[ExI] The digital nature of thermostats

Eric Messick eric at m056832107.syzygy.com
Sat Jan 23 23:34:36 UTC 2010

Gordon writes:
>Modern thermostats contain digital circuitry but they do not equal
> digital simulations of analog thermostats.
>To see this, imagine that you have an instrument for scanning objects
> to create digital simulations. You scan an analog thermostat and
> observe the resulting simulation on your computer. You will not see a
> real digital thermostat appear on your computer screen. Instead you
> will see a digital simulation of a non-digital object. That
> simulation will not have the properties of the original; it will not
> have the capacity to regulate temperature in your room.

I don't understand your assertion that the thermostat hanging on my
wall does not simulate an analog one.

What does a thermostat do?  It commands a furnace to turn on or off
based on the difference between the room temperature and a set-point.
An analog one does this by means of a tilt switch connected to a
bi-metallic strip.

We could simulate the behavior of that bi-metallic strip at the atomic
level in a giant supercomputer, and connect the simulation up to a
silicon temperature sensor and a relay.  That simulation would
regulate the temperature of a real room.

We could use less power by noticing that there was a simple
relationship between the input temperature and the state of the relay,
and replace the supercomputer with a small embedded processor.  We've
placed an abstraction boundary around the bi-metallic strip, and
simplified it to a compare instruction in the embedded program.

That compare instruction is still simulating the behavior of the
bi-metallic strip.  The box we've built looks just like the one hanging
on my wall.  It's got screw mounts for wires to control the furnace,
just like the analog one.  It reacts to changes in the room
temperature, just like the analog one.  It regulates the temperature
of the room, just like the analog one.  Inside, it has a CPU running a
program which simulates an abstract analog thermostat by using a
simple compare instruction.

We build a p-neuron the same way.  We give it sensors to detect the
firings of neurons around it.  We give it a CPU with a program to
simulate the behavior of a neuron.  We give it output devices for it
to trigger to signal other neurons.  In this case, the simulation is
more complicated than a simple compare instruction, because the
behavior of a neuron is much more complex than the behavior of a

Note that we can already do all of these things.  Cochlear implants
turn sound into neural signals.  Nature reported in 2006 on a system
which used neural signals to control a prosthetic hand (similar to the
EPOC EEG headset being discussed in another thread, but more


 Neuronal ensemble control of prosthetic devices by a human with tetraplegia

     Neuronal ensemble activity recorded through a 96-microelectrode
     array implanted in primary motor cortex demonstrated that
     intended hand motion modulates cortical spiking patterns three
     years after spinal cord injury. Decoders were created, providing
     a 'neural cursor' with which [the subject] MN opened simulated
     e-mail and operated devices such as a television, even while
     conversing. Furthermore, MN used neural control to open and close
     a prosthetic hand, and perform rudimentary actions with a
     multi-jointed robotic arm.

This paper describes precursor technology to the p-neurons we
discussed earlier.  This experiment is very similar to the partial
replacement experiment we discussed earlier.

MN was able to open *real* email and was able to move a *real* robotic
arm using his simulated p-neurons.  People with cochlear implants hear
*real* sounds through their simulated p-neurons.

And before you say that we're not actually simulating neurons here,
note that in both cases there are complex encoding and decoding
processes to generate and interpret the neural signals involved.  That
encoding and decoding is information processing that would otherwise
have been performed by other neurons.  The algorithmic programs in the
prostheses are replacing and emulating those neural sub-systems.

The key point here is that we hook our simulation up to the real world
through input/output devices of some sort.  A computer with no I/O
devices is not very useful (except perhaps to it's inhabitants).
Those I/O channels allow our simulation to affect and be affected by
real objects.

The I/O devices are where symbols are grounded.


More information about the extropy-chat mailing list