16 replies to this topic

[Just One Question]

Hi all,

As I remember from reading about this very interesting subject the Blue Brain simulation is working VERY SLOW, about 10% of the real time speed, this is very strange becouse all the time Ray kurzweil argue that a simulation of a brain on a computer will work AT LEAST a million times FASTER than a biological brain!

How can you explain this contradiction ?

Thanks for any explanation!

Edited by Just One Question, 01 September 2009 - 05:21 PM.

[forever freedom]

It's working at 10% (if this number is correct) of the speed of a mice's brain. This is how it starts. Then, as computer power constantly increases, we'll get faster and faster simulations. The simulation of a brain in a computer will work millions of time faster, in the future.

[Vgamer1]

Yea, forever freedom is right. Kurzweil talks about once we actually simulate a human brain we will be able to "speed it up." That point hasn't come yet, but it's what projects like what you're talking about are striving towards.

[Just One Question]

Not at all, you are both wrong, Ray is very clearly talking about TODAY's computers that are working a million times faster than our biological electrochemical brain, from his book "The Singularity Is Near" -

"Our thinking is extremely slow: the basic neural transactions are several million times slower than contemporary electronic circuits...."

"Our electronic circuits are already more than one million times faster than a neuron's electrochemical processes, and this speed is continuing to accelerate...."

He is very clearly talking about TODAY's computers, they still only have the power to simulate a half rat brain (one "column", the blue brain project) but if listening to Ray kurzweil this simulation Should ALREADY work a million times faster than the speed of a real (half) rat brain.

How can you explain this contradiction ?

How can he be wrong about something so basic ?

Thanks.

Edited by Just One Question, 01 September 2009 - 06:35 PM.

[Vgamer1]

OK, yes a computer can run computations millions of times faster than humans today. But just because a computer can do math really fast, doesn't mean it knows how to think yet.

This powerful computing is being used to simulate the brain, which requires billions of calculations per second. That's why it runs slower than the actual brain.

We're not actually conscious of those billions of calculations going on in our heads. What Kurzweil is referring to when he says human thinking is "slow" is conscious thought, not the countless operations that go on in our heads without our awareness.

Does that make sense?

[Just One Question]

No Vgamer1, it doesn't make sense to me, becouse if I read the words of Ray Kurzweil - As It Is, I understand that if you build/simulate a piece of a brain on today's computers it will ALREADY work a million times faster than the same piece of brain in a biological brain.

No, it doesn't make sense.

Edited by Just One Question, 01 September 2009 - 06:49 PM.

[niner]

"Our thinking is extremely slow: the basic neural transactions are several million times slower than contemporary electronic circuits...."

"Our electronic circuits are already more than one million times faster than a neuron's electrochemical processes, and this speed is continuing to accelerate...."

Ray's not wrong, he's just not being clear. The problem is that he's comparing the speed of a fundamental neural operation to a fundamental computer operation. We have a trillion neurons all doing their fundamental operations in parallel, more or less, while the vastly faster computer has to do tons of operations serially in order to simulate the behavior of a single neuron. That's the main reason why Blue Brain is so slow.

[Just One Question]

niner, it looks to me like a fraud, if what you say is true, then why does he at all talking about it? why should I be care if computers today works a million times faster than a biological brain, if ACTUALLY they work much slower?

It's like someone coming to me and telling me: Hey, this car is fantastic! it's engine works a million times faster than a normal engine! but then I take the car for a ride and to my surprise I can see that it's driving Much Slower than a normal car, so I go back to him and I ask him about it, and he says: Yes, OF COURSE, this engine works much faster than a normal car's engine, BUT it's connecting to a MUCH SMALLER transmission wheel, so the car is driving much slower.....

Then why at all did you bother telling me about this fast engine?? if it actually doesn't help the car going faster?

I don't like this, maybe also when Ray is talking about a singularity in about 35 years from now he ACTUALLY mean that it will happen in about 500 years from now.

Come on this is stupid.

Edited by Just One Question, 01 September 2009 - 07:54 PM.

[niner]

Come on this is stupid.

Right. I don't think anyone here is going to be able to help you. Maybe you should take some classes and learn how computers work, and then it will become more clear.

[Just One Question]

OK, someone else ?

[Vgamer1]

OK, think of it this way....

A human mind consciously does calculations very slowly. If you had a human and a computer compete for who could do 1,000 algebra problems first, the computer would win. In that way computers are millions of times faster than humans.

