Christopher Lord

Vibr8gKiwi: "but the actual calculation from inputs to outputs is deterministic (you'll get the same set of outputs for the same set of inputs every time)"

My point is that the brain has many inputs, such as heat, temperature, chemical concentration, electrical charge, gravity, age of the cells involved, as well as inputs from other neurons and their ilk.

All of these vary not only the output, but also the function. they can feed into each other, and produce error and noise that do not appear in pure software neurons such as those you've probably worked with.

The brain, given the exact same sensory input, will not behave deterministically, like a net in a software emulation would. It would only behave deterministically as far as the error bars allow it. tiny fluctuations (randomness) will be amplified by neurons in such a case and produce divergant behavior.

I agree that if all inputs, including perhaps even the exact state of the universe, were fed in, we could have a deterministic organic brain. Otherwise, there is too much randomness in input, output, and function.

This is why I hold the opinion that a computer does not 'choose', but we can.

Choosing implies a choice, as was stated earlier. A typical game AI MUST respond in a certain way to certain input. There is no choice in the matter.

A brain is not constrained to 'responding only' because it is not constrained by a 'certain input'. Input is entirely analog, and so is not entirely certain.

K

Christopher Lord:

It would be quite easy to have a computer program take ambient temperature, humidity, time of day, and a host of other things into its decision. Would it then be making a choice? And how does randomness imply choice? Randomness would seem to imply lack of choice - that is unless the individual can somehow control the choice.

Christopher Lord

No, it would not be easy for a computer to take these into account. These factors actualy vary the function of neurons. A transistor operates as a transistor under all situations.

That does not preclude an emulation layer, but that is not what we are discussing! The qeustion was specifically 'do [contemporary] computer games choose?'

My answer was no for all of the reasons I outline in posts above.

Amen-Moses

Military flight control software is now at the stage where it can learn enough about a particular pilot to predict what the pilot is going to do before the pilot is even aware of having made a decision. The reason for these systems is that in the height of a dog fight or during the stressful part of final approach to ground targets the pilot will start to concentrate on a single task to the detriment of other ancillary tasks. By learning what the pilot does in particular circumstances the computer can take over those tasks whilst the pilot is busy, acting like a virtual co-pilot. This happens so seemlessly that the pilot is unaware that it isn't him doing things but a sort of virtual twin.

In the industry the big joke amongst software engineers is that for at least ten years the only component in the cockpit that isn't really needed is the biological part, in fact so much effort goes into software to correct pilot errors that we could probably cut the cost of the system by 50% if the pilot was removed.

I wonder how many people flying across the US are aware that in most cases it is software that taxis the plane to the end of the runway, software that takes off, cruises and lands the plane and software that taxis them to their arrival gate. The pilot has become a back up system!

(oh, another joke was when we were talking about a replacement for the Harrier ond A10, both tank buster coming to the end of their lives, we eventually decided that a cruise missile with a refuelling probe and a 30mm cannon would be the ultimate replacement!)

Amen-Moses

Liquidrage

Exactly. And you are making up your own definition of choice. You are attempting to turn choice into a synonym of free-will. But that is not what choice means. Nor does free-will mean the ability to choose. As I tried to point out earlier, use a dictionary. You may think you are using these words/terms properly, but you are not.

"A brain is not constrained to 'responding only' because it is not constrained by a 'certain input'. Input is entirely analog, and so is not entirely certain."

Computer games more typcially then not use analog inputs. But get back to defining your terms.

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

Vibr8gKiwi

This is really the meat here I think. Computer can be put in the same state again and again to see how their decisions come about, and can be shown to be deterministic or non-deterministic. People cannot be. However as I think humans are "meat machines", I'm with you that if we could back up and get into the exact same state again, the same decision would be made.

Starboy

I don't think it is necessary to go to such lengths. We are awash in random events, everything from radio noise to cosmic ray events. Who is to say that such things don't randomize everything on the planet from time to time? I am not sure, but I suspect that there have been a few times in my thirty years of working with computers that a random change in a memory bit did indeed occur. There was this time when my computer crashed with the blue screen of death, and my computer kept saying over and over "Fuck you asshole!" in Arnolds voice of course.

Starboy

Kent Stevens

In computer games there can be times where there is no randomness by the computer and a move is still made. It might be because the computer is forced to make a move or else it will lose the game immediately. Therefore randomness is not absolutely necessary for the computer to make moves.

Starboy

Yes Kent, what is your point? Are you saying that creatures with free will don't do the same thing?

Starboy

Is human intelligence required to choose? If your answer is yes, then only humans have choice, if it is no, then a computer program can choose. What is your answer?

Starboy