Liquidrage

You're not looking deep enough which is my main point. You are falling prey to the abstraction layer that you yourself are defining to meet you preconvieved conclusions.

Obviously, unless you believe the universe is not made of disrete entities analog imputs are the same as binary in the most fundamental sense.
And of course human minds and computers are not exactly the same. Humans minds and current analog processors are able to take approximations. Approximations based on predefined thresholds. Else the onus of proof lies on you to show me how a analog system can even make an decisions when they have no thresholds allowing them to make a decisions.

You use determinism. I find this odd.
What sense do you use this? Are you using it with the ability to predict the future? Which has nothing to do with determinism by the way. How are you meaning free will?

Also, choice needs to be defined better, else we are wasting our time arguing on the same side.
To me a choice is the act of selecting or deciding when an alternative option is present. Use a dictionary for Pete's sake. Choice has nothing to do with free-will.

You are caliming computer games don't make choices. But to do so (since you are obviously familiar with programming) you would need to remove basic logic from the programs. A simple "if" statement qualifies as a choice.

Vibr8gKiwi

This is easily shown false. Analog inputs contain an entire range of values while a digital input is either "on" or "off." Take for example a driving game with a digital left/right pad vs an analog stick that gives a range of how far left/right. Certainly these are very different in the fundamental sense.

Neural networks don't put thesholds on individual inputs but work on a sort of weighted function of combined inputs. The output of the network depends on the combined inputs not a simple threshold of inputs. Here's a possible example network input output table:
input A, input B, Output
0.51, 0.25, 1
0.49, 0.25, 0
0.49, 0.30, 1

From the first two lines it might seem the network has a "threshold" for input A at 0.50 that flips the output value. However the third line shows us the output is actually computed by a combination of both inputs. Simple "thresholds" are not used. There are literally near infinite possible outputs for two floating point inputs with a neural net (limited by the resolution of the floating point numbers and network structure).

I agree with you here. I think the claim that computers don't make choices is rather strange. If they are running deterministic algorithms, those choices will be consistent--for the same inputs, the computer will "choose" the same outputs every time. For a non-deterministic algorithm, the same set of inputs might cause the computer to "choose" different outputs even if the same inputs were used. It's even possible for a computer to be truly random (though this is difficult), for example there was a server on the web that provided truly random values (from the measured decay of a radioactive isotope) when queried. Someone using such real random data would not be even remotely deterministic as pseudorandom algorithms can be argued to be.

[ September 08, 2002: Message edited by: Vibr8gKiwi ]</p>

Liquidrage

This is easily shown false. Analog inputs contain an entire range of values while a digital input is either "on" or "off." Take for example a driving game with a digital left/right pad vs an analog stick that gives a range of how far left/right. Certainly these are very different in the fundamental sense.

Read what I wrote again. Notice the word "inputs"? I agree with what you wrote. It just had nothing to do with what I wrote.

From the first two lines it might seem the network has a "threshold" for input A at 0.50 that flips the output value. However the third line shows us the output is actually computed by a combination of both inputs. Simple "thresholds" are not used. There are literally near infinite possible outputs for two floating point inputs with a neural net (limited by the resolution of the floating point numbers and network structure).

Don't really disagree. But it's not what I wrote again. You here are invoking complexity. You've added a simply algorithm and decided that it's result was your threshold.

Edit:
I see what you mean in the first part now. My bad in missreading what you wrote. However, it is still a fallacy you are clinging too and had nothing to do with what I wrote.
With what you are claiming is analog and binary systems live in different universes with different types of fundamental particles. Which means you are also clinging to an abstraction layer here.
What is an analog signal made of?
A binary singal?
Exactly! The same damn thing.
They are just interpreated differently.
Realistically, in a computer sense the inputs are made to arive in different forms.
My point was if you break it down to it's MOST FUNDAMENTAL level (I've used that term before notice) the inputs are are made of the same things as long as you believe in a discrete universe.

