cyberneticzoo.com

Internals showing N-S, E-W carriage, Relays, Uni-selector, motors, amongst other electrical components. 

See 18 mins 51 secs in for 27 seconds.

See 9 mins 16 secs in for 32 seconds.

As the 1952 maze solver was recently at the MIT Museum.

Picture from Life Magazine 28 July 1952. Top trace is showing the first pass of the maze solver learning the maze. The second run showing that it has learnt the maze and the mouse goes direct to the cheese.

Detail of a trace showing to mouse rotations and making contact with the wall.

Picture above from Popular Science March 1952 showing another pair of  time-lapse photos showing the learning of the maze in the first run, and the solving of the maze.  A modified mouse is also shown. It included a lamp to ensure a trace showed in the time-lapse photography. Full pdf here.

The above maze photograph from Electrical Engineering July 1952. It took two minutes to learn the maze, and between 12-15 seconds to reach the "cheese" once solved.

Problem-Solving Electric Mouse Aids in Improved Telephone Equipment Research

An electric mouse with a man-made super-memory is busily at work these days, repeatedly threading its way through a series of complicated mazes at Bell Telephone Laboratories. The handiwork of Dr. C. E. Shannon, a mathematician associated with the Bell Telephone Laboratories, Inc., the mouse uses for its "brain" some of the same kind of switching relays found in dial telephone systems. The reason it exists is to provide fundamental knowledge which will help improve telephone service.
The mouse, in reality a 2-inch bar magnet with three wheels and copper whiskers, can solve quickly more than a million million different mazes, learning each new one rapidly, then instantly forgetting it in order to be ready to learn the next one. Its goal is an electric terminal with a bell which rings when the mouse nudges it with its copper whiskers.
The maze is about half the size of a desk top. It has aluminum fences which can be rearranged at will in 40 different slots to create the hardest possible problems for the mouse. The mouse is placed at some arbitrary point in the maze and the goal at a different arbitrary point. After a brief pause to get its bearings, the mouse goes up and down corridors, bumping into walls, backing up and turning, and exploring until, a minute or two later, it reaches its goal and rings the bell.
Having learned the correct path to the goal, the mouse now can be set down at any point that it visited during its explorations and, without making a single false move, it will proceed directly to the goal in 12 to 15 seconds. If it is placed in a part of the maze not previously visited, it will explore until it reaches a known part and then move directly to the goal.
After this, if the maze is altered, the mouse will have to learn the new paths by further exploration, but it readily will remember those parts of the path which remain unchanged.
This is the way the mouse works. When it is set down on the metal floor of the maze, it trips an electric switch which signals its position to a mechanism under the floor. A motor-driven electromagnet moves swiftly to the spot directly beneath the mouse and from then on holds it in a magnetic grasp. The magnet turns through a 90-degree angle, carrying the mouse with it, then guiding it forward. If the mouse hits a barrier and detects, by means of its copper whiskers, that it is in a dead end, the magnet will back away, shift the mouse to another direction, and start it forward to try again to find an open path. It keeps trying until it finds the way to the goal. Then it remembers the successful path and can solve the maze directly without error.
To regulate the sequence of movement, a "programming" circuit has been built, consisting of 40 electric relays. Another part of the mouse's "brain," which serves as its memory, contains 50 relays. Two small motors complete the equipment.
By working with such problem-solving equipment, it is hoped that more will be learned about what man can do with machines. Many of the techniques by which machines are able to remember are currently being applied in the Bell System in dial switching, in automatic accounting, and in other equipment.
The real significance of this mouse and maze, lies in the four unusual operations it is able to perform. It has the ability to solve a problem by trial and error means, remember a solution and apply it when necessary at a later date, add new information to the solution already remembered, and forget one solution and learn a new one when the problem is changed.

The above two sequences are interesting in that the 'learnt' maze is altered (2nd panel before the finish), and the mouse is still capable of re-learning the change and solving the maze.

Shannon with the mouse.

The original mouse was carved from wood hollowed out to take a two-inch magnet bar of aluminium, nickel, and cobalt. It has two beady, button eyes, three small brass wheels for legs, and an pipe cleaner for a tail. Two copper whickers guide it through the maze to the "cheese" which is an electrical terminal that rings a bell when toughed by the whickers.

Bell built several versions of Theseus for demonstrations of the technology. One of them was known as Philbert as used by Southwestern Bell Telephone Company.  As late as November 1976 they were still being demonstrated.

Time-Life have about 70 images of Shannon, the mouse, and time-exposures of the maze. They can also be found in Google images by adding the option source:life .

The Thomas Ross Maze Learning Machine showing its feeler tracking the slots of this comb-shaped maze.

See complete Scientific American 1933 article titled "Machines That Think" – pdf here.

In 1951 Marvin Minsky teamed with Dean Edmonds build the first artificial neural network that simulated a rat finding its way through a maze.

They designed the first (40 neuron) neurocomputer, SNARC (Stochastic Neural Analog Reinforcement Computer), with synapses that adjusted their weights (measures of synaptic permeabilities) according to the success of performing a specified task (Hebbian learning) The machine was built of tubes, motors, and clutches, and it successfully modeled the behavior of a rat in a maze searching for food.

As a student, Minsky had dreamed of producing machines which could learn by providing them with memory "neurones" connected to "synapses"; the machine would also have to possess past memory in order to function efficiently when faced with different situations.

In 1951 the "machine" was born, consisting of a labyrinth of valves, small motors, gears and wires linking up the various "neurones". Some of these wires were connected up at random to the various memory banks in order to achieve a degree of causality of events. The reason such a machine had been put together was to try and find the exit from a maze where the machine would play the part of a rat whose progress would be monitored on a light network.

When the system was completed it was possible to follow all the movements of the 'rat' within the maze and it was only through a design fault that it was found more than one 'rat' could be introduced which would then interact together. After various casual attempts the rats started 'thinking' on a logical basis helped along by reinforcement of correct choices made and the more advanced rats would then be followed by the ones left behind. This first practical example, built by Minsky with the help of Dean Edmonds, also included numerous casual connections between its various 'neurones', acting like a sort of nervous system able to overcome any eventual information interruption due to one of the neurones failing.

Image courtesy Gregory Loan:

Gregory visited Marvin Minsky and enquired about what happened to his maze-solving computer. Minsky replied that it was lent to some Dartmouth students and it was disassembled. However, he had one "neuron" left, and Gregory took a photo of it.

An extract from an interview Jeremy Bernstein did with Marvin Minsky – The New Yorker, Dec 14, 1981

