Thorsen memory tubes

Thorsen memory tubes are electronic devices that record an operation demonstrated through manual control and then direct a machine to repeat it without ongoing human supervision. In Dan Davis’s account, they turn mechanical capacity into reusable learned routines: paired with search and safety circuits, they let machines handle household work, suggest a route toward a programmable automaton, and offer a basis for storing the sound codes of a spoken-language stenographer.

Operation and Machine Learning

The tubes function as a machine’s operational memory. An operator first guides a connected machine through a task; the tube retains the sequence and can direct the same operation again, repeatedly, without a human supervisor. Davis distinguishes this form of stored routine from human thought: for an automated machine tool, recorded operation is enough, but a more complex system needs additional circuits to respond when conditions do not exactly match the remembered sequence.

Those supplementary circuits search within specified limits for a relevant condition, then trigger the basic instruction stored in the tube. Davis calls them “judgment” circuits only provisionally: in his view, they do not confer genuine judgment, but act instead as hunting circuits that look for a programmed circumstance. A further circuit can interrupt the stored routine whenever the current cycle departs from the one originally recorded. The resulting system combines a fixed learned procedure with limited recognition of variation, without claiming that the machine understands what it is doing.

Flexible Frank and Domestic Tasks

This architecture makes Flexible Frank’s domestic competence scalable. Davis could demonstrate once how to clear a table, scrape dishes, and load a dishwasher; a Thorsen tube could then guide Frank through the same routine whenever he encountered dirty dishes. An electronically duplicated tube could instead let him manage dishes the first time he encountered them, while a separate tube could store the procedure for changing a wet baby. In Davis’s formulation, the machine’s reliability depends not on humanlike understanding but on a routine sufficiently specified for its purpose.

Frank’s square head could accommodate as many as a hundred tubes, each holding the electronic memory of a different household task. A guard circuit around the search or “judgment” circuits supplied a practical boundary: if Frank met a situation outside his instructions, he was to stop and call for help rather than risk damaging dishes or harming a baby. The tubes therefore promise breadth through accumulated routines, while the guard circuit acknowledges that no catalogue of routines can anticipate everything.

From Servant Machine to General-Purpose Automaton

The tubes also matter because Davis recognizes that Frank’s design is not the only possible use of the technology. Frank’s framework as a powered wheelchair and his role as a household servant constrain what he can do; Davis reasons that alternative design choices could make a machine more flexible, provided it still had access to Thorsen memory tubes. Since the tubes were commercially available to anyone designing a cybernetic sequence, Miles could not prevent their use in a different machine.

Davis imagines an all-purpose automaton programmable for anything a human being can do, so long as the task does not require true human judgment. That ambition rests on the tubes’ capacity to hold and execute complex instructions, supplemented by circuits that detect relevant conditions and interrupt a routine when needed. The tubes do not erase the distinction Davis draws between mechanical performance and human judgment; instead, they make that distinction the design problem around which his imagined robot is organized.

Speech Coding and Broader Uses

Later, Davis applies the same general principle of stored, reusable codes to a proposed stenographic machine. With memory tubes and advances in miniaturization, he estimates that a cubic foot could hold a hundred thousand sound codes, enough in principle to encode every word in a Webster’s Collegiate Dictionary. He considers that unnecessary for ordinary stenography: ten thousand words would suffice, with uncommon terms spelled out when needed and then entered into a client’s special vocabulary.

The proposed system would also sound-code punctuation and formats, retrieve addresses from a file, and take instructions about the number of copies and their routing. At least a thousand codes could be reserved for specialized business or professional vocabulary, with the owner able to add terms personally. The example extends the tubes’ significance beyond a robot’s physical tasks: a stored sequence can organize not only movement, but also the recognition and processing of spoken information.

Significance

Across these applications, the tubes stand for a practical model of machine competence: demonstrate or encode a procedure, store it, and repeat it under defined conditions. Their promise is both expansive and bounded. A hundred stored household routines, an adaptable robot, or a stenographer equipped with specialized vocabulary all become conceivable through memory and circuitry, yet each system still depends on instructions, limits, and provisions for cases it cannot handle. In Davis’s account, the tubes do not make a machine human; they make complex work reproducible without requiring a human mind.