[ September 07, 2002: Message edited by: Liquidrage ]</p>

hinduwoman

It seems to me that philosophically there is not much difference between advanced computers and humans choosing.
Humans have a limited number of options from which they choose one.
Computer programming gives the software a limited number of options from which it chooses one.

Is there really a difference?

Vibr8gKiwi

No. You can sample a continuous wave or function to binary values, but the original analog function is continuous. The binary is a finite model, the continuous function is infinite (between any two points in a continuous function is yet another point).

If you've ever seen those fractal viewer programs, that's the perfect example of what I mean. You can see that you can zoom in on "locations" in a fractal function infinitely. Those fractals show simple functions contain infinite universes of data with infinite "zoom" capability that can never be sampled to binary (because you can always zoom more). It's pointless to try to make a binary "capture" of all the pictures that a fractal function contains--you could fill CD after CD and still not get them all. Yet a single simple mathematical statement holds them all. Fractal functions are a perfect example of where analog fits a problem much better than binary.

Computers and mathematics are not limited to just the discrete (I don't think the universe is either).

But really this entire argument is old and complex (I recall reading a chapter on the argument in Bart Kosko's "Fuzzy Thinking"), and certainly no discussion board will do it justice. But the bottom line is I agree you can model analog with many digital values, but sometimes that's more complex than going with an analog or fuzzy logic solution in the first place. I don't think there is a right or wrong as both can work in various situations, but there certainly is an easier and more elegant solution for each problem. The best thing is to understand both and know when to apply each.

There's a reason the Playstation controller has a binary/analog mode switch

[ September 07, 2002: Message edited by: Vibr8gKiwi ]</p>

Liquidrage

I don't claim there isn't a difference between the two types of systems. Hence the need for the switch. It is only in the most fundamental form that the inputs are the same. Hence, my use of the term "abstraction" being used often here.

And if you don't believe the universe is discrete I'd like to know why.
Surely you are in the vast minority here and I doubt either you have the information needed to make such a claim. Surely I dont either, but I've read much on the subject and I have not seen a single source even claim that possible the universe is not discrete in it's most fundamental form.

In terms of the system, yes you are right. No message board will do justice to the discussion. Which I why I mentioned in the first place. Such a profound claim as originally put forth by Christopher is absurd.

Computers make choices. It is clear and in my opinion nondebatable. To debate it you would have to redefine the word "choice" to include free-will. And that is itself a fallacy.

Vibr8gKiwi

This is way off topic, and it's just a small thought I have. I don't think the universe is discrete because the universes in fractals are not discrete. Nature loves fractals, they are everywhere. I suspect nature and the Universe is a sort of fractal (perhaps Stephen Wolfram does also) and as well as going on forever in space, can be "zoomed in on" infinitely. Why would it have an end to the zooming? What would be at the end? Between every two numbers is another number. Dig it.

Starboy

Maybe this is the wrong kind of response, but asking the question "does the computer choose" implies that it has more than one course of action. In other words, if when you run the program again and enter the same moves as before, does the same thing happen. If it diverges from previous sessions then, operationally the computer has choices, if the same entry always gets the same response, then operationally the computer has no choice. Since it is easy to put a randomizer into any program, from an operational point of view it is possible to give the computer a choice. Maybe when you get down to it exercising choice is just rolling the dice.

Starboy

Liquidrage

Well as was nicely poined out previously, you can tie randomization algorithims into quantum uncertainty by using atomic decay and thus quantum uncertainty.
I myself have never read this as happening, but I do not doubt it for a minute.
Though it isn't used for games, I don't believe we need to go this far to meet the accepted definition of "choose".
"Accepted definition" being paramount.

K

Vibr8gKiwi:

This is kind of off topic, but analog systems are really only continuous modes of discrete systems - at least as far as electronics are concerned. There is no such thing as charges that are fractions of the charge on one electron (unless you're talking quarks, but that's another story).


For everyone who believes randomness is so important to the decision, does anyone believe that if a human brain was put several times in the exact same state, with the exact same choice to make, the choice would be random? If so, what convinces you that this is true? Based on the behavior of neurons, I would believe that it's far more likely that the choice would always result in the same outcome.