What Baby Can Still Teach Us

Going to see Baby at Manchester’s Science and Industry Museum is a useful corrective to the way computers now present themselves. Modern computing prefers a smooth surface. The difficult parts happen elsewhere, behind the screen, beyond the network or inside something nobody expects you to open.

Baby offers rather less concealment. The machinery occupies space and makes no serious attempt to resemble furniture. It looks like something people had to persuade into working, which is closer to the truth about computers than their manufacturers generally prefer to suggest.

The museum’s machine is a working replica, built in 1998. The original Manchester Baby first ran a program on 21 June 1948. Between those facts lies a story about invention; standing before the reconstruction raises another question. How much of what made the achievement possible have we learned to overlook?

A small machine with a precise ambition

Its formal name was the Small-Scale Experimental Machine. “Small-scale” is doing historical work there. It was small in relation to the problem its designers were investigating, rather than something you might reasonably clear a corner of the kitchen table for.

Baby was built at the University of Manchester by a team centred on Frederic Williams, Tom Kilburn and Geoff Tootill. Its particular achievement was running instructions held in electronic, writable memory.

That qualification matters. Other calculating machines existed. Computing history becomes less useful when every important machine must win the same competition to be first.

The University of Manchester’s technical history explains the practical distinction: instructions could be fetched from memory, and a different program could be loaded without rebuilding the machine’s circuitry.

The hardware could remain while its behaviour changed. We now take that arrangement so thoroughly for granted that we mostly notice it when the behaviour changes without our permission.

First, make it remember

Baby’s immediate purpose was narrower than creating the modern world. It was built to test the Williams–Kilburn tube, a form of electronic memory using a cathode-ray tube.

As Manchester’s account of the subsequent Mark 1 explains, the experiment subjected the memory to a demanding test of speed and reliability. Running a program demonstrated that the necessary components could work together.

This is an appealingly concrete ambition. Before imagining every possible application, establish that the machine can reliably retain and retrieve what it needs.

The lesson is not that ambitious thinking is foolish. It is that an ambitious consequence can begin with a carefully bounded question. A working answer to one difficult problem may contribute more than an elaborate description of everything that will become possible once somebody else has solved it.

Baby did not need to describe itself as a platform. It needed the memory to work.

The calculation was not the point

The machine’s early test program searched for the highest proper factor of a number. Manchester’s historical account describes tests progressing to 262,144, with the calculation taking fifty-two minutes.

That number is a power of two. There are quicker ways for a person to identify its largest proper factor than constructing an electronic computer and waiting for most of an hour.

But the inefficient route was useful precisely because it exercised the machine. The result mattered as evidence that the process worked, rather than as an answer humanity had been impatiently awaiting.

That distinction is worth keeping. A demonstration should establish the capability under investigation. It need not pretend that its immediate output has transformed somebody’s business.

In modern terms, Baby’s test was not a particularly impressive productivity story. Its significance lay in proving that a different kind of productivity would become possible.

Memory you can almost count

Baby’s main store held thirty-two words of thirty-two bits: 1,024 bits altogether, equivalent to 128 eight-bit bytes. That is a quantity small enough to contemplate without an analogy involving several million football pitches.

The comparison with a modern device is inevitably absurd. But simply announcing that a phone is vastly more powerful misses the interesting part. The storage limit was a working condition the designers had to understand, not a number available for a nostalgic caption.

Every instruction occupied part of a very limited resource. Memory was not an invisible background assumption. It was central to what could happen next.

There is no virtue in recreating that scarcity unnecessarily. Modern abundance makes extraordinary work possible. Nobody needs to surrender useful tools to prove that they could have survived with valves.

Still, seeing such explicit limits raises a reasonable question about the software we build with far fewer of them. Are we using the additional capacity to accomplish more, or merely becoming less curious about where it goes?

The replica is another achievement

The machine in the museum also tells a later story. According to the Science and Industry Museum, the 1998 replica used vintage electronic components and guidance from the original designers.

Reconstruction was not equivalent to finding an old object, polishing it and writing a label. A working machine had to be produced from historical knowledge and physical parts.

The Computer Conservation Society credits Chris Burton with initiating and leading the project, followed by work maintaining and demonstrating the replica. The achievement includes the continuing effort required to keep the reconstruction intelligible and operational.

That is a form of preservation we can underestimate. Saving a description and retaining the ability to make something work are different undertakings.

A photograph preserves appearance. A running reconstruction can reveal sequence, constraint and failure. It asks the people caring for it to understand more than where the cables belong.

Progress has made the mechanism harder to see

The replica makes the physical character of computing difficult to ignore. Information requires a material arrangement. Instructions must be represented, retained and acted upon. Nothing becomes less physical because we call it software.

Modern abstraction is immensely useful. I do not want to investigate semiconductor behaviour every time I save a document. The ability to work without understanding every underlying layer is part of what makes computing valuable.

But convenience can encourage a peculiar forgetfulness. We speak about digital services as though they float above the world, then express surprise when they depend on electricity, equipment and people who know how to maintain them.

Baby is too substantial to sustain that illusion. It reminds us that computation has always involved making physical things behave dependably enough for an abstraction to become useful.

The neat interface is an achievement. So is everything it has successfully persuaded us not to notice.

The useful kind of humility

There is a familiar museum response in which an old machine becomes evidence that its makers were heroic and we have become frivolous. Baby deserves better than being recruited into an argument against whatever young people currently do with computers.

Its designers were solving a problem with the tools and knowledge available to them. Our responsibility is not to imitate their constraints. It is to bring comparable seriousness to our own.

That means knowing what a test demonstrates, distinguishing a proposal from a working system and respecting the expertise required to keep either understandable. It also means accepting that consequential work may look modest before its consequences arrive.

What I value about going to see Baby is the adjustment in scale. The machine is physically large, its immediate task narrow and its historical consequence immense. Those proportions are almost the reverse of a contemporary technology announcement.

Back outside, a phone can offer maps, messages and more computation than Baby’s designers had at their disposal. It may also ask for a moment while something loads. The machinery has become smaller. The need to make it work has not.