But now think about all of the subconscious activity your brain undergoes. We are constantly processing information from our senses, which interacts with our complex memory system. This requires the activity of billions of neurons with trillions of connections. It gets even more complicated than that because sadly our brains do not work like computer chips. We have to model our biological system with a computer program. So, if we want to simulate one neuron firing (which is the smallest unit of calculation in a human brain), we need more than just 1 bit (the smallest unit of calculation in a computer).

As computers get faster and our models improve, we will have better, faster brain simulations. We don't even really know completely how a single neuron works. We're approximating brains, we don't actually have a real brain simulated, just pieces of one.

[forever freedom]

OK, someone else ?

Electronic circuits are at least a million times faster than neurons, yes. But this doesn't mean anything. It doesn't mean that computers are a million times faster than the human brain. We will need computers to first simulate human neurons and then try to simulate an entire human brain. It doesn't work at all the way you're thinking. To simulate all the neurons in the human brain we will need computers vastly more powerful than the human brain, because computers are not made like human brains, they don't "think"/work like human brains. Cappiche?

By the way, i'm just saying what niner and vgamer1 already said.. but hopefully in more simple terms. If you still don't understand i suggest you do what niner said and learn how computers work.

Edited by forever freedom, 01 September 2009 - 09:26 PM.

[valkyrie_ice]

Actually let me.

To understand why the simulation is working slower you have to understand that in order to duplicate the action of one single neuron, a computer must do approximately a million or more steps.

It has to calculate where every connection is in relation to every other neuron
It has to calculate the cells state at any given second
It has to calculate the chemical balances at any given second
It has to calculate the electrical potential of every synapse at any given second.
It has to calculate the effect of every impulse affecting the neuron at any given second.

Each one of these processes required hundreds of thousands of calculations, which the computer must do one right after the other. And that same processor must do this for hundreds of thousands of cells one at a time.

In order for a computer to run a simulation of a brain using current processing techniques, a computer must run millions of times faster than the brain it is simulating. This is inherent in both the type of processors we must use, and the software which runs it.

Massively parallel networks help, but since each processor is still having to run a certain number of neurons it still cannot match the quantum processes of the neuron which can nearly instantly determine it's state.

However, Quantum computers are in the works, and I expect they will provide solutions to such issues by the mid to end of next decade.

Kurzwiel is correct that modern computers exceed the speed of our minds, but this is solely in terms of speed of a single calculation matched to a single neuron firing. What you are failing to grasp is the enormous amount of extra calculations required by a computer to imitate the functions of the brain.

[niner]

Good explanation, Val.

[Just One Question]

Thanks for the extra explenations, this point is clear, I'm just saying that Ray Kurzweil is very misleading in what he say, when he talks about today's computers a million times faster than a biological brain, he gives the impression that if you have a simulate of a small piece of brain on a computer it should work a million times faster than the biological piece of brain that it's simulating.

But as we see it's not working like that at all, so I'm asking why did Ray Kurzweil bother to emphasize this point so many times in his book and in his lectures? in what way does it help us? OK it's a nice piece of information, but it doesn't gives us reference to anything, we can't calculate anything from it, it's just misleading, I was misleading by it.

He should be more clear in his words.

Edited by Just One Question, 02 September 2009 - 03:58 PM.

[ben951]

The blue brain run slower simply because it doesn't have enough power to run a mouse brain yet, so they try to compensate by running the program slower.

It's like if you want to play Crisis with an old Pentium, the only way for the computer to keep up is to slow down the frame.

When we'll have the necessary computing power to run a mouse brain it'll run faster than a real mouse brain.

Edited by ben951, 02 September 2009 - 05:09 PM.

[Vgamer1]

I'll take your example quotes from Ray. Try taking what he's saying at face value instead of reading into it.

"Our thinking is extremely slow: the basic neural transactions are several million times slower than contemporary electronic circuits...."

It's true - neuron-to-neuron communication is very slow compared to the speed between components of a computer. In this quote he doens't say anything about a computer brain simulation being faster than a real brain. You are thinking that is implied when it is not.

"Our electronic circuits are already more than one million times faster than a neuron's electrochemical processes, and this speed is continuing to accelerate...."

Again, a computer can do computations very quickly compared to a brain, but in order to simulate the way a brain computes, the program must do many many many "computer" calculations in order to simulate a single "brain" calculation. The more efficient we make our software and the better we make our hardware, the faster this process will be. In your quote it doesn't say that simulated brains run faster than real brains, it says that circuits transmit information faster than neurons. Those are very different things.