condition (a computer having thoughts, experience, concsciousness, etc.) is impossible.

"What I can show you is that what the human brain is made out of (the hardware) is simply neurons."
We had also better be careful of equating bits and neurons too hastily. We're dealing with entirely different substrates here. In particular, it's not really clear how consciousness emerges as a property of certain kinds of matter. This is a philosophical problem of interaction. For a mind has to have certain properties to qualify as a mind (and not just a brain or a computer, for instance):

1) Non-localized. That is, even though we can say my thoughts are *in* my brain in some sense, it remains impossible to point at the part of the brain that creates a conscious experience. The 'mind' doesn't occupy any actual space like the brain does.

2) Non-deterministic. Here's the big problem for AI, since computers are completely deterministic and minds are not. Your thoughts are free from causual interaction in a way that a computers operations can never be; while it's already a stretch to say computers 'think,' 'know' or 'remember,' it is even more difficult to think of a computer having a 'choice'.

 

[Light]

Thats really not a great argument. I have no idea how-to build a computer to mimic emotions. My best guess is that emotions are the filtering of primitive instincts (lower brain) through the consciousness, logic and understanding of our higher brain. I see no reason why this can't be the same for a machine.

The reason I say your argument isn't great is because just because I can't design a computer in no way says it can't be done. What I can show you is that what the human brain is made out of (the hardware) is simply neurons. We can analyze neurons and see exactly what they do. Its not really complicated. They're basically ion transfer cells. We can model what individual neurons do with some clever math. Just take a look at the field of neural networks. Now how does the a collection of these simple machines (neurons) turn into the intelligence that is the human brain? Its all in the massive complexity and in the specific configuration. There are animals with brain sizes comparable to ours yet they do not show the type of intelligence we do. Therefore the answer is obviously in the configuration of these neurons. We can build silicon circuits which can mimic the actions of neurons and implement functions actually more complex. There is no reason then to believe that a certain configuration and complexity of these silicon circuits would not result in intelligence just like humans.

Image for a moment that we took a single cell from your brain and analyzed it, we completely understood its function precisely. We then built a circuit which could exactly mimic that single neuron. So far its really not that far fetched an idea. Then imagine we were able to insert that circuit back where your neuron was. Your brain would function EXACTLY the same. Now why couldn't we simply do this for 2 neurons? 4 neurons? 100 neurons? A million neurons? a billion neurons? The fact is there is nothing magical about our brains. Its vastly complex and very special in that no other species or thing that we know of has the complexity and configuration of ours. But in the end, its just hardware.

-AntonK

Just hardware, eh? Wow! That's got to go down in the books as the understatement of the century!

Keep in mind that even though I am a scientist by profession (biology, psychology) I still have to say that your expectations for science and technology are unreasonable.

And while you expect great complexity, you still tend toward over-simplifying the problem. While it may someday be possible to precisely mimic the operation of a single neuron and then replicate it as many times as we choose, that still cannot function as a human brain. To use your phrase - THAT would still be just hardware.

It might be well for you yo remember the old maxim "the whole is greater than the sum of it's parts" and that applies perfectly when speaking of the brain.

A computer is a computer and is just a computer. Regardless of it's size, computing power and how much data it can store, it will still be just a computer - a number-cruncher and data analyzer. You cannot program it to develop a consciousness, to feel true emotions and many other human attributes. The brain and it's mind are much more complex than any inorganic mechanism. The whole really IS greater than the sum of all it's parts.

 

[AB_saratov]

The beginning of the forum is very interesting.
But after it going every day discussions “possible or not” like yours. Stupid or not.

2 Light
Of course without psychology and biology it just a computer and a hardware. And i.m.h.o. when it will be created a half of scientific work on it will be psychology work.


ps: sorry for my eng

pps: someone here already have a parry possible or not ?

 

[Light]

The beginning of the forum is very interesting.
But after it going every day discussions “possible or not” like yours. Stupid or not.

2 Light
Of course without psychology and biology it just a computer and a hardware. And i.m.h.o. when it will be created a half of scientific work on it will be psychology work.

Interesting, but it immediately raises the question: how do you do "psychology work" on a computer? Easy enough to say... but how do you actually do that?

 

[Cottontop3000]

Interesting, but it immediately raises the question: how do you do "psychology work" on a computer? Easy enough to say... but how do you actually do that?

Maybe someday we'll have to figure that out. MRI, CPU-Scan?

 

[AB_saratov]

I ? I just interesting in Psychology, etology, programming, and data sequence. And now searching for latest news about MI creating.

If some one know how to do this, he would be a little richer than Bill-gey tssss...

 

[funkstar]

Keep in mind that even though I am a scientist by profession (biology, psychology) I still have to say that your expectations for science and technology are unreasonable.

Why?

And while you expect great complexity, you still tend toward over-simplifying the problem. While it may someday be possible to precisely mimic the operation of a single neuron and then replicate it as many times as we choose, that still cannot function as a human brain.

Why not?

A computer is a computer and is just a computer. [...] You cannot program it to develop a consciousness, to feel true emotions and many other human attributes.

Why not?

The brain and it's mind are much more complex than any inorganic mechanism.

True, but it doesn't mean that it transcends mechanics.

The whole really IS greater than the sum of all it's parts.

You don't know that. You assume it. I don't find it to be a reasonable assumption, without an argument.

 

[Blue_UK]

So, Light, do you suppose that there is something other than 'just neurons'? Anything supernatural? If we assume there is no supernatural element then there is no reason to assume that logical construction of the brain cannot be duplicated by another medium.

For example, it does not matter to you that the atoms which make up your brain change (in part) from day to day. Why should it matter that nerve impulses are controlled by voltage gated ion channels and not MOSFET switches?

It's not the material that counts, it's the what the system does - and you cannot possibly state that 'emotions' and 'feelings' are anything more than that if you disregard the supernatural.

Current AI is a (as you would probably agree) blind, mechanical 'dumb' computer trick that looks as if it's clever. However, that is the so-called top down approach. to build AI from artifical neurons (...bottom up) would create something that works in the same way as a brain.