“It is difficult to make predictions, especially about the future.”(1)

The foolishness of these words might seem obvious, but there is a case to be made that it is indeed possible to predict the future. In his address to the Royal Institution (24/01/1902) entitled The Discovery of the Future, H.G. Wells argued just that. The gist of Wells’ argument is that, given the scientific advances in geology, palaeontology and archaeology that have given a vast time depth in knowledge of the past, similar principles should be applicable to the future. In my favourite section of his address, Wells remarks:

Angular pit-sand has grains of the most varied shapes. Examined microscopically, you will find all sorts of angles and outlines and variations. Before you look you can say of no particular grain what its outline will be. And if you shoot a load of such sand from a cart you cannot foretell with any certainty where any particular grain will be in the heap that you make; but you can tell—you can tell pretty definitely—the form of the heap as a whole.

Some aspects of the future are predictable and it turns out that industry has, over the years, explored the mechanics of prediction in some detail.

Experimental piles of sand


In a recent conversation with Cornelius Holtorf about some work he is doing on the possibility of periodising the future, I was reminded of some questions about product warranty that had interested me for some time. I first encountered this issue when building Windows PCs for myself and others, beginning in the late 1990s. The core element at the time was the hard disc (the more recent advent of SSDs has changed things radically). Being an electromechanical component, it was the most likely to fail. I, like many other PC builders, noticed the wide variation in warranty offered by leading manufacturers such as Western Digital, Seagate and IBM (later Hitachi), with some offering up to a 3 year warranty. Logically it was obvious to choose the latter since if the manufacturer had such faith in their product, its durability/reliability seemed promising.

Over the past decade or so, such long warranties have become much more common. When we bought a new central heating boiler in 2011, the manufacturer offered a 7 year warranty, which they have now increased on new models to 10 years. We have just bought a new washing machine which has a 5 year warranty with a 10 year warranty on its main motor. In the car industry, warranties between 3-7 years are now common, and it is the case that compared with the 1960s-70s, the expected life of a car in terms of mileage has increased from 100 000 to 200 000 miles.

Do these manufacturers know something about the difficult task of predicting the future that we don’t?

Time, scale, nature, culture.

In fact science offers a whole load of predictions about the future, many of which we take for granted. That the sun will rise tomorrow is what philosophers call inductive knowledge, but it is a pretty safe bet. If one considers the solar system, it is a relatively simple system where the chances of some chaotic event disrupting the orbits of the planets is fairly remote. And at this huge scale processes are pretty slow, it will probably be about 7.5 billion years until the earth is absorbed by the sun.

Down here on earth, things happen much more quickly and are much more complex. Yet in recent years, since the application of super computers, the ability of meteorologists to predict the weather has increased markedly. Unfortunately, the same is not true of economics, which is hardly surprising given the earth’s population of 7.7 billion people, each in possession of 86±8 billion neurons. But things, at least, are more predictable than people.

Predicting technological futures

One assumes that early technologies were highly predictable. Given the state of many surviving Palaeolithic handaxes, they are pretty robust. Equally most technologies until the post-medieval period were relatively simple and easy to repair. In some cases, particularly Roman concrete, the formulation actually includes a self repairing quality. Generally, in my opinion, it is when technology becomes more complex that issues of reliability become prominent, and this arises in parallel with the refinement of manufacturing techniques to produce artefacts to increasingly exact tolerances. Thus, for example, the advent of the internal combustion engine was facilitated by advances in machine tools that allowed for more precise cylinders and pistons than those found in steam engines. Clearly there has been a learning curve involved here; for example, the infamous failings of the De Haviland Comet airliner were due to ignorance of the effects of metal fatigue. Although the existence of metal fatigue had been known since 1837, its effects had been underestimated in the design of the aircraft (the World’s first passenger jet), tellingly in the use of square rather than oval windows, as are seen on all modern passenger aircraft.

By the late 1960s, engineers had begun to visualise what is known as the “bathtub curve”, which consists of a period of “infant mortality” or burn in period; a period of constant failure rate and a final phase of wear out failure. In this sense, engineers do indeed periodise the future.

