Harvard Mark I

In the front hallway of Harvard’s glittering new billion dollar Science and Engineering Complex, behind glass and surrounded by hyper-modern architecture, one the earliest proto-computers sits occupying the space of two phone rooms. In a new space christened to push Harvard into the 22nd century, one selected artifact ties Harvard’s legacy to the present: the Harvard Mark I. 

First, this project will explore the history of the Harvard Mark I. Then, the project will evaluate the different chaine operatoire and materials used for the Harvard Mark I and the more recent Frontier supercomputer while reflecting on the impact of globalization. Finally, the project will interrogate the politics of the history of innovation by examining Harvard University and IBM as stakeholders while also paying attention to the elevation of women's contribution to the machine in curatorial design. 

The Harvard Mark I, also known as the IBM Automatic Sequence Controlled Calculator (ASCC), was the first operating machine that could execute long computations automatically (IBM). It was developed in the 1940s as the largest electromechanical calculator ever built and the first automatic digital calculator in the United States (IBM).

The development of the Harvard Mark I was led by Howard Aiken PHD ‘39, then a physics graduate student at Harvard University. Aiken had been working on the design of a large-scale computing machine since the 1930s, and in 1939 he received funding from IBM to build a prototype. In 1944, after alleged collaboration with Columbia University, it sat at 51 feet long, weighing 5 tons with 750,000 parts. An addition took a third of a second with multiplication taking 1 second (da Cruz).

While development was led by Professor Howard Aiken, the big breakthrough of the Mark I was incredible programmability. Its programmability stemmed from an extension of an architecture proposed by Charles Babbage almost 75 years prior and was applied by Grace Hopper, then an assistant to Aiken. Hopper graduated from Vassar in 1928 with a degree in mathematics and physics  and later received a doctorate in mathematics from Yale University (Yale). In 1943 she joined the United States Navy and was assigned to the Bureau of Ships Computation Project at Harvard University. She was responsible for developing the first compiler, a program that translates human-readable instructions into machine-readable code. This compiler, known as the A-0 system, was a crucial step forward in the development of programming languages and paved the way for modern languages like C++ and Java.

After decommission the computer was split up in 1959, with about half of the machine kept initially at the Harvard Science Center as part of the Collection of Historical Scientific Instruments, with the rest of the machine given to IBM and the Smithsonian (IBM). The Smithsonian is keeping the computer in archives, while the remainder of the machine is now on display at Harvard University in the Science and Engineering Complex.

Despite weighing 5 tons with 750,000 parts, the components of the Mark I were built and assembled in one facility in Endicott, New York. To contextualize the development process of the Mark I, I will compare it with the development process of the Frontier Supercomputer, the most powerful computer in the United States.

Howard Aiken, a graduate student in the Harvard Physics department, was struggling with calculating the numerical solutions to differential equations that had the combination of trigonometric and logarithmic functions. Aiken envisioned a precise “analytical machine” that, in his words, could “economize time and mental effort in arithmetical computation” (Aiken and Hopper). Inspired by the ideas of Charles Babbage in the 1800s and convinced that the electromechanical technology of the late 1930s could finally compute numerically solutions to generally complex systems he took his idea to IBM in November of 1937 (IBM). IBM engineers thought the idea held promise and encouraged Aiken to refine his designs. By January of 1938, Aiken brought his designs to Clair Lake and Frank Hamilton at IBM's Endicott Laboratory. While Aiken wanted to discuss "the possibilities of designing a machine of a very complex nature for solving equations pertaining to the ionosphere to work on television and also equations pertaining to astronomy", Lake and Hamilton actually guide the infrastructure of the computer towards what they consider much more feasible (IBM). By September 1938, the machine was estimated to cost $100,000 (about $2 million in 2022). Over the next 3 years, the majority of the effort dedicated to the machine was expended by unnamed engineers at Endicott as they designed, tested and assembled the machine with Aiken dropping in occasionally to receive status updates and teach some IBM math classes (IBM). The Navy began supporting the project in 1943, with IBM delivering the completed machine to Harvard in 1944. For the duration of the war the Navy exclusively used the Mark I for calculations. It was also in 1944 when Grace Hopper was assigned to the Mark I (Yale).

