The most unrealistic thing in Terminator

It’s not time travel: it’s how Skynet produces its robots

So there’s one more Terminator movie coming out. Yawn. By this stage the franchise has been so thoroughly messed up that it’s almost excusable to forget how awesome the two first movies were. The Terminator, and Terminator: Judgment Day, are classics, and justly so.

However, one thing always bugged me about them. And no, it’s not all of the problems that come with time travel. It’s not either the impracticality of the design of the Terminator.

It’s the fact that Terminators would be built in series.

The Terminator portrayed by Arnold Schwarzenegger is of course the T-800. The liquid metal Terminator in Judgment Day is the T-1000, and a host of other models have appeared in the various sequels and offshoots along the years.

Now what is wrong with that, in my opinion, is that mass manufacturing series of identical machines is a typically human thing, and more specifically a 20th century human thing. This is not something that an advanced AI would be doing: in fact, I’m certain that every single Terminator produced by Skynet would be unique. Let’s dig into the why.

Why we make series of things

The 20th century was the age of the mass-manufactured product for two reasons: the tooling costs involved with the creation of a new product are high, on one hand, and manufactured goods were relatively few.

In our age of constant new product announcements, it’s easy to forget that this is a very new thing: throughout the 20th century, everyday products weren’t improved upon at such a rapid clip. Take the telephone as an example: the Model 202 was the standard telephone in the US from 1930 to 1939. It was replaced by the Model 302, which was manufactured from 1937 to 1955, which was replaced by the Model 500, whose production ran from 1950 to 1984. That’s 34 years of continuous production of a single model of phone! Think about that next time you see The Verge wondering why Samsung hasn’t announced a new gizmo in all of 2 months.

This wide availability of a limited number of different devices brought their retail prices down, of course. But, perhaps counter-intuitively, the cheaper the retail cost of a product is, the more expensive its tooling and supply chain tends to be, and therefore the more expensive it is to get a replacement into production.

The Reis telephone, a very early device, from 1861

The Western Electric Model 500, manufactured from 1950 to 1984

Let’s take the example of the telephone again to explain that. Consider a very early device such as the Reis telephone pictured here, built as a prototype by an inventor. Had that device been commercial, its price would have had to be very high, since it required so much hand labor. But the investment necessary to get there is zero. You don’t need to machine expensive molds for plastic parts, secure the right partnerships or organize a global supply chain: just pop into your workshop and build one.

On the other hand, the retail price of a widely available set such as the Model 500 is very low. But to be able to get there, the investments are enormous: they include tooling, specific machinery to mass manufacture the different components, and setting up and managing a global supply chain.

Now, there is a circular relationship between tooling costs and the limited range of goods available: the more of a certain device you intend to manufacture, the higher your tooling costs are going to be, and therefore… the more you’ll want to sell of that same device. On the other hand, should you intend to sell only one copy of a certain device, you will not invest in tooling, and it will not be a problem if your next customers asks for some tweaks.

The problem with mass market

The main problem with markets where few products are introduced in large quantities is that, as a manufacturer, you have precious little opportunity to learn. From 1950 to 1984, Western Electric was selling the same telephone, and could therefore not have a feeling for shifting preferences, demographics, or culture.

In a competitive market, this presents obvious drawbacks, since every newcomer entering the market with an updated product will stand to learn a lot more about the market. Many of these will fail, sure, but whenever one comes across an under-served niche, incumbents stand to lose market share.

The new manufacturing ethos

One of the first industries to catch on to this was software, of course, where people have long known of the advantages of building and delivering products in small increments. This spawned the agile development processes in the 1990s.

Software has a tremendous advantage compared to hardware, in that it typically doesn’t involve any tooling costs at all : while launching a new product may require you to build a new factory, a software is just code, so what you stand to lose is just the time you spent writing it.

However, this ethos is rising in manufacturing too. Thanks to a broad array of new technologies, loosely grouped under the appellation Industry 4.0, it is now possible to turn a profit manufacturing products on demand. For instance, Shapeways will 3D print any shape for you, Pediapress lets you publish your own book constituted of Wikipedia articles. Companies are exploiting this by using algorithms to automatically create product listings — sometimes with horrifying results such as the phone case pictured below.

That product is horrible, and there is an argument to be made that it is symptomatic of a deeply broken system. But for the company behind it, releasing it came at no cost. No one spent time selecting the picture or editing it. If no one bought it, not one case would have been produced, and no one would have unsold product on their hands. If someone bought one, though, the company would have turned a profit. It’s all upside, and no downside.

The technological breakthrough that makes these products possible is the digitization of the manufacturing tools as well as all steps of the communication: the Toenail Fungus phone case was generated by a computer mashing together stock pictures with generic phone cases. It was automatically listed as a product on Amazon. Should anyone have bought it, an order would have been sent to the production plant to digitally print this terrible picture onto a phone case and ship it to its happy owner. No human would have been involved in the entire process.

In a pre-digital era, the lone purchaser of a Toