Comment by tsimionescu
7 hours ago
And yet you can build a device with the exact same functionality using vacuum tubes, semiconductor transistors, field effect transistors, water pipes, ant molehills, and any other substrate - and you could even replace some of the components with a software-defined hardware component that does the same thing. The computation is the thing that is objectively the same between all of these different realizations of the same device - the software that they are running. And for many of these, the software is indeed a physical object, one whose presence you can precisely measure. A hard disk containing a copy of quicksort has different physical properties that the same hard-disk containing a copy of Windows. A CPU currently running quicksort is likewise different from a CPU currently running ChatGPT, in perfectly measurable and observable ways.
A computational description of a system is no more and no less rigurous than any other physical model of that system. To the same extent that you can say that billiards balls interact by colliding with each other and the table, you can say that a processor is computing some function by flipping currents through transistors.
> software-defined hardware component that does the same thing
No, you cannot.
A hard-drive needs to a have a physical hysteresis. An input/output device needs to transmit power, and be powered, by an electrical field. A visual device needs to emit light on electrical stimulation, and so on.
The only sense, in the end, in which a "computer" survives its devices being changed is just observer-relative. You attribute a "3" to one state and a "1" to another, and "addition" to some process. By your attribution, does that process compute "4".
But it computes everything and computes nothing. If you plug in a speaker to VGA socket, the electrical signal causes an the air to move, sound.
The only sense in which a VGA signal is a "visual" signal is that we attach an LCD to that socket, and we interpret the light from the LCD semantically.
The world is a particular way objects in space and time move, those exhaust all physical properties. Any other properties are non-physical, which is why this kind of computationalism is really dualism.
You suppose it isnt your physical mechanism and its relationship to your environment which constitutes your thinking -- rather it's your soul. A pure abstract pattern which needs no devices with no specific properties to be realised.
Whatever this pattern is, if you played it through a speaker, it would just be vibrations in the air. Sent to an LCD, whitenoise. Only realised in your specific biology is it any kind of thinking at all.
Again, this is simply and provably false. I can build a system that opens a door when I'm near it using a photodiode connected to a measurement pin and have the CPU trigger the door opening motor if the diode is indicating no light, and the door closing motor if it indicates light. Or, I can buy a camera and build a complex software solution to analyze the output of that camera, and open the door if the software sets the "is_present" bit and otherwise closes the door.
In either case, the door will open if you're in front of it, and close after you've gone. This will happen regardless of whether you undertsand what it represents, it will open for a basic robot as well as for a human or a squirrel or a plant growing towards it very slowly or a rock rolling downhill.
Of course, you can't replace every single piece of hardware with software - you still need some link with the physical world. And of course, there will be many measurable differences between the two systems - for a basic example, the camera-based system will give off a lot more heat than the photo-sensitive diode one. I'm not claiming that they are perfectly equivalent in every way, not at all. I am claiming that they are equivalent in some measurable, observer-independent ways, and that the specific way in which they are equivalent is that they are running the same computation.
"opening a door" is an observer-relative purpose
Yes, you can intepret systems as having a goal and realise that goal using a vareity of different devices.
Reality itself doesnt have purposes, there are no goals. "A device that opens a door" isnt a physical process, it's a goal.
Go do the same with chemistry, physics, biology -- no, actual relaity doesnt have purposes. Hexane isnt methane, gravity isnt electromagnetism, the motion of air molecuels isnt the emission of light.
Any time "one thing can serve the purpose as another" you are, by definition, working in the world of human intention.
Your entire observer-relative purpose-attributing "engineering mania" here is anti-naturalistic dualism. Reality is a place of specific causes, not of roles/pruposes/goals/devices
Fire is the thing which is a plasma disposed to burn in oxygen which results from a specific chemical/etc. process etc. etc. There is no "water fire".
Insofar as an object can causally interact with another such that it "pushes it out of the way" -- the property had by all such objects relates to the pauli exclusion principle (essentially) and refined by surface area, volume, density and the like. To "open a door" is to displace wood in a certain location, to do that is to exist such that the femionic structure of the wood is excluded from that place.
2 replies →