Bathtub Curve – a periodisation of the future?

The concept of a burn-in period was, supposedly, first adopted as an element of the Minuteman missile programme. Returning to my PC building experience, this is a familiar concept. When building a PC it is common to “breadboard” the components before final assembly (2). It makes sense to find out if all the basic components; motherboard, processor, memory, storage device; will actually produce a bootable computer before going to the trouble of fitting them into a computer case. If new components are going to fail, it is often when they are first turned on. By contrast the constant failure period, which represents most of the life of an artefact, represents the occasional faults that develop in use. The final phase is the inevitable approach of death, as it were. Obviously this model resembles the human life cycle and perhaps adds a new dimension to the notion of artefact biography.

Breadboarding a PC

Clearly, the overall longevity of an artefact will depend on the predictability of its components and assembly, and here refinement of manufacturing technique intervenes. Six Sigma is a production management model, introduced by engineer Bill Smith at Motorola in 1980. The term refers to the standard deviation, sigma, in behaviour of components, etc. with the goal of producing 99.99966% of parts free of defects. In effect it involves a set of heuristics to refine out sources of error, contamination etc, and resembles other production models, such as the Japanese concept of Kaizen. I suspect that this kind of technique is at the heart of the increasing length of product warranty noted above (3). Interestingly, one criticism of such approaches is that they stifle creativity.

Use life also plays a part. The rather dry publication Reliability Engineering (Kapur and Pecht 2014) notes that “it has been found that some Asians use a dishwasher to wash vegetables, in addition to eating utensils.” and hence this affects the life expectancy of the dishwasher. But beyond this, there are also elements built into technologies to prolong use life, the most obvious example being fuses and other cutouts designed to protect electronics from overload. In the case of critical technologies, redundancy of components is also included. As noted above, Roman concrete (either by accident or design) includes components of volcanic ash that give it self repairing properties; calciumaluminosilicate crystals form in the concrete matrix and act to repair cracking. In recent years, similar properties have been sought in industrial materials, particularly in attempts to produce self-repairing polymers. In this instance one might wonder whether reliability is always desirable; given the growing problems with waste plastics and microplastics, self-repairing plastics might turn out to be disasterous.


As mentioned above, the durability of cars has increased markedly in recent years, so much so that manufacturers are apparently looking to make interior materials more durable, so that they don’t look worn in relation to the rest of the vehicle. What is odd here is that it was in the motor industry that the idea of planned obsolescence had its origins, with General Motors introduction of the annual model change in the 1920s. Clearly obsolescence is the antithesis of durability, it implies, as the title of Giles Slade’s (2007) book suggests that goods are “made to break”. Actually I think there are two distinct aspects to this; there are artefacts that are expected to be disposable, either because they are cheaply made or because their use life is expected to be short, a kitchen scourer for example. But it is also the case that some products simply become outdated; this is certainly the case in the history of computers and mobile phones, for example, where technologies have developed so rapidly that in effect, your computer is out of date as soon as you have bought it (4). Obviously in this context there is less utility in a device lasting a long time if it rapidly becomes obsolete in the sense of technically outdated.

In recent years, partly due to environmental concerns, the counter concept of future proofing has gained currency, including in the field of heritage (Rich 2014). Generally future proofing is about making things that are resilient, in the sense of being repairable, updatable or versatile. To some extent this applies to heritage sites, although I doubt that updatability is really an issue. But here too there are contradictions. As DeSilvey (2012) notes, the National Trust aims to preserve its sites “For ever, for everyone” an ambition which would seem unlikely to be achieved. Indeed the case of Mullion Cove is one in which some form of managed decay has been accepted, again in connection with environmental issues. Elsewhere apart from the flexibility of use which can help keep the relevance of historic buildings (Rich 2014), the emphasis is on prevention of decay and here it is interesting that repair itself is functionally linked to resilience: “Interventions in existing buildings should use equally durable building materials. Don’t mix short-term materials with long-term materials.” (Rich 2014: 39).

