cyberneticzoo.com

Self-recognition and the Mirror Dance

[Image source: An Imitation of Life,  Scientific American, May 1950, p42-45.]

7 . Self-recognition. The machines are fitted with a small flash-lamp bulb in the head which is turned off automatically whenever the photo-cell receives an adequate light signal. When a mirror or white surface is encountered the reflected light from the head-lamp is sufficient to operate the circuit controlling the robot's response to light, so that the machine makes for its own reflection; but as it does so, the light is extinguished, which means that the stimulus is cut off — but removal of the stimulus restores the light, which is again seen as a stimulus, and so on. The creature therefore lingers before a mirror, flickering, twittering, and jigging like a clumsy Narcissus. The behaviour of a creature thus engaged with its own reflection is quite specific, and on a purely empirical basis, if it were observed in an animal, might be accepted as evidence of some degree of self-awareness. In this way the machine is superior to many quite 'high' animals who usually treat their reflection as if it were another animal, if they accept it at all.

Source: p115, W. Grey Walter, The Living Brain, 1953 -see chapter 5 – Totems, Toys, and Tools

What can be seen or determined in the photo of Elsie below?  

1. tracer candle visibility;

2. low batteries (because it enters the hutch which is strategically placed to the right of the mirror).

Figure 7. Elsie performs in front of a mirror, but is probably responding to the candlelight rather than to her pilot light. [RH 2010 -Most earlier comments by others are of this rather un-clear image of the so-called 'mirror dance'.]

Prior to the release of the clearer Life image of Elsie performing the 'mirror dance' (see pic below) Holland in "Legacy of Grey Walter" describes it as follows:

Recognition of self
A pilot light is included in the scanning circuit in such a way that the headlamp is extinguished whenever another source of light is encountered. If, however, this other source happens to be a reflection of the headlamp itself in a mirror, the light is extinguished as soon as it is perceived and being no longer perceived, the light is again illuminated, and so forth. This situation sets up a feedback circuit of which the environment is a part, and in consequence the creature performs a characteristic dance which, since it appears always and only in this situation, may be regarded formally as being diagnostic of self-recognition. This suggests the hypothesis that recognition of self may depend upon perception of one’s effect upon the environment.

The below from Discussions on Child Development,  1971, see Book II 1954-56 p35-6.

p35.

With Fig. 6 we come to some of the refinements which emerged only some time after these creatures had been made. This mode of behaviour and the next one were, quite frankly, surprising to us though, of course, we ought to have been able to predict them. Fig. 6 illustrates the situation when a creature of this type is confronted by its reflection in a mirror. It has on its nose a small pilot light, put in originally to tell us what was happening inside; it is so arranged
p36.
that it is turned off when the creature sees another light; that is, it tells us when the photo-tropistic mechanism is in operation.
In this case, the light which the creature was allowed to see was its own pilot light in the mirror. In this situation, the act of 'seeing' it makes it automatically extinguish the light which it sees. The apparent stimulus light having been extinguished, it turns it on again, then off and so on, so that you get a characteristic oscillation. You can see how peculiar and regular it is by the zigzag going up the side of the mirror. This is an absolutely characteristic mode of behaviour, which is seen always and only when the creature is responding to its own reflection. This is an example of the situation I described in the second proposition, where the reflexive circuit includes an environmental operator; in such a situation you get a characteristic mode of behaviour which occurs always and only when the model is reacting to itself.

Narcissism

“The creature therefore lingers before a mirror, flickering, twittering and jigging like a clumsy Narcissus” (Grey Walter, 1963, p. 115). Grey Walter interpreted this famous mirror dance as evidence of self-recognition.

The drawing of the famous `mirror dance’ in `An imitation of life’ [from Scientific American] is nothing like the regular alternation between the tortoise's approach and avoidance as shown in the photograph, being an altogether more irregular and complex trajectory. There may well have been a mirror dance that could have been argued to be a form of self-recognition, but unfortunately this photograph cannot be said to be a record of it. The brightest light visible to the camera, and presumably to the photocell, is the candle on the tortoise’s back and its reflection in the mirror. The trace is far more likely to reflect the alternation of behaviour pattern P (approach to the reflected candlelight) with behaviour pattern O (obstacle avoidance on contact with the mirror). We can be sure that Walter used this image as an example of the mirror dance because it appears in the form of a diagram in the transcript of a talk he gave in 1954 (Walter 1956b); the text matches closely the account given in `Accomplishments of an artefact’. Interestingly, the description of the mirror dance in de Latil’s book also matches this photograph rather than Grey Walter’s original description and Bernarda Bryson Shahn’s sketch.

For most people, with regards to the image above (see figure 7), one could hardly refer to this behaviour as "flickering, twittering and jigging like a clumsy Narcissus". However, you could do so to the above illustration by Bernarda Bryson (partner and later married to the artist Ben Shahn), as illustrated in Scientific American (Walter, W. Grey, "An Imitation of Life," Scientific American, May 1950, p42-45.). The above illustration is actually of Elmer, and not Elsie as is the below photo. This also gives more credence to Grey's use of the word Narcissus, being the son of a Greek god who became obsessed by his own image. [Elmer scans clockwise, the opposite of Elsie and the bump aviodance traverse therefore is from right to left. see here.]

[Narcissus : In Greek mythology, a beautiful youth who fell in love with his own reflection. He was the son of the river god Cephissus and the nymph Leriope. His mother was told by a seer that he would have a long life, provided he never saw his own reflection. His callous rejection of the nymph Echo or of his lover Ameinias drew upon him the gods' vengeance: he fell in love with his own image in the waters of a spring and wasted away. The narcissus flower sprang up where he died.]

Although Elmer was then long gone, Grey Walter continued to use this more interesting description of self-recognition along with the below image, although it didn't and couldn't match with the sometimes erratic behaviour of the original tortoise, Elmer  and could no longer be reproduced with the newer models.

In my opinion, in the cycloidal trace seen above, the 'bottom' of the cycloid appears flattened and bright spots at the start of the cycloid 'flats' appear. To me, this is indicative of 'bump' avoidance behaviour, not self-recognition.  When I visited the Bristol Robot Labs in 2009 to see the replica tortoises, the comment was passed to me that they were unable to satisfactorily reproduce the self-recognition behaviour as described by Grey Walter.

A relatively recent , clearer image of the so-called 'mirror dance' as released by Life Magazine.

A comment on Time-Lapse Photographs in General:
In interpreting all the time-lapse photographs, there are several aspects to keep in mind.
As already mentioned in pervious posts on the Tortoises, the cycloidal gait makes Elsie traverse to the right as her scanner turns in a counter-clockwise direction. Elmer, on the other hand, scans clockwise and because of the trailing action of the rear-wheels, will veer to the left. I must say, though, that the illustrations suggest that with no light source to track towards, Elmer tends to move in a forward direction and not sideways.
Most of the pictures show Elsie heading towards a light, either a candle or the hutch light, sometime a light out of sight near the camera.
Where you see two identical Elsies, it is actually the photographer’s technique of photographing Elsie at the start of the run, then  Elsie at the end of the run. There are not two separate Tortoises except where they look physically different i.e. Elmer has the ‘scaled’ Bakelite sheeting shell. The single trajectory is also an indicator of only a single Tortoise being traced.

Notice also that the flame of the target candle is placed at the same height as the PEC (the Photo-Electric Cell) in the scanning turret. 

Hobby Electronics magazine (November 1979-January 1980) describes an autonomous robot.

HEBOT is a free roaming robot which can negotiate obstacles, steer towards a light (infra red) and follow a wire (A.C. current) around your home.

HEBOT emits a squeak when it detects light or following a collision. Control is transferred to the wire following circuitry unless HEBOT encounters an abstacle.

HEBOT also has a microphone, and can be wired to go into reverse, spin, or just stop.

HEBOT's 'random walking' is a bit of a misnomer, as it actually executes a series of spirals.

There is room for development. Board one will support a further four levels of control. It also has a simple constant-current nicad charger for its recharging 'nest'.

Click image for pdf.

Click image for pdf.

Click image for pdf.

The QMW Mk. IV Mobile Research Robot

The final version (Mk. 4) of our first attempts at a mobile robot for machine learning research. It had a somewhat unusual (and not entirely satisfactory) drive layout with wheels at the front of the vehicle. D.C. motor drive was complemented with gray-code shaft encoders. Forward facing sensors included sonar rangefinder, colour photo-receptors and a 32×32 binary camera. Other sensors were all round touch bars and battery level monitoring. The contacts at the top of the front face allowed the vehicle to nestle against a charger pad. This robot had no on-board computing(!), relying on communications via an umbilical wire to a DEC PDP-11. The machine was programmed to give a plausible emulation of the Pavlovian conditioned reflex; and was used extensively by D.H. Mott in his thesis "Sensory-motor Learning in a Mobile Robot", an investigation into Piagetian forms of learning. Despite its simplicity this robot made several television appearances and enjoyed considerable press coverage.

Text and image from here.

Computer scientists Mark Witkowski, left, and Dave Mott at the Artificial Intelligence Laboratory of Queen Mary College in London's Mile End Road, with a prototype (foreground) of the robot on which the CMC group are working which will be capable of learning for itself and making its own decisions. (AP Photo/Press Association) – image dated 23 February 1978.

photo caption: THINK PIECE—Mr Mark Witkowski and members of the team with Mr Cube, the thinking robot at Queen Mary College. [RH 2012 - In the background to the left of the camera you can see two bowed strips on the wall. These are the contacts used for the battery recharging.]

Evening Standard Tuesday February 21 1978 p7.