From the perspective of manufacturing and computer development the vast majority of the work in the written record (composed of memos kept in the IBM archives) was conducted at Endicott. In this period, most large manufacturing companies were involved in the war effort regardless of the businesses specialization. However, IBM emerged as the leader in mechanical tabulation machines (machines that could count) prior to the war and was well represented internationally with offices in Europe (IBM). After the war halted international commerce, IBM focused the majority of its manufacturing power to supporting the American war effort. However, in this period IBM continued to expand its fundamental businesses capacity opening new facilities in Poughkeepsie, New York, Washington, D.C., and San Jose, California in addition to its flagship factory in Endicott (IBM).

Despite this new expansion, IBM’s Endicott Engineering Laboratories, opened in 1904, was still the technical center of the company. The Laboratories were conveniently integrated with the Endicott factory, so parts could be made to order and transferred between them. Given the dynamics of World War II, steel for Endicott was probably sourced nearby in Pittsburgh, Pennsylvania. From a labor perspective, the Endicott IBM factory was an early mover towards salaried employees and group life insurance from as early as 1934.

This production process itself is irrelevant without proper contextualization and placing this history in conversation with modern methods of computer production. Unlike the Mark I, the development process of the Frontier Supercomputer (OLCF-5) was global. The Frontier, like the Mark I was, is the most powerful computer in the United States located in Tennessee. Costing 600 million dollars (300 times more than the Mark I), the Frontier was built by Hewlett Packard Enterprise in collaboration with AMD. 

Unlike the Mark I, the Frontier was completely designed by the Department of Energy (for nuclear physics simulations) then contracted out to be manufactured by HPE and AMD. While precise manufacturing details about the Frontier haven’t been released for national security reasons, the components list and details about past supercomputers built in the same frame are publicly available. HPE’s supercomputer manufacturing division, Cray, designed the structure and assembled Frontier in two locations: Chippewa Falls, Wisconsin and Bloomington, Minnesota.

However the components they used from the “brain” of the machine came from partner AMD and took a circuitous journey. The 2 components that drive the majority of the computational performance of the machine are the AMD Epyc CPUs and AMD Radeon Instinct MI250X GPUs of which there are 9,472 and 37,888 respectively. While the actual components were likely designed with a cross national R&D team largely based out of Santa Clara, California and Austin, Texas in the United States all manufacturing of the chips are outsourced to Taiwan Semiconductor Manufacturing Company (TSMC). As the operator of the largest chip foundries in the world, TSMC operates key factories and R&D centers globally. Given the common 6nm manufacturing process used in the CPU and GPU, it’s most likely the chips were produced at TSMC’s high volume Nanjing foundry in China. The Frontier has likely been touched by tens of thousands of people at different levels of the industrial process, from strategic product managers in the United States to administrative assistants in China. 

While the two supercomputers are most famous for their contributions to simulations of nuclear interactions, their chaine operatoire are incredibly different. The majority of this difference can be directly traced to new globalization at the end of World War 2 driven by American imperialism. Decision makers and businesses like IBM were given a fairly simple choice: pay salaried workers with health benefits to build parts that harm their health or outsource the work to countries with less developed labor protections. Businesses chose the latter and while it has caused an outstanding increase in production, their growth also has brought around substantial cultural differences. Cartographically, it’s easiest to see the air quality differences between Endicott of the Mark I and Nanjing of the Frontier.

For both the Mark I and Frontier, the primary consumer was the United States Federal Government. We can clearly trace that in the time since from the Mark I to now, the world has become more economically interdependent as key technological infrastructure is spanning across countries. In 1944, everything that composed a computer could be found within a couple hundred mile radius. Now these machines require coordination across time zones and languages. This change in the value chain has also caused significant changes in how innovation supports local communities. In 1950 Endicott had a population of 20 thousand people, today it has a population of 13 thousand. At a national population adjusted level, Endicott has shrunk by over 3.5 times. In the 1940s innovation allowed communities to spring up around it providing horizontal value as support for businesses that fed engineers and workers grew. Now innovation is confined to an office tour, where any chance at horizontal value being created is shifted thousands of miles away. Due to the legacy of colonialism and imperialism the locations where value is shifted, are susceptible to high degrees of corruption and often lack basic civil rights such that no value accrues to individuals in the ground. Instead of a technological development driving substantial positive externalities, our example shows how increasing globalization has often divorced innovation from local and immediate communities. 