Past and Future

To my mind there are symmetries between the past and future, not least in that the further one goes from the present, the less one can know. In the work that Cornelius is doing, and in the above example of the bathtub curve, it is the case that aspects of the future can be periodised, in a way analogous to the past. When I was a student at Sheffield University we joked about the fact that the 14 storey Arts Tower had a predicted life of 20 years (it was about 20 years old at the time). But the glaringly obvious divergence of past and present is in that the former is over; we may learn more about what happened, but it will never change what happened. With respect to the future, whilst the domain of things may be relatively predictable, and probably is becoming more so, it is actors who make for the stochastic elements, as the aforementioned dry tome suggests “Systems do not fail, parts and materials do not fail—people fail!” (Kapur and Pecht 2014, quoting Lewis 2003). Very much an engineers perspective, we may think, but how people fail is in failing to be predictable, and here I would want to suggest that by actors I don’t mean those defined by Latourean symmetry, specifically because such non-human “actors” (excluding other living beings) are predictable. This is clearly another symmetry between past and future; if we think of Leroi-Gourhan’s concept of the chaîne opératoire, it is the ways in which materials constrain the actions of humans that make the sequence of actions understandable in, for example, the reduction sequence of a flint core.

Sheffield University Arts Tower – still there in 2015

In both past and future, one of the key drivers of unpredictability is entropy. As we know from the laws of thermodynamics, which are themselves a key component of time’s arrow (5), there is a universal tendency towards increase in entropy in the universe, even if, as is often claimed, human actions in creating order in the world are “negentropic”. That which remains of the past in the present is the product of both systematic and entropic processes; people have made and discarded, created and destroyed. Apart from the entropic aspects of post use taphonomy, past entropic actions are also to some extent “fossilised”; the ancient rubbish heap achieves a relatively static stratigraphy. The same cannot be said of the future. Wells’ heap of sand reiterates a certain form, whereas the disposition of grains is unpredictable. But in the human world, the grains themselves are highly complex agents, perhaps suggesting that even the form of the heap may not be predictable.


1)This quote originates from the Danish politician Karl Kristian Steincke in his autobiography Farvel Og Tak – in the original Danish “Det er vanskeligt at spaa, især naar det gælder Fremtiden.” 2006, The Yale Book of Quotations by Fred R. Shapiro, Section Niels Bohr, Quote Page 92, Yale University Press, New Haven.

2)The concept of breadboarding originates in the early days of radio, where amateurs often assembled radio recievers on a breadboard. I tend to use some corrugated cardboard (an insulator) spread on my workbench

3)Obviously there are other factors – competition will lead manufacturers to offer a longer warranty, thereby factoring in an addition cost of failures. But evidence of significant failures where products have to be recalled suggests that lack of genuine reliability can be very expensive.

4)Although one might argue that computer technology has reached something of a plateau, where innovation has stalled, e.g. the fact the Microsoft has decided to introduce no new versions of its operating system beyond Windows 10. Interestingly, the idea of disposability seems to have deeper roots in the USA; de Tocqueville notes this in his Democracy in America (1835/40; see also Rolt 1965)

5)It seems to me that time’s arrow has at least two distinct components. The first is the Einsteinian concept of time as a 4th dimension. But within this frame, the thermodynamics of entropy, which presumably have their roots in quantum physics, are a quite separate trajectory within which time is measured and perceived.


De Silvey Caitlin 2012 Making sense of transience: an anticipatory history. Cultural Geographies 19(1) 31–54

Lewis, N., 2003 “Reliability Through the Ages,” Presented at Canadian Reliability and Maintainability Symposium, Ottawa, Canada, October 16–17.

Kapur, Kailash C., and Michael Pecht. 2014 Reliability Engineering, John Wiley & Sons, Incorporated,.

Rich, Brian D 2014 The Principles of Future-Proofing: A Broader Understanding of Resiliency in the Historic Built Environment. Preservation Education and Research 7

Rolt, L.T.C. (1965), Tools for the Job: a Short History of Machine Tools, London: B. T. Batsford,

Slade, Giles (2006), Made to Break: Technology and Obsolescence in America, Harvard University Press,

Wells, H. G. 1902 The Discovery of the Future http://www.gutenberg.org/ebooks/44867

Leave a Reply