In Design for a Brain, W. Ross Ashby speculates about the possibility of creating a mobile homeostat "with its critical states set so that it seeks situations of high illumination." This talk explores an embedding of Ashby's homeostat within a simulated robot and environment, exploring the question as to whether the classic homeostat architecture is able to adapt to this environment. Remaining faithful to the physical design of Ashby's device, this simulation enables us to quantitatively evaluate Ashby's proposition that homeostasis can be achieved through ultrastability.
2. W. Ross Ashby’s Homeostat
• four surplus RAF bomb
control switch-gear
kits, refashioned into
electro-mechanical
artificial ‘neurons’.
http://www.rossashby.info
3. Not just stable but
ultra-stable
• homeostasis: As a queen bee
lays her eggs, bees maintain a
stable cluster temperature at
around 20ºC. Bees on the
outside of the cluster form an
insulating shell while those on
the inside can generate heat by
contracting their thoracic
muscles.
• Maintaining temperature is
essential to bee survival.
• Ultrastability is about
maintaining essential variables
within limits.
5. Simulated Environment
•The homeostat represented both brain & anti-brain.
•This changes with the mobile homeostat.
Velocity is
average of
the 2 wheels
• 2D kinematic model based on the Rossum Project
<http://rossum.sourceforge.net/papers/DiffSteer/>
Proportional to
the difference
between the 2
wheels
Cosine of the
angle to the
light source
(brightness
constancy)
6. Ashby’s Mobile Homeostat
“Suppose U is mobile
and is ultrastable,
with its critical
states set so that it
seeks situations of
high illumination.”
7. Experiment
http://stevebattle.github.io/homeostat/
• “Marina severed the attachments of the ..
muscles of a monkey’s eyeball and re-
attached them in a crossed position”
• “Sperry severed the nerves supplying
the .. muscles in the arm of a spider
monkey, and rejoined them in a crossed
position”
• See Di Paolo, “Homeostatic adaptation to
inversion of the visual field and other
sensorimotor disruptions”.
8. Sensory Inversion
1. Manual
deflection
on unit 2
2. Positive
response
from unit 1
3. Reverse
polarity from
unit 1 to 2
4. Unit 1
goes out of
bounds
5. Unit 1
restabilizes
6. Manual
deflection
on unit 2
7. Negative
response
from unit 1
9. Systems of variables
left eye
right eye
distance
left motor
right motor
unit 1
unit 2
manual parameter
reduced connectivity
10. The Law of
Requisite Variety
“An organism should be as intelligent as its
environment - no more no less.”
• If we add more units than are necessary, the
time taken to reach stability increases.
• If we reduce connectivity (while maintaining
functionality) the time taken is reduced.
“coordination can take place through the
environment; communication within the nervous
system is not always necessary.”
12. U Robot
“Suppose U is mobile and is ultrastable.”
Eyes mounted 90° apart.
Forward and rear facing
sensors return cosine of
light intensity.
Brightness constancy
maintained by scanning
sensor.
13. Connections between even a
small number of brain cells can
produce complex behaviours.
William Grey Walter, Bristol 1948
15. Machina sopora
• Walter likened the
homeostat to a “fireside
cat or dog which only
stirs when disturbed,
and then methodically
finds a comfortable
position and goes to
sleep again.”
16. The dimensions of autonomy*
A. The extent to which response to the
environment is direct or indirect.
B. The extent to which the controlling
mechanisms were self-generated rather than
externally imposed.
C.The extent to which inner directing
mechanisms can be reflected upon, and/or
selectively modified.
Autonomy is the greater, the more behaviour is directed
by self-generated (and idiosyncratic) inner mechanisms,
nicely responsive to the specific problem-situation,
yet reflexively modifiable by wider concerns.
*Boden, Autonomy and Artificiality
17. M.speculatrix is not
just reactive, it has
internal state
M.sopora is not just
reactive but has
internal feedback loops
M.spec's mechanisms
were hand-crafted by
Grey Walter
M.sopora’s internal
connectivity is
randomly self-generated
M.spec is adapted for
survival, anticipating
threats to its
essential variables
M.sopora is highly
adaptive to change
being ultrastable
A.
B.
C.
✔
✘
✔
✔
✔✔
18. A shortcoming
• “Further variables are
put between the
environment and the
essential variables
(the relay). The relay
thus never 'sees' the
environment directly.”
• “This picture must be
used if any severe test
of a reacting system
(artificial brain) is
to be applied.”
• This is like growing
a shell rather than
using intelligence.