Turing Machines
 

I'd like to open up this for discussion: for anyone that knows of the Turing Thesis, what do you think of it? Do you believe that anything is computable? If so, do you think that one day a fully functioning human can be reproduced, and will it be the exact same as the rest of us? What do you think on the matter of human free will and intentionality, do humans have it or not? Feel free to get as abstract or specific as you'd like. This doesn't necessarily need to turn into debate, as this topic tends to get volatile in discussion, just tell me what you think! :)

The Turing thesis is pretty narrow. In fact I think it's narrower than your assertion. There is a formalism to it. It concerns "computable functions." In that scope I think it's a rather trivial assumption. I just don't see how that applies to something as esoteric as freewill.

Note that certain classes of problems have been proven to be non computable. A classic example would be the halting problem. Closely related is Chaitin's constant. Also tangentially related would be Godel's incompleteness theroem. What I find important is these topics tell us that certain things are unknowable. It's part of the structure of reality.

By the way I believe in freewill. Not from a religious perspective. More based on quantum mechanics. We live in a world that gets fuzzy at the small scales. Luck of the draw rules there. The mechanisms that nerve cells function on summation and probability.

Of course not; Turing himself proved that false. However, as far as anyone can tell, the laws of physics are computable.
Spirit is all that matters. However, the difference between spirit and software is only one of viewpoint. ;)

It's perfectly possible - software can be replicated, it just needs the right hardware to run on.

Good post, and I also entirely agree with your last paragraph.

Interesting you should mention that because of the idea of replicating mind as well as the brain. Does the physical brain give rise to the mind, or does the mind operate on a completely different level? Any dualists out there lol? I would consider myself one.

If the universe is computable, and we have every indication that it is, then the mind is just as replicable as the brain.

The mind is just software, much like an OS. It can be transferred to other hardware if the profile is the same or it's within its capability to have a degree of abstraction for different ones. For that reason, emulation is also possible - reproduce the basic neural functions and with enough computing resources, you have something equal to or greater than a biological brain.

Good Morning Hal.

I think you need to go a couple of steps up if you want to use less than planets' worth of materials in said computer. You need to start directly executing the "program", not just blindly simulating hardware and hoping you've got everything hooked up right.

Hal was just doing what he was told, if you read 2010. (The issue appeared because he was given contradictory orders.) He also saves the entire crew at one point in that story, so he's certainly not dead weight.

I've been accused of being dualistic beyond reason.



Being computable in theory doesn't mean computable in practice.


Again there is that computable in practice thing. Neural networks that we know today are not representations of biological neurons. They are vastly simpler. I doubt there is enough computing power on earth to even simulate a tiny portion of the neurons that make up a human brain. I do know that there have been attempts to simulate the 302 neuron brain of the flatworm. We pretty much know what each of those neurons do and it's a very very tough problem. There is a big gap from there to our more like 1 trillion neuron brain were we have little clue as to function of each part.

It's easy to draw the line between hardware and software for our computers. It's not so simple with your brain. I would venture that they can not be separated. Even more the brain you have now is a lot different than what you were born with. It's constantly changing.

We've got software-controlled hardware already. We also know that the mind can be computed in a fairly small amount of space; the brain already does it. ;) (And the brain was semi-randomly evolved, which means that we should be able to make it far smaller.)

http://upload.wikimedia.org/wikipedi...en-auswahl.jpg
Remember those?
Never assume improvement will stagnate. Remember Moore's law.

I run virtual machines on my computers all the time - as far as the OS there can see, it's running on normal hardware, because all possible functions are implemented on top of the physical layer. Would that have been possible on a 286? no.

Very true, but that's implementable. There are almost certainly no people who could describe the function of all hardware down to the level of what that individual capacitor does and where that trace goes, yet they can for discrete components, and it's through their interfacing in a standard manner that something can be built across the top.

VMs run software, not hardware - they do not simulate the electric currents going through wires. That would be vastly inefficient, and is also what we're basically doing with neurons right now.

Exactly. Both have a layer of abstraction on the actual hardware.

Impossible to say it can be done now, we barely know something about how mind works, much harder to even emulate with current processing technology, which is, largely, an evolution of the Intel 4004 invented in 1971 (yes, we are currently using a technology developed more than 40 years ago). But later this century, we will certainly change the processing technologies and then, we will be able to make a better prediction.

We appear to be violently agreeing. :P

Everything's a Turing machine eventually. :cool: (Or a Wang machine, if you want to be slightly more practical.)

Whether a processor is x86 or not is immaterial as long as it's Turing complete - an architecture may be more or less efficient, particularly for specific instruction or operation types, but any one can implement the other at a higher level.

Which, of course, makes the high level implementation dependent of the knowledge of the system to be emulated, we are not yet there, therefore, in this moment it is impossible to make an accurate prediction, unless one with time constrains. And there's the problem of hardware implementation, which, at any level, is the major constrain to any machine.

It's implementable even if performance is low - Turing-complete languages are capable of calculating anything calculatable even with only a single conditional branching or loop type and operations to increment or decremen a value at a pointer, that is just excessively complex and inefficient even compared to the opcode level where there operations such as shifts. For that reason, even when something biological does not run on opcodes, even if it needs to be a biochemistry simulation, or indeed, even a simulation of an entire world/universe, it's implementable - just scaled to the point of current implausibility.

Indeed, it is completely implementable at any scale. but right now, we are constrained in resources, hardware in short. Those constrains, along with the emulation or simulation make a Turing machine just that, a simulation, no matter how complete it is (at least, a priori, it will change in the future). Obviously, the simulation could be much better than the actual system.