p69
For a while, I studied topology, and then I ran into a young graduate student in physics named Dean Edmonds, who was a whiz at electronics. We began to build vacuum-tube circuits that did all sorts of things."
As an undergraduate, Minsky had begun to imagine building an electronic machine that could learn. He had become fascinated by a paper that had been written, in 1943, by Warren S. McCulloch, a neurophysiologist, and Walter Pitts, a mathematical prodigy. In this paper, McCulloch and Pitts created an abstract model of the brain cells—the neurons—and showed how they might be connected to carry out mental processes such as learning. Minsky now thought that the time might be ripe to try to create such a machine. "I told Edmonds that I thought it might be too hard to build," he said. "The one I then envisioned would have needed a lot of memory circuits. There would be electronic neurons connected by synapses that would determine when the neurons fired. The synapses would have various probabilities for conducting. But to reinforce 'success' one would have to have a way of changing these probabilities. There would have to be loops and cycles in the circuits so that the machine could remember traces of its past and adjust its behavior. I thought that if I could ever build such a machine I might get it to learn to run mazes through its electronics— like rats or something. I didn't think that it would be very intelligent. I thought it would work pretty well with about forty neurons. Edmonds and I worked out some circuits so that —in principle, at least—we could realize each of these neurons with just six vacuum tubes and a motor."
Minsky told George Miller, at Harvard, about the prospective design. "He said, 'Why don't we just try it?' " Minsky recalled. "He had a lot of faith in me, which I appreciated. Somehow, he managed to get a couple of thousand dollars from the Office of Naval Research, and in the summer of 1951 Dean Edmonds and I went up to Harvard and built our machine. It had three hundred tubes and a lot of motors. It needed some automatic electric clutches, which we machined ourselves. The memory of the machine as stored in the positions of its control knobs—forty of them—and when the machine was learning it used the clutches to adjust its own knobs. We used a surplus gyropilot from a B-24 bomber to move the clutches."
Minsky's machine was certainly one of the first electronic learning machines, and perhaps the very first one. In addition to its neurons and synapses and its internal memory loops, many of the networks were wired at random, so that it was impossible to predict what it would do. A "rat" would be created at some point in the network and would then set out to learn a path to some specified end point. First, it would proceed randomly, and then correct choices would be reinforced by making it easier for the machine to make this choice again—to increase the probability of its doing so. There was an arrangement of lights that allowed observers to follow the progress of the rat—or rats. "It turned out that because of an electronic accident in our design we could put two or three rats in the same maze and follow them all," Minsky told me. "The rats actually interacted with one another. If one of them found a good path, the others would tend to follow it. We sort of quit science for a while to watch the machine. We were amazed that it could have several activities going on at once in its little nervous system. Because of the random wiring, it had a sort of fail-safe characteristic. If one of the neurons wasn't working, it wouldn't make much of a difference —and, with nearly three hundred tubes and the thousands of connections we had soldered, there would usually be something wrong somewhere. In those days, even a radio set with twenty tubes tended to fail a lot. I don't think we ever debugged our machine completely, but that didn't matter. By having this crazy random design, it was almost sure to work, no matter how you built it."
Minsky went on, "My Harvard machine was basically Skinnerian, although Skinner, with whom I talked a great deal while I was building it, was never much interested in it. The unrewarded behavior of my machine was more or less random. This limited its learning ability. It could never formulate a plan. The next idea I had, which I worked on for my doctoral thesis, was to give the network a second memory, which remembered after a response what the stimulus had been. This enabled one to bring in the idea of prediction. If the machine or animal is confronted with a new situation, it can search its memory to see what would happen if it reacted in certain ways. If, say, there was an unpleasant association with a certain stimulus, then the machine could choose a different response. I had the naive idea that if one could build a big enough network, with enough memory loops, it might get lucky and acquire the ability to envision things in its head. This became a field of study later. It was called self-organizing random networks. Even today, I still get letters from young students who say, 'Why are you people trying to program intelligence? Why don't you try to find a way to build a nervous system that will just spontaneously create it?' Finally, I decided that either this was a bad idea or it would take thousands or millions of neurons to make it work, and I couldn't afford to try to build a machine like that."
I asked Minsky why it had not occurred to him to use a computer to simulate his machine. By this time, the first electronic digital computer— named ENIAC, for "electronic numerical integrator and calculator"—had been built, at the University of Pennsylvania's Moore School of Electrical Engineering; and the mathematician John von Neumann was completing work on a computer, the prototype of many present-day computers, at the Institute for Advanced Study. "I knew a little bit about computers," Minsky answered. "At Harvard, I had even taken a course with Howard Aiken"—one of the first computer designers. "Aiken had built an electromechanical machine in the early forties. It had only about a hundred memory registers, and even von Neumann's machine had only a thousand. On the one hand, I was afraid of the complexity of these machines. On the other hand, I thought that they weren't big enough to do anything interesting in the way of learning. In any case, I did my thesis on ideas about how the nervous system might learn.

To date, I have not been able to locate a diagram of SNARC.  It's possibly in Minsky's thesis.