amkerman wrote:

LucidFlight wrote:

Positron wrote:But I was clear from the start that the simulation included data from a virtual world. Clearly a virtual world would itself be an algorithm. And I don't know what dictionary you are using but it is perfectly possible to embed all the data within an algorithm.

Normally, I wouldn't embed any data in an algorithm. An algorithm is a set of instructions — a procedure to process data. Do you see the embedding of data as necessary for the simulation? Would you be so kind as to explain why, please?

The algorithms for the "data" which is "embedded" in positrons proposed algorithm would be running simultaneously within postirons main algorithm I am guessing. I think that mights be where the confusion lies.

But I might be talking out my ass as well, so there is always that...

You are indeed. You can't synthesise any data unless you have a theory based on some prior data. That's evolution, for you. If you start everything with data that have been synthesised, you can build castles in the air. You can build the skyhooks used to construct the castles. Und so weiter.

GrahamH wrote:

Positron wrote:

GrahamH wrote:Just a side note: "sense data" is suggestive of a black-box view of a brain / world boundary. That would not be realistic. "the brain" is extended in "the world" and "the world" influences "the brain" in more ways than neuron firing events.

In what ways, for example? Do you mean like the temperature of the brain's environment? Physical shocks?

I am not sure how much else the brain is extended into it's environment - but it is all physical so it can be modelled.

One example is the way hormones disperse through the body. Where could you divide out the brain and say that hormone levels become a simple value input? We can model the whole body to avoid this problem. The nervous system extends throughout the body and is all part of the neural circuitry. Damasio makes the case that this connection to the body and viscera is critical to our sense of self.

Thinking of the brain as a black box is an error. At least include the whole body in the model.

But how many times do I have to point out what I said in the initial description of this - that there would be input from a virtual world which would obviously include the body?

SpeedOfSound wrote:

Positron wrote:So the question is simple, can we model the neural mechanism of the brain with an algorithm?

Yes. We can. They are already on it.

Well then, an algorithm can be desk checked. It doesn't take quintillions of workers all at once trained for billions of years to desk check and algorithm.

SpeedOfSound wrote:What was the point again?

What happens in the desk check is what happens in the computer and so if we have reason to doubt that we could be the desk check then we have reason to doubt that we could be that algorithm running on an electronic computer.

All the computer is doing when it runs an algorithm is taking numbers, making calculations and storing them someplace - like the desk checkers.

But the computer (and the desk check) would exhibit the behaviours of the physical system it is modelling and when that feeds back into the virtual world it will result in the sort of behaviour that we show.

So one of the outputs of the algorithm is that the virtual brain will report feeling a full bladder or feeling nauseous and my question was - why would it report feeling nauseous when it didn't feel nauseous?

And my original point was that if the computer, modelling the same neural architecture that we have, would claim to be conscious then it would suggest - if it is not conscious - that when we claim to be conscious we don't do so because we are conscious.

But I imagine that when I say I am conscious that it is because I am conscious.

So thinking about this leads me to either one or another conclusion that is, to me, absurd. And avoiding one absurd conclusion seems to run me smack bang into another absurd conclusion.

And finally - it would suggest that there is a non-conscious mechanism that can display the outward behaviour of a conscious being and since the hypotheses that science can have about consciousness are verified by behaviour - in particular language - then there can be a non-conscious mechanism that can verify any hypothesis that science can make about consciousness.

That is not an argument for Idealism or Dualism, but it is sufficient for me to remain a firm agnostic metaphysics - including Materialism.

GrahamH wrote:

Positron wrote:

SpeedOfSound wrote:

Positron wrote:So the question is simple, can we model the neural mechanism of the brain with an algorithm?

Yes. We can. They are already on it.

Well then, an algorithm can be desk checked. It doesn't take quintillions of workers all at once trained for billions of years to desk check an algorithm.

Well, it might do, depending on just how complex the algorithm and how large the data set.
Blue brain takes a lot of runtime on IBM's fastest supercomputer to simulate a few seconds of just one cortical column of a rats brain.

But if it is an algorithm then it would require very few people at any given point int the simulation to actually run the algorithm no matter how complex the algorithm, no matter how large the data set. That is the nature of an algorithm.

It would take a very long time to simulate just a second or two of mind simulation as I have already stated and it might take even longer than the billions of years I originally estimated.

But it would not take many workers at once and they would not need any long term training.

GrahamH wrote:Sure, it only takes one clerk to be a Turing machine and it can be learned in minutes.
You won't get any meaningful output in a thousand generations of clerks, and they will spend most of their time merely filling interim results (moving the tape) rather than computing, but in principle it can be done.

