Knowledge Preservation and Decay


Part 3 of the essay series Research and Knowledge Accumulation


Preserving knowledge

It is easy to believe that knowledge accumulates automatically. We see many instances of knowledge accumulation — in math and physics, on Wikipedia and the internet, in our own lives as we learn things and teach things to others.

However, knowledge accumulation is not automatic.

To see this, we can think about cases of lost knowledge — or cases where knowledge would have been lost, had we not worked very hard to preserve it. If we sometimes lose knowledge on a societal or civilizational scale, or if we manage to preserve our knowledge but only through substantial effort, it follows that knowledge accumulation is not automatic.

Let’s consider two cases, each that followed a different course. First, the Saturn V’s F-1 rocket engine. Second, Damascus steel.

The F-1 rocket engine

The Saturn V was the most powerful rocket ever launched. It was used thirteen times by NASA from 1967 to 1973, including in all of the missions that put people on the moon. Its first stage was powered by five F-1 rocket engines.

The Rocketdyne F-1 engine (Photo by Daniel Mennerich)

The F-1 engine itself was the most powerful single combustion chamber liquid-fueled engine ever deployed. To give a sense of this, a single F-1 weighs 18,500 pounds and burns more than 5,000 pounds of oxidizer and fuel each second.

Rocketdyne, the company that built the F-1, developed it from 1955 through 1965. It then started developing an upgraded engine, which was called the F-1A. Demand for F-1 engines declined though, production of the F-1 stopped, and the F-1A was never used.

Can we build the F-1 today?

People sometimes ask whether we can still build the F-1 today. The answer is a bit complex.

On one hand, we have a lot of information about the F-1. In 1969, Rocketdyne ran a program called the F-1 Production Knowledge Retention Program, with the goal of exhaustively documenting how the F-1 engine was made. They collected 20 volumes of written material and also tape-recorded interviews with the skilled engineers who worked on the project. The purpose of the interviews was to capture things the engineers knew that would not show up in the written documentation.

In addition to this, there are more than 25 original F-1 engines that never flew, as well as some that did fly that Jeff Bezos fished out of the Atlantic Ocean.

That’s a lot of information. Unfortunately though, it’s probably not enough. This is because of the confluence of two factors.

First, the F-1 engines were handcrafted. We did not have computer-controlled machine tools when the F-1s were produced. Instead, the F-1s were built by many very skilled engineers fitting thousands of pieces together with thousands of individual, complex welds.

Second, the fabrication industry has changed. In the time since the F-1s were built, friction stir welding and laser welding were invented, aluminum-lithium alloys and carbon composites became available, and computer-controlled machine tools that weld, mill, and bend metal to make more accurate parts were developed. There were other technological advances as well.

These advances led to a decline in demand for a particular type of highly skilled engineer — the exact sort who would have worked on the F-1.

In all probability, complete knowledge of how to build an F-1 engine does not exist in either the written documentation, the tape-recorded interviews, or the heads of skilled engineers. In this sense, we have forgotten how to build the F-1.

Two routes to recovery

All is not lost though, and for two reasons.

First, in 1992, Rocketdyne tried to figure out if it could still build F-1As. They concluded that they could. The price tag was hefty: $315 million (1991 dollars) to reboot production, and then $15 million per engine for a batch of 40 engines. $100 million of the $315 million would go to the production of test engines. $215 million would go to everything else, including labor. In 1992, Rocketdyne found that they still had 248 employees who had worked on the original F-1s and they had 76 who were willing to come out of retirement. Undoubtedly many new people would need to be trained.

It’s almost 30 years later. Many of those skilled engineers will have retired or passed on, and the market has been producing far fewer of the relevantly skilled people. But we built the F-1 once. If we were similarly motivated, and if we spent the time and money, it seems likely we could do it again.

Second, rather than actually rebuild the F-1 or the F-1A, it is more likely that we would make the F-1B. In 2012, Rocketdyne created a design for a modernized F-1, taking advantage of the advances in computer-guided machine tools, laser welding, and the like. The new design is much simpler: one major part of the engine, for instance, is composed of 40 parts, rather than 5,600. It is also less expensive and intended to produce the same lift at the F-1A.

In this sense, we still can build the F-1 engine. If we decide to put in a lot of money and skilled effort, we can build the modernized version of the F-1, which is better.

Damascus steel

Now let’s consider Damascus steel.

Starting in the 3rd century AD, a special type of steel known as wootz steel was shipped from India to Damascus, where it was forged into blades. These blades were especially strong, flexible, and resilient. They had a signature surface pattern resembling oil on water, were said to have been carried by many Indo-Persian rulers, and gave rise to the legend of blades so sharp that silk would part if dropped on them. The forged steel that composed these blades was called Damascus steel.

(Blades and other objects made through pattern welding are also sometimes called “Damascus steel” as a result of the similarity in surface pattern. Here we’re just talking about wootz Damascus steel, which is where the name originally comes from.)

Production of Damascus blades with a high-quality pattern ceased around 1750. Production of blades with a low-quality pattern ended in the early 19th century. Researchers have proposed many explanations, including the disruption of long trade routes, the lack of ore with the right impurities, or a ban on production by the British government.

Loss and redevelopment

Whatever the cause, it is clear that there was no Damascus Steel Production Knowledge Retention Program. We didn’t get 20 volumes of technical documentation and tape-recorded interviews with expert blacksmiths. The techniques for making wootz steel were lost. The techniques for turning wootz steel into Damascus steel were lost. No known samples of wootz steel or the original ore remain.

Nevertheless, we still have some original Damascus steel swords and daggers, as well as some historical commentary on the material and process. These combined with modern chemistry and advanced microscopy have given us a number of clues. Working very hard, some researchers are making progress figuring out how to recreate the original Damascus steel. Some even think they’ve figured it out.

A continual effort

The effort to preserve knowledge is not limited to rocket engines and swords.

In mathematics, a small group of mathematicians is working to simplify and consolidate a 10,000 page proof scattered across more than 500 journal articles, so that it isn’t “lost to the living world of mathematics.” They hope to be done by 2023.

In the intersection of religion and music, over the past 30 years one scholar worked with the last representatives from over two dozen musical lineages and recovered several musical instruments on the brink of extinction in an attempt to preserve the Sikh tradition of sacred music. There are now regular meetings for an international group of musicians to ensure the relevant knowledge is passed along.

Indeed, there are efforts to preserve knowledge across all disciplines, at all scales, from notes we jot down to scrapbooks to official written records to multimillion-dollar programs like the F-1 Production Knowledge Retention Program. This is because knowledge does not accumulate automatically. The preservation of knowledge is something we are all engaged in. It is a continual effort.

The case of the F-1 engine is especially illustrative here. The F-1 is world-famous, the best of its kind, and was involved in a world-historical event, the moon landing. But these things alone were not enough to protect it from the impermanence of human memory, a changing technological landscape, and shifting societal priorities. Likewise, Damascus steel was also world-famous and the best of its kind. But this did not protect knowledge of how to produce it from being lost.

The decay of knowledge is the default, not the accumulation. This is why we have to work so hard to preserve it.

The mechanism of accumulation

If the accumulation of knowledge is not automatic, the next question is: what causes it to accumulate? This is the topic of the next essay. By understanding what allows knowledge to accumulate, it will then be possible to diagnose a problem besetting a substantial portion of present-day academic and scientific research.


Next essay: Intellectual Schelling Points
Previous essay: How to Understand Research Productivity