Thread: Hal

Skynet is approaching, fucking Judgment Day. Get yer guns ready!!

No, but really. That program seems to be pretty good.
But it's not conscious. It runs on Windows XP on a standard desktop with CPU. It cannot be conscious. People seem to think that systems that exhibit identical behavior are necessarily identical on all levels. That is absurd.

Biological chauvinism is absurd.

Originally Posted by Serkat

Skynet is approaching, fucking Judgment Day. Get yer guns ready!!

No, but really. That program seems to be pretty good.
But it's not conscious. It runs on Windows XP on a standard desktop with CPU. It cannot be conscious. People seem to think that systems that exhibit identical behavior are necessarily identical on all levels. That is absurd.

I kind of agree and disagree.
I think it has more of a minor-conscious. Not quite on par with humans yet, but obviously it can learn and be self aware. Those are some big steps towards consciousness.

1. I'm talking about qualia, not behavior.

Originally Posted by ClassyElf

I kind of agree and disagree.
I think it has more of a minor-conscious. Not quite on par with humans yet, but obviously it can learn and be self aware. Those are some big steps towards consciousness.

No. シ

Originally Posted by Xei

Biological chauvinism is absurd.

I'm not a "biological chauvinist". Refer to other thread. Non-carbon-based consciousness is possible, but not on CPU-based systems.

Originally Posted by Serkat

I'm not a "biological chauvinist". Refer to other thread. Non-carbon-based consciousness is possible, but not on CPU-based systems.

Suppose a CPU was programmed to simulate what a neural network would do by calculating inputs and outputs in small, discrete steps. Then suppose such a CPU was big and fast enough to simulate these steps in real time, such that an outside observer would have no way of knowing whether it was a real neural network or a simulated neural network. Now, you have already contended that consciousness doesn't depend on what produces it by saying it can be non-biological. Therefore, how can you say a physical neural network is conscious, but an equivalent simulated one is not?

Originally Posted by drewmandan

Suppose a CPU was programmed to simulate what a neural network would do by calculating inputs and outputs in small, discrete steps.

This is one problem. First, brains don't work in terms of "data" or "input and output", they are continuous processes. The concepts of input and output are used to model these processes, but they are not identical. Second, they are holistic processes. They don't work in discrete steps. Instead all neurons exist at the same time in reality, as specifically organized matter.
CPU processing isn't anything like that. The relevant information processing is limited to computations of tiny pieces of data. Most of the time none of the alleged neurons are anywhere except in the imagination of the observer. They don't exist.

Then suppose such a CPU was big and fast enough to simulate these steps in real time, such that an outside observer would have no way of knowing whether it was a real neural network or a simulated neural network.

Outside observers play no role in evaluating consciousness because observers observe behavior and that doesn't have anything to do with it.

Now, you have already contended that consciousness doesn't depend on what produces it by saying it can be non-biological.

The microscopic chemical elements are irrelevant. The macroscopic organization of these is not. If the neuron isn't matter, it's not a neuron. Information is a very valuable social construct but it's nothing that has any relevance outside of a human observer. Neurons don't spin at 10,000 rpm and they don't move through conductors in forms of electrons. They exist as material entities.

I'm not a "biological chauvinist". Refer to other thread. Non-carbon-based consciousness is possible, but not on CPU-based systems.

Although certainly not currently possible in practice, a computer could theoretically emulate the brain down to the cellular level... quantum effects are not computable but the brain is not quantum, it works on macroscopic, computable functions. If such a computer were to simulate an entire brain, I can't see why it wouldn't be conscious. It would act identically to a conscious person, anyway.

Originally Posted by Xei

Although certainly not currently possible in practice, a computer could theoretically emulate the brain down to the cellular level... quantum effects are not computable but the brain is not quantum, it works on macroscopic, computable functions. If such a computer were to simulate an entire brain, I can't see why it wouldn't be conscious. It would act identically to a conscious person, anyway.

Because behavior is just one aspect of a system. it wouldn't be conscious because there wouldn't be any neurons to be conscious.

It would still have physical embodiment though. If the physical makeup of the neural nodes is so unimportant, why not have them embodied as electrons and transistors on a microchip? Consciousness is the result of a mathematical system... and it seems to me that neural networks are isomorphisms.

Originally Posted by Serkat

This is one problem. First, brains don't work in terms of "data" or "input and output", they are continuous processes. The concepts of input and output are used to model these processes, but they are not identical. Second, they are holistic processes. They don't work in discrete steps. Instead all neurons exist at the same time in reality, as specifically organized matter.