Well of course it wouldn't be an actual Turing Machine - no actual computer program is.

But, of course, a thousand generations wouldn't begin to do it, that is why I said billions of years for a couple of seconds of mind simulation - and that may be an underestimate.

GrahamH wrote:

Positron wrote:

GrahamH wrote:Sure, it only takes one clerk to be a Turing machine and it can be learned in minutes.
You won't get any meaningful output in a thousand generations of clerks, and they will spend most of their time merely filling interim results (moving the tape) rather than computing, but in principle it can be done.

Well of course it wouldn't be an actual Turing Machine - no actual computer program is.

But, of course, a thousand generations wouldn't begin to do it, that is why I said billions of years for a couple of seconds of mind simulation - and that may be an underestimate.

So, what is the value of this absurdity to the discussion?

What value was it when you said "you won't get any meaningful output in a thousand generations of clerks"?

In my original statement of this I had already pointed out the long time frame.

So I am puzzled as why you felt the need to point this out. You seemed to be implying that I had not thought it out.

I was only pointing out that I had.

GrahamH wrote:
You didn't answer my question.

I did. Your question below is a different question.

You posted "It doesn't take quintillions of workers all at once trained for billions of years to desk check an algorithm.", which is of no relevance.

Of course it is of relevance. Quintillions of workers all at once would obviously be impossible. A small number of workers over a long time period is possible.

We can just as well suppose quintillions of workers, or billions of years, or any other absurdly large numbers. What is your point?

It is you guys who are making such a hoo-ha song and dance chorus line over the desk check thing. I am only responding to your points as you make them.

So you tell me what is the point of the hoo-ha you are making over the example and then you will know the point of my reponses to them.

I would much rather discuss the questions this raises, which you guys don't seem to want to touch with a barge pole.

And by the way, why do you insist on putting "absurdly" in front of "large numbers"? There is nothing inherently absurd about a large number.

Positron wrote:
And my original point was that if the computer, modelling the same neural architecture that we have, would claim to be conscious then it would suggest - if it is not conscious - that when we claim to be conscious we don't do so because we are conscious.

But I imagine that when I say I am conscious that it is because I am conscious.

So thinking about this leads me to either one or another conclusion that is, to me, absurd. And avoiding one absurd conclusion seems to run me smack bang into another absurd conclusion.

I can only agree with you, the hard problem looks as intractable as ever. Materialism doesn’t help while I am sure from second to second that I am conscious. But, like all the materialists here (if not the theists) I feel strongly that there has to be at least some way of coming to terms with the possibility that I’m a machine made up of predictable elements.

It does seem easier if I imagine there isn’t any other person around. The distinction between “I” and the external world shows signs of disappearing. But imagining other people around without imagining them as conscious feels possible, but seriously wrong.

zoon wrote:I can only agree with you, the hard problem looks as intractable as ever. Materialism doesn’t help while I am sure from second to second that I am conscious. But, like all the materialists here (if not the theists) I feel strongly that there has to be at least some way of coming to terms with the possibility that I’m a machine made up of predictable elements.

Yes, in fact I think we can't avoid that we are made up of elements that are either deterministic, random or some combination of the two.

I can't see any alternative. But then again that might just be my limitations in thinking.

It does seem easier if I imagine there isn’t any other person around. The distinction between “I” and the external world shows signs of disappearing. But imagining other people around without imagining them as conscious feels possible, but seriously wrong.

As I said before I find it easier to dispense with any metaphysical dimension and just accept that we are what we are.

GrahamH wrote:
But can you refute the possibility that "a machine made of predictable elements" can not all the work of understanding?

Is anything understood beyond its raw recognition that it is present or its relations to other things? Is there any reason why brains, or machines, cannot do that raw recognition and contextual interpretation?

If it is valid to say a brain can understand, handle knowledge, then it can understand itself as a subject having experiences. It can be absolutely certain that this is the case. As certain as you or I are of our experiences, as self-blind as you and I are of the events underlying our experiences.

Predictability of the parts does not lead to practical predictability of the whole. See chaos theory.

I agree with what you say (keeping open the possibility of practical prediction), but I think it’s easier to think about in materialistic terms if there’s some materialistic reason for seeing oneself as a subject having experiences. For a solitary individual, this looks like a waste of brain activity. It makes sense to me to suppose that we first see other people as conscious, because it’s useful to predict them (or rather, guess more accurately what they will do) by simulating their brain states. It is then useful to see myself as other people see me, a conscious being, because this gives me a better chance of predicting (again, improved guesswork) how they will react to what I do.

So I think it’s in a social context that I understand myself as a subject having experiences. Without the social context this becomes unnecessary.