I found the stakeholders of the Mark I asking important questions about who owns the history of a technological development. With a machine as powerful and influential as the Mark I, those who claim responsibility for the machine also claim the valence associated with transformative technology. In a modern era where conventional histories have been shaped by narratives around power leading to “just” conquest, groups that are seen to have produced exceptional tools are often seen to inevitably fall into power. Therefore, control of the narrative of innovation has exceptional consequences on the scale of civilization and barbarism. The US Government is the most direct beneficiary of ownership of the Mark I. The narrative of inevitably surrounding the eventual nuclear bombing of Nagasaki, conclusion of World War II and ensuing US global hegemony fits quite nicely around the superior technical capabilities of the machine that charted naval paths and simulated atomic weapons.

Perhaps differently than other objects, the Mark I reflects unique stakeholder interests from groups that are often overlooked in archaeological analysis. In particular Harvard University, IBM and women scientists. 

The Mark I’s ambiguous naming (Harvard Mark I and IBM ASCC) points to the first two key stakeholders to consider: Harvard University and IBM. After decommission in 1959, half of the machine was kept by Harvard while the other half was given to IBM and the Smithsonian. The half of the machine given to the IBM and Smithsonian has been photographed but is kept out of sight in the Smithsonian Archives (it should be noted the Smithsonian incorrectly identifies the place made as Harvard University). The IBM digital archives related to the ASCC are entitled “IBM's ASCC introduction (a.k.a. The Harvard Mark I)”. This naming is consistent with the paper that inventor Howard Aiken and programmer Grace Hopper ultimately published entitled “The Automatic Sequence Controlled Calculator”. However, the totality of the curatorial design choices by the Harvard Collection of Historical Scientific Instruments have insisted on referring to the machine as the Harvard Mark I. The central question for both Harvard and IBM is “who is responsible for the production of this artifact?” This question could be answered for both parties by engaging in open theoretical stratigraphy that relies on excavated documents and ephemera (like laboratory notes) from both IBM and Harvard facilities. Putting these documents in conversation could trigger new discoveries about the development process within a systematic framework.

IBM interests in the ASCC extend as part of building the historical framework of the more than century long commitment to tabulation that undergirds IBM market position today. The IBM archives serve to categorize intellectual property owned by the corporation for the purpose of demonstrating the enduring value IBM has added to the world. This work is intimately tied to a public relations campaign where IBM weaponizes history to support share prices, by currying favor with the public. At the same time, these IBM histories often forgo individual contribution to machines in favor of large corporate groups. Changing the common name of the Harvard Mark I to the IBM ASCC would directly allow IBM to claim more ownership around early contributions to computing. Importantly it would disassociate the development of the computers from academia (which is less subject to market forces) and place innovation firmly in the hands of corporations and the capitalist elite. 

Harvard’s interest in owning the history of the Mark I is evident in curatorial decisions since 1959. Harvard immediately displayed the machine when the Harvard Science Center opened in 1973. When the new Science and Engineering Complex opened in 2021 the Mark I was front and center. It’s no coincidence that the name of both locations that have housed the Mark I start with science. Curators are trying to draw a direct link between the caliber of scientific exploration that the Harvard faculty is pursuing today to the heights of progress that the Mark I represents. This is incredibly important as Harvard embarks on building an engineering school for the 22nd century. Engaging in anecdotal conversations with engineers and prospective students at Harvard and elsewhere, pieces of history like the Mark I can have a substantial impact in understanding the legacy of an engineering school. In fact , many students have been influenced in their decision making by the apparent legacy of world class contributions to science and technology Harvard has made. 

Many statements made in the Harvard Mark I exhibit are very hard to justify in the written or material record. For example Harvard asserts responsibility for the production of the artifact through statements like: “Mark I was originally called the “Automatic Sequence Controlled Calculator” by IBM, and often referred to as the “Harvard Calculator” when first installed in Cambridge in 1944. It started to be known as Mark I as its successor machines were built” (Collection of Historical Scientific Intruments). There is no systematic store of information that backs this assertion up and the University provides no evidence for the claim. If anything an inscription on the machine clearly says ASCC. Absent written records, most archaeologists might conclude that since an inscription forged into the steel case of an object was made it is related to the naming of that object. 