'Mr Cube'- the robot that thinks …

Meet the new Mr Cube. He's shy and retiring and has only one eye. But he has one big asset–he's a robot which thinks.
The robot is the brainchild of computer scientists Mark Witkowski and David Mott of the Artificial Intelligence Labarotary at Queen Mary College, University of London.
It has a very primative intelligence which enabled it to avoid obstacles, but larger and stronger version which will work in hazardous environments or intolerable places such as mines or nuclear power stations are planned.
Nest
The 20-inch high plastic box, on wheels, bristling with sensors and flasing lights, trundles forward when activated and then returns to its "nest"– a battery charger which gives an electrical award.
Labarotary director Mr Alan Bond said: "Robots as we know them are not capable of dealing with unexpected emergencies.
"We are developing a robot which will be responsive to its environment and will learn from its own experience."
Sullen
When I first introduced myself to "Mr Cube" he remained sullen, silent and motionless.
"Speak a bit louder," said scientist David Mott.
I did and the robot emittted a mechanical squeak and backed away. "He tends to do that if you shout," said David.
He will also back away if you knock into him, responding to his battery-powered sensors.
He is due to make a public demonstration later this week and although he can at the moment "speak" only with morse code signals, with barking, the scientists the scientists plan a more sophicticated character.

See pdf here.

ENTROPY— BUILDING A ROBOT FROM SCRATCH
Gene Oldfield, began building his first major homebrew robot around 1976. Entropy, as it was called, was a mobile, three-wheeled robot powered by a car battery. A KIM single-board computer was interfaced to the sensors and relays by only seven microchips, which means that most of the processing was done in the computer itself. To give you some idea of the process a homebrewer goes through, we have outlined the steps in Entropy's development:
"I began," says Gene, "with the mechanical construction. Entropy's back wheels were on a common axle, which means that the turning center must lie somewhere on the extended axle. Rather than build one from scratch, I used the axle yoke and wheels from a discarded toy red wagon." The front wheel was motorized and attached by a vertical axle (called the scan). The electrical connections to the motorized wheel were commutated using generator brushes
from a car motor. "Had I wired the wheel and motor instead, the wires would have become twisted and eventually broken from all the turning required to steer the robot." A second motor with gearhead drove a gear on the scan axle along with a cam assembly that allowed the robot to be set in any direction.
"I made the frame of wood," Gene continues, "and painted, waxed, and covered it with copper foil." While not a traditional material for building robots, wood has advantages. It's an electrical insulator; light, strong, easily worked; and it's very simple to attach switches, wires, and terminal blocks as you go along. Also, wood is not as high tech and makes the robot acceptable, more like furniture. The visual impact is an an important design criterion. A robot should be friendly, fun, and nonthreatening. "As a homebrewer, you can get away with a lot and still make your creations legitimate. R2D2's popularity was due, in large part, to being cute."
Gene decided to wire Entropy on a protoboard (also called a breadboard). 'Since I was developing Entropy for the first time," he explains, "I did not know how it was go ing to be wired." The protoboard permits endless experimentation. Once power wires were in place, it took about four hours to connect Entropy's relays using patch cords (wires). "This is both fun and satisfying—you feel that you could emulate anyone's robot by simply changing a few wires." At that point, Entropy was in the same category as the mechanical/electrical rats and turtles of the fifties. It could center its scan axle and stop.
Next, Gene wired the motors, sensors, and relays to the electronic components that interfaced with the on-board computer. Again, because the pattern of connection was not established, he used a protoboard.
Entropy was to operate without external wires, hook-ups, or data links. Data from sensors was processed on board. While doing the dishes and other seemingly simple tasks were impossible, Entropy did have the sensors and memory to be effectively mobile. It could travel from room to room and navigate through doorways. Sonar, using a tone- decoder chip, measured multiple reflections on distances ranging from one inch up to fifteen feet.
The KIM computer on board allowed Entropy to be programmed for motion at over 100 instructions per minute. Operations were performed through the record/playback program, much the way you save, load, and run programs on your computer. In the playback mode, the computer sends out commands, which the robot acts upon.
Since Entropy lacked a tape recorder or disk drive, Gene used nonvolatile RAM. With a special CMOS RAM chip (which requires  very little electricity) and a couple of nicad batteries, he could keep the RAM powered up even when the rest of the system was turned off. The system is both cheap and reliable. Programs are always present. Simply turn on the robot, select a program's starting address in memory, and press go.
Entropy was a successful project. The robot wandered around the house, moving from room to room.

Above extract from the book "Everyone Can Build a Robot Book " by Gene Oldfield and Kendra Bonnet, 1984.

Entropy is on the right.

Source: Infoworld, 8 Nov 192.

Gene Oldfield had the original Robot Repair in Sacramento in the early 1970's.  He's more into Art robots and electric bikes these days, with the Horse Cow Art Collective. See Youtube video below.