Instead of using Conway's game of life to represent life forms, try using it to represent chemical reactions, explosions, and the like. Or even movement. As I said earlier in this blog, we can call this Conway's Game of Physics. If we find problems with the Conwayesque representation of inanimate lower/less-complex forms, then we would be reasonable in suspecting greater problem to attend the representation of (them being higher level / more complex) life forms.
One indicator of the sort of problems that I anticipate would be the fact (or purported fact) that nature doesn't operate in a voxel-like way. Rather, any two or three dimensional pixel/voxel representation of nature must piggy-back on a nature that is utterly different.
Nature is not voxels "all the way down."
Can a voxel or pixel even represent movement? Take the 8x8 square (with rows represented by letters A through H and columns represented by numbers 1 through 8) At time-1 A1 is white while all of the other pixels are dark; at time-2 B2 is white while all of the others are dark. Then C3; D4; etc. We can interpret this transition as continuous diagonal movement, but we don't have to: we can also interpret it as a kind of quantum-like leap in the location of the same item. Or (we can interpret it) as different squares remaining immobile but going on and off. If the same "given" can be interpreted in such different ways, then there is a lack of constraint on the first sort of interpretation (which Dennet expects us to employ).
This lack of constraint is the sort of thing one finds with an abstraction. 1+2=3 can be about apples or oranges--or just about number itself. Conway's game of physics can be about a grid-like universe entirely lacking in movement, or it can be a purely geometric object--or it can be, I suppose, the sort of thing that Dennett hopes we will interpret it as. But it doesn't have to be. And if we do interpret it in that hoped-for way, then it's because we've engaged in a bit of isogesis.
Unlike the abstractions distilled from nature, nature itself always has more to say than what we've abstracted from it. Voxels, on the other hand, are like mathematical objects: as abstractions, they have no more to tell us about themselves than we've told them they can say. They lack the sort of thickness that nature itself possesses.
Actually, two different things said by Dennett are somewhat at odds with each other. Later (than his discussion of voxels and pixels) in Evolution Evolves, Daniel Dennett criticizes those who think neural networks, which operate non-algorithically (we'll therefore call them NaNNs), can do what Turing machines cannot. He points out that proponents of the virtues of NaNNs base their claim on observations made of computer models of networks. But the models that they use rely on the algorithmic properties of computer programs to represent the non-algorithmic properties of neural networks. Hence the purportedly non-algorithmic operations attributed to these models are some very basic level, algorithmic. It's as if the proponents of the virtues of NaNNs are cheating by piggy-backing on algorithmic software to score points in favor of non-algorithmic hardware.
But isn't Dennett himself cheating by having us imagine a Democratean universe when the one we live in is manifestly not such? And any computer that he used to model a pixel or voxel-like universe would do that modeling only by piggy backing on a universe that is quite different.
Then again, he may be glad to acknowledge all this as long as I grant that Turning machines rely at the most basic level on something non-algorithmic (i.e., on hardware) do the sorts of things that he supposes they can do.
Okay, I'm not all that sure of the significance of all this, but I'm posting it with the hope that I'll see the big picture better later on.
One indicator of the sort of problems that I anticipate would be the fact (or purported fact) that nature doesn't operate in a voxel-like way. Rather, any two or three dimensional pixel/voxel representation of nature must piggy-back on a nature that is utterly different.
Nature is not voxels "all the way down."
Can a voxel or pixel even represent movement? Take the 8x8 square (with rows represented by letters A through H and columns represented by numbers 1 through 8) At time-1 A1 is white while all of the other pixels are dark; at time-2 B2 is white while all of the others are dark. Then C3; D4; etc. We can interpret this transition as continuous diagonal movement, but we don't have to: we can also interpret it as a kind of quantum-like leap in the location of the same item. Or (we can interpret it) as different squares remaining immobile but going on and off. If the same "given" can be interpreted in such different ways, then there is a lack of constraint on the first sort of interpretation (which Dennet expects us to employ).
This lack of constraint is the sort of thing one finds with an abstraction. 1+2=3 can be about apples or oranges--or just about number itself. Conway's game of physics can be about a grid-like universe entirely lacking in movement, or it can be a purely geometric object--or it can be, I suppose, the sort of thing that Dennett hopes we will interpret it as. But it doesn't have to be. And if we do interpret it in that hoped-for way, then it's because we've engaged in a bit of isogesis.
Unlike the abstractions distilled from nature, nature itself always has more to say than what we've abstracted from it. Voxels, on the other hand, are like mathematical objects: as abstractions, they have no more to tell us about themselves than we've told them they can say. They lack the sort of thickness that nature itself possesses.
Actually, two different things said by Dennett are somewhat at odds with each other. Later (than his discussion of voxels and pixels) in Evolution Evolves, Daniel Dennett criticizes those who think neural networks, which operate non-algorithically (we'll therefore call them NaNNs), can do what Turing machines cannot. He points out that proponents of the virtues of NaNNs base their claim on observations made of computer models of networks. But the models that they use rely on the algorithmic properties of computer programs to represent the non-algorithmic properties of neural networks. Hence the purportedly non-algorithmic operations attributed to these models are some very basic level, algorithmic. It's as if the proponents of the virtues of NaNNs are cheating by piggy-backing on algorithmic software to score points in favor of non-algorithmic hardware.
But isn't Dennett himself cheating by having us imagine a Democratean universe when the one we live in is manifestly not such? And any computer that he used to model a pixel or voxel-like universe would do that modeling only by piggy backing on a universe that is quite different.
Then again, he may be glad to acknowledge all this as long as I grant that Turning machines rely at the most basic level on something non-algorithmic (i.e., on hardware) do the sorts of things that he supposes they can do.
Okay, I'm not all that sure of the significance of all this, but I'm posting it with the hope that I'll see the big picture better later on.
Comments