We are on the threshold...

[iMac722]

I would imagine that a few years from now computers will be everywhere and the entire world's population will have them because they will be so cheap by then. Then, by the year 2015 the first computer that is the equivilent of the human brain will appear( assuming that computer power doubles every eighteen months) it will have the same processing power and the same memory capacity, in about ten years we have that power on our desktops and in about 5 years we have that pocket sized. By then we will have friends that we can talk to like any individual human, or a Matrix scenario when those computers take over the world. Around the time when that first computer pops up our technology will skyrocket we will be much smarter, stronger, and be able to do anything we want pretty much. We will be able to travel between the stars colonize new worlds and become the Homo Superior like in the game, we can not avoid this it is our destiny(providing that we dont kill ourselves first) and eventually transend.

[Lord Maxwell]

Aren't we the little optimist. Computer power will likely not continue to double every 18 months, as consumer interest is actually going down. (Well, it is generally high, but people more often buy 6 months old products today than two years ago. Maybe due to less "techiness" among consumers.)

Besides, a mere increase in computer power will not simulate a brain. As long as they are based on integer operations in linear space they cannot do it efficiently. (We have 200 billion (or 1/10, I never get these words right in English. Or 10 times as many. Well, look it up, it's 200*alot! (Where, oh where is my dictionary?) "nodes" in our brain, and each connects to approx 5000 others. These connections were "well understood" until a few years ago, when people figured out that there was several powers of ten more things going on in each connection than previously believed. (They exchange large-scale electrochemical signals, as well as a host of smaller scale signals.)
There is also a likelihood that we will need to base them on organic matter to replicate the behaviour, but then we loose the so valued predictability with computers.
And there is certain limits that are approaching. (We need a better understanding of the randomness of the quantum world pretty soon, and that field isn't looking to bright. We need room-temperature superconductors. etc etc. We will get there. But in the 80s I heard that we would have colonized the moon by now. Still waiting to go skiing on the sea of tranquility. Leave the predictions to Nostradamus. (Besides, with history as a guide, we will likely think of something new and interesting and more or less drop development. We have the collective attention span of a 7-year old ODing on amfetamine.

[Obi Wan Kenobi]

I'm pretty sure there are 200 billion neurons in our brain, if that's what you mean...

[Vi Vicdi]

I am quite certain that computer power will easily surpass that of the human brain if it already hasn't.

The problem ain't the hardware; it's the software.

"It ain't the crate -- it's the pilot who flies it."
-- "Red Baron" Manfred von Richtofen

[Qilue]

Plus, the human brain can update the the state of individual neurons with no effect on other parts of the brain.

Lastly, apart from some corruption by outside influences, there are no bugs in the brain's programming, at least none that are known.
[This message has been edited by Mouse (edited January 30, 2000).]

[Urban Ranger]

Unfortunately, there are. Known in some other circles as "genetic defects."

------------------
If I can't believe in my own eyes, whose eyes can I believe? Yours?!

[Lord Maxwell]

VV: Nope, it isn't even within a factor of a million in connection power. Computers are still way behind, they are to linear to be able to simulate true intelligence. (Off course, there has been some promising research on fuzzy chips.)
But sure, it counts the decimals of pi faster than you. (It's a hassle to simulate a neural net using software only, all the floating point calculations mess it up. You need to make 200billion*5000 floating point operations per second, merely to update the state of the synapses. )

[Qilue]

True, but our brains don't need rebooting after they generate General protection faults.

[edgecrusher]

yes, but what about seizures? i work with people who have hardcore medications for their seizures... beasties with side effects that we can't discern because they're taking so many other meds. besides, there is a chemical reboot, known as psychotropic drugs. sounds fun, right, but one of the side effects could cause you to walk around with your arm sticking 45 degrees in the air four your entire waking life, without a damn thing you can cousciously do about it.
[This message has been edited by edgecrusher (edited January 31, 2000).]