Hogwash. The inputs to the human brain are discrete. They are the individual firings of transducer neurons, which behave in a very digital way. There's no such thing as 'half' a neuron firing, for example. This is the first thing you learn in psychology, by the way.

Now, if it were true that there are an infinite number of input neurons, then you might have a case. However, there is clearly a finite number of inputs in the human body, and not even all of those can be activated simultaneously anyway. But at any rate, a finite number of inputs means that you can, in principle, divide the input stream into small packets and solve the entire system one packet at a time, computational power permitting. In fact, I don't even need to argue with you here, because many scientists today already simulate neural networks with CPUs routinely.

Originally Posted by Serkat

Most of the time none of the alleged neurons are anywhere except in the imagination of the observer. They don't exist.

Do you use a calculator to perform difficult arithmetic or do you count shells? I trust you have the intelligence to see how this analogy applies.

Originally Posted by Serkat

Outside observers play no role in evaluating consciousness because observers observe behavior and that doesn't have anything to do with it.The microscopic chemical elements are irrelevant. The macroscopic organization of these is not. If the neuron isn't matter, it's not a neuron. Information is a very valuable social construct but it's nothing that has any relevance outside of a human observer. Neurons don't spin at 10,000 rpm and they don't move through conductors in forms of electrons. They exist as material entities.

Ah, the Chinese Room argument.
http://en.wikipedia.org/wiki/Chinese_room

I will allow you to read the various refutations at your leisure. I prefer to see the room and occupant as the system as a whole. The counter-argument is then that if the occupant memorized the rule book, then the system is just the occupant. This is incorrect; the system is still and always is the occupant and the rules, which just happen to be written in memory rather than ink. The point here is that the system can have any inner workings and still be, in totality, conscious.

And here's some further reading for you:
http://en.wikipedia.org/wiki/Neural_network_software

I've taken philosophy and neurology, thanks.

Also, I still feel that you're not talking about qualia...

Isn't Hal from 2001: A Space Odessy? And look how he turned out..

To be honest HAL was programmed that way...

And yeah, neurons are simple, discrete inputs adders, not continuous.

Searle's arguments piss me off greatly. The man in the Chinese Room does not understand, no, no more than one of my neurons 'understands' anything, both being a part of a system. The system, however, does understand.

I find the China brain argument even more ridiculous. The way he just says 'and therefore it isn't conscious because... well, that would be dumb, having consciousness hover over China'. That's not a proof at all, and it's no more dumb than consciousness 'hovering over our brains'.

Actually no. Only states of organized matter can be conscious. Systems can't be. Because they don't exist. It's real simple.

Well congratulations, apparently you've solved the hard problem. I suggest you publish your argument in a philsophy journal and become an overnight superstar.

All wordly systems are 'states of organised matter' bound by causality. It's just that it's harder to tell where the physical correspondense lies in some. A simulated neural network on a microchip is just as physical as a carbon one.

You're gonna have do a little explaining as to why the causal network needs to be obvious in order to be consciousness rather than just asserting it if you want to win me over...

Lets all log on and teach HAL how to lucid dream. Oh wait...

Originally Posted by Xei

Well congratulations, apparently you've solved the hard problem. I suggest you publish your argument in a philsophy journal and become an overnight superstar.

All wordly systems are 'states of organised matter' bound by causality. It's just that it's harder to tell where the physical correspondense lies in some. A simulated neural network on a microchip is just as physical as a carbon one.

You're gonna have do a little explaining as to why the causal network needs to be obvious in order to be consciousness rather than just asserting it if you want to win me over...

They don't need to be obvious, they need to be material. When I say organized matter, of course I mean organized in a specific macroscopic way, not just any way. The calculations performed by a CPU are no different for either a simulated neural network or a side-scrolling shooter. They are only different under the presumption of a bunch of purely subjective unfounded theories about what holds information together and over a specific time span. A brain is in states and these represent something and all neurons exist at the same time whereas in CPUs this is not the case at all. Wow, this is terrible English. Anyway, My point is that in brains all neurons exist at the same time in states but in a CPU you only have some basic computations going on. boom

Originally Posted by Serkat

My point is that in brains all neurons exist at the same time in states but in a CPU you only have some basic computations going on. boom

Actually you're wrong. The CPU simulates all of the neurons at the same time.

Originally Posted by Serkat

The calculations performed by a CPU are no different for either a simulated neural network or a side-scrolling shooter.

The calculations are actually extremely different. The two aren't even compatible with each other for comparison.

Originally Posted by ClassyElf

Actually you're wrong. The CPU simulates all of the neurons at the same time.

The calculations are actually extremely different. The two aren't even compatible with each other for comparison.

How so? As far as I'm aware, CPUs use a defined set of machine instructions to move and modify tiny data packages in registers. Your assertion that all of the neurons, their interconnected states and the data they represent can be contained within the few KBs of L1 cache or even the maximum 6-12 MB L2 cache of a CPU seems pretty absurd to me. Now unless this monster of a program requires some supercomputer, it will run the same on an old 32bit 1Ghz CPU as it will on a new x86-64 quad core. This is because the CPU is constantly exchanging data with the much larger RAM, an entirely different structure. Even if it fit within those tiny boundaries, it still wouldn't do it because that's not how computers work. There is so much other stuff going on (OS anyone?) that a single program's data is never going to be a permanent resident on the CPU.

A couple billion operations a second doesn't mean a couple billions operations at the same time. It only gives the illusion of synchronicity on the screen.

Why would the neurons have to be constantly there for consciousness anyway? And every neuron must be constantly 'there' in a computer during the simulation in the memory of the program... otherwise it'd keep forgetting where the neurons were...

Originally Posted by Xei

Why would the neurons have to be constantly there for consciousness anyway?

Because consciousness can't arise from nothing... Something has to be there for it to occur. If only a miniscule fraction of my neurons were to exist at every moment, alterating so that my whole brain would be complete as a mapping through time, that wouldn't be the same as just having one complete brain at all. And I don't think any notable consciousness would arise from that what is basically just a tiny number of neurons, even if in the next secpmd another couple of neurons were to exist. That's not how stuff works. Stuff is either there or not, and it only has an effect on other stuff so long as it is there.

And every neuron must be constantly 'there' in a computer during the simulation in the memory of the program... otherwise it'd keep forgetting where the neurons were...

Yes, it is in the memory, very much so. But that is an entirely different concept from having it all within the same structure that does the computation, because the neurons in the RAM aren't connected in any meaningful way and neither are they connected to the CPU in which the actual computation is happening. They're just sitting there, not doing anything until they're being copied to the CPU and then you get computation (which doesn't connect them either - just transforms some data that's then sent back to sit around some more) Until then, they're sitting there like on a hard disk. So by that logic, if I were to copy the temporary memory to the hard disk and turn the computer off, consciousness would still exist because of the data saved to the hard disk. Which would seem odd. If you're saying that neurons in RAM and CPU have anything to do with one another you might as well say that my neurons and yours are somehow cooperating and producing a common consciousness. Which would seem odd as well.

Originally Posted by drewmandan

Suppose a CPU was programmed to simulate what a neural network would do by calculating inputs and outputs in small, discrete steps. Then suppose such a CPU was big and fast enough to simulate these steps in real time, such that an outside observer would have no way of knowing whether it was a real neural network or a simulated neural network. Now, you have already contended that consciousness doesn't depend on what produces it by saying it can be non-biological. Therefore, how can you say a physical neural network is conscious, but an equivalent simulated one is not?

I agree. But it is impossible to test, as are so many things in philosophy of mind. Ironically, it basically comes down to people saying 'how can something that is just ones and zeros be concious?'. It is exactly the same as dualists saying 'how can something that is just matter be concious?'. It isn't just the ones and zeros or just the matter that causes conciousness. It is very much against your intuitions to see something in a computer, in a way outside of time and space (or atleast as we and everything around it experience it), as something with conciousness, but that doesn't make it nessecerially so.

On the A.I. at hand:

This A.I. is incredibly interesting, and a big step forward, but not concious. That was not really what they were aiming at. A new level of computer intelligence might result from this 'child' approach to creating A.I..

I think maybe there are tests for consciousness... I talked about it before here. Most neuroscientists would agree that consciousness is the result of physical manifestations, but what's not generally realised is that there are physical manifestations which are the result of consciousness; consciousness is not some kind of added layer with no causal consequences but rather an integral part of physical reality, neuroscience is making that clear.

Think about this thread for example. It is a physical manifestation of consciousness. If consciousness did not exist nobody would have made a thread about consciousness because we would have no concept of it, clearly. One could perhaps argue that something can delude itself that it is consciousness, but I think that that's a contradiction of terms.

Basically what I'm saying is that you can test somebody is conscious or not by asking them 'are you conscious'?

Ahh, semantics. I thought we were talking about qualia... not self-awarerness. Yes, a PC program can behave as though it has a concept of itself. That doesn't mean it has a subjective experience of having this concept.