Thursday, January 08, 2015

Deceived by invisibility

The Newlyn-Phillips machine is an analogue computer for calculating the value of certain macroeconomic variables when other such variables are changed. It is a hydraulic machine that circulates colored water through a series of pipes and valves:

If you showed a bunch of people this machine in operation, and then asked them, "So, this machine sure knows a lot about the macro-economy, doesn't it?" I'd bet most would reply, "What? It's just a bunch of water running around in pipes. It's only humans who interpret the levels as economic aggregates." In fact, it is entirely possible (albeit unlikely) that someone built a machine that functions identically, but is actually used to flush clean some manufacturing equipment or run a fancy bathroom. In those cases, no one would even imagine that there was anything to do with macroeconomics going on.

But once the parts of a machine of essentially the same type are hidden in tiny microchips, and the machine circulates electrons instead of water, many people act like a cargo cult, willing to attribute to the machine powers like "knowing" how to play chess.

UPDATE: This post is not meant to prove that computers don't think. Perhaps they do, for all I know! It just shows we have no more reason to believe they do than we have to believe the Newlyn-Phillips machine thinks about macroeconomics.


  1. Couldn't sufficiently advanced circuitry give rise to thought?

    1. Could sufficiently advanced plumbing give rise to thought?

    2. "Asking if a computer can think is like asking if a submarine can swim."

      (It is worth googling to find out whom I am quoting.)

  2. Good one Gene, but alas it won't work on people who are still materialists. ("Was I any different at his age?")

    If you think human thoughts are "really" just electrical impulses in the nervous system, then...



"If your approach to mathematics is mechanical not mystical, you're not going to go anywhere." -- Nassim Nicholas Taleb