[Qilue]

Ok, I guess I shouldn't bother posting on this thread any more, since few here understand humour and are more interested in being intellectually superior.

[Bkeela]

iMac722, I think you've been playing SMAC a bit too much...

Bkeela.

[Lord Maxwell]

Ok, some more info on the brain. It consists of 200 billion neurons. Each is made up of the cell body proper, an axon and several dendrites. (Up to 10000+ in extreme cases.)
The dendrites are the incoming ports, and the axon the outgoing port. (It branches near the end to interface with several dendrites belonging to other neurons.)

It takes an axon 1/1000 second to flash. It does so by sending a jolt of about 55 millivolts down it's length. It seems to create this current by some smart gates, that only let certain types of ions through. By stocastic processes it thus creates the needed voltage to signal in less than 1/1000 second. Luckily for us, it is binary in it's nature, it flashes, or it don't.
To avoid the loss of voltage over distance it is sheated in a fatty tissue, known as the myelin sheath. This sheath is broken up at intervals (it consists of short tubes) and these intervals are called ranvier nodes. They play an important role in keeping the speed up.
The signal that travels down these axons consists of the ion gates opening, and rapidly closing again.
Teh signal is sent out from the cell body when it is stimulated enough by the data it receives from it's dendrites.
It regulates signal "strength" not through the voltage level, but through the interval between firing. The more it wants to make itself heard, the more often it fires.

So, we have 200billion neurons (by conservative measurements, Nauta and Feirtag, 1986, placed it at 1000 billion neurons), which can accept about 1000 values each, at intervals of 1 millisecond. Weigh the income and decide whether or not to send out a signal. Everything except the weighing is simple binary switching. So, we need to make 2*10^12 binary adjustments per second (this is the maximum of what the brain can do), each binary adjustment is the result of a poll of about 1000 other adjustments. So we need to poll 2*10^15 times per second.
Now, your average supercomputer could do 8*10^12 polls per second. (Given 8192 processors with 1Ghz speed each, making one poll per clock tick, which is unreasonably fast.) You see that we lag by a factor of 100000 even under ideal conditions.

Now consider the fact that the brain never "blocks" on an I/O. (A state that is often encountered in a computer, and serious part of os programming is made to avoid it.)

We need a new model for mass-scale calculations, as well some serious increase in hardware.

Now, if the cutting edge research is correct there is a chance that the brain does a few million times as many operations per second. (Circumstantial evidence point to the brain using quantum processes in the gap between axons and dendrites as well. This is not proven yet, but if this is the case we suddenly have a much more complex situation.)

Can a computer be built to model this. Sure. How do we train it? Nobody knows? How do we pre-weigh the neural net? Nobody knows? And experimenting in the field is damn expensive, as you need computers about as good as those at fort meade. And I fear the nsa has better uses for it's computing power.

Off course, we could get rid of all the muscular neurons, and most of the sensory neurons. The vast majority are in the gray area between, where the "real action" seems to be. These are called interneurons. We can't get rid of these just yet.

So, hands down boys, we are only a few steps closer to this goal today than we were in the seventies.

And even when we do, how do we know if androids dream of electric sheep? (ie, there are quite a few philosophical grey areas here.) As of yet no computer has been able to pass even the simple Turing test, much less proven "true" intelligence according to more exact measurements.

[Thaum]

Very interesting thread. Tell me Lord Maxwell, where are you getting this information as I think it is fascinating and like to read up on it.

------------------
For it is the folly of men that they forget!

Excalibur

[Bblue]

Blade Runner fan Lord Maxwell?

[edgecrusher]

sorry if i came off like an ass. it's just that with my line of work, nobody seems to understand these things. people seem to think nowadays that prozac is cool, but it, like all meds are a last resort. people whose brains have tripped the 'general protection fault' as it were don't come back. it can't be fixed, or rebooted. again, i apologize if i killed the thread, but something struck a nerve.

------------------
"edgecrusher" Spartan Probe Team 'angelis'