There is a dearth of written journals from Howard Aitken or his professors in the Physics department that could support many of Harvard’s claims about the faculty's contribution to the development of the machine. This stands in contrast to the plentiful records at IBM where notes from almost every meeting about the ASCC have been meticulously archived. In archaeological research, a multiplicity of sources about the same event are welcome as they can help elevate voices previously left out of discourse. A discovery or donation of ephemera related to Harvard faculty, students or staff that worked on the Mark I would have a high probability of substantiating claims made by either IBM or Harvard on the nature of the production of the Mark I.

Women scientists are a shareholder group whose interests are less tangled in the IBM-Harvard web. Grace Hopper’s work on the Mark I was groundbreaking and was recognized contemporaneously as she was quickly promoted in the Navy to Rear Admiral. Her work on the Mark I led her to be a key figure in the development of the COBOL programming language which is essential to modern life. Her achievements cannot be understated and enumerated because her contributions pervade every aspect of society. She won many prizes including the National Medal of Technology and the Presidential Medal of Freedom.  Importantly, Hopper was an advocate for women in the computer science field. As a result, there are dozens of prizes named after her with a focus on elevating women’s voices in science and engineering (Yale). Interestingly, Hopper’s education and success was much less unusual for her generation than a generation after. 15 percent of the total number of American math doctorates in the 1930s were women. Just 20 years later that number plummeted to 4%. In 2020 that figure stands at just 25%. 

A key aspect of the new Mark I exhibit in the Science and Engineering Complex is the elevation of Grace Hopper’s story as the author of Mark I manual. This was a previously un-excavated story that was raised as engineering schools made a push for gender parity in mathematics and computer science. Despite the noble goals, this curatorial elevation can be problematized as Harvard is co-opting the story of a woman scientist who never received formal funding or financial support from the University. Grace Hopper was on assignment from the US Navy and posted at Harvard while Harvard still only awarded degrees to men. 

An important question to ask might be “where are the other stories of women in science at Harvard?” Given the lack of women at Harvard we might specify, “where are the other stories of women in science at Radcliffe?” Clearly, there were other women of Hopper’s generation also doing substantial work in mathematics and computer science across the country and elevating their stories should be important work. I would propose research into Hopper’s contemporaries at Radcliffe including folks like: Dr. Payne-Gaposchkin who was the first to propose the accurate chemical composition of stars, Martha Thomas who established some of the first phosphor plants in the United States or Dr. Sears who served in World War II and founding editor of the journal Deep Sea Research and Progress in Oceanography.

Understanding the lives and contributions of these women might give more opportunities for young women to connect with scientists and engineers who leave important but under-emphasized legacies. 

Wading through the theoretical stratigraphy that contextualizes the Harvard Mark I has been extremely gratifying. I revealed some strings that tug at how we construct narratives around progress. In the historical era, the emphasis on written sources can dominate conversations. However in this case despite overwhelming written evidence suggesting a path production for the Mark I that lies with IBM, persistent curatorial decision making by Harvard University elevated Harvard’s voice. Whereas the Smithsonian has kept the Mark I in a warehouse, Harvard’s insistence in making the Mark I the centerpiece of display has elevated the object and privileged the story that aligns with Harvard’s interest. 

I also learned that this power of curatorial preference can be used in targeted ways to balance inequities and champion for change. The focus on Grace Hopper’s legacy has inspired women engineers that I know. Being conscious of the power of curation, has allowed exhibit designers to make a profound impact in the world around them.

Finally, comparing the Mark I with the Frontier supercomputer allowed me to think about the married impact of innovation and globalization in a new light. The transformation in material process and looking at the environmental impact on different communities shows that innovation without ethics can cause feedback cycles that embed inequities between populations. Tracing the legacy of computation and manufacturing from World War 2 to today, revealed how the US government engages in policies that accelerate these themes. At the same time, it also reaffirms the miracle of our contemporary global economy. So many people and countries are interdependent and inter-reliant on one another to make technological progress.

Walking past the Mark I every morning on my way to class, I should feel inspired and also reminded about how the technology I build impacts underserved communities.