RE: wonderings
10-22-2018, 09:08 PM
Most people would accept that a computer is Turing-complete -- ignoring the fact that a truly Turing-complete machine has infinite memory while (most) of our computers don't. Most people would also accept that something is Turing-complete if it can simulate something that is already Turing-complete.
Is an infinite bag of electronics parts -- wires, resistors, capacitors, inductors, transistors, and diodes -- Turing-complete, then? You can compute anything with them, including a computer, and oftentimes you can make a much more efficient computation machine by building it with electronics rather than by building it with a computer then programming the computer.
The issue is that putting in the input feels an awful lot like building the machine, rather than just telling it what you want to compute. Are the collection of parts or the electromagnetic field really computing anything?
What about a 20 kiloton pile of iron, copper, and various rare earth metal ores? You can make electronics parts from that. But this seems an awful lot less like "simulating" something, and an awful lot more like just "making" it.
What about a single wooden block? You can probably carve something Turing-complete out of wood if you have a good enough chisel.
At each of these examples (it seems intuitive that) less and less computing power is actually present in our "computer" and more and more computing power and work is put into configuring the "inputs" and interpreting the "outputs". Clearly, we have to include other information about the system in the part we consider Turing-complete, probably including the method of adding inputs and instructions on interpreting outputs.
But at what point could we just take the input and output system alone, replacing our "Turing machine" with some arbitrary object? And at what point do we consider the human integral to the operation of our machine?
Is an infinite bag of electronics parts -- wires, resistors, capacitors, inductors, transistors, and diodes -- Turing-complete, then? You can compute anything with them, including a computer, and oftentimes you can make a much more efficient computation machine by building it with electronics rather than by building it with a computer then programming the computer.
The issue is that putting in the input feels an awful lot like building the machine, rather than just telling it what you want to compute. Are the collection of parts or the electromagnetic field really computing anything?
What about a 20 kiloton pile of iron, copper, and various rare earth metal ores? You can make electronics parts from that. But this seems an awful lot less like "simulating" something, and an awful lot more like just "making" it.
What about a single wooden block? You can probably carve something Turing-complete out of wood if you have a good enough chisel.
At each of these examples (it seems intuitive that) less and less computing power is actually present in our "computer" and more and more computing power and work is put into configuring the "inputs" and interpreting the "outputs". Clearly, we have to include other information about the system in the part we consider Turing-complete, probably including the method of adding inputs and instructions on interpreting outputs.
But at what point could we just take the input and output system alone, replacing our "Turing machine" with some arbitrary object? And at what point do we consider the human integral to the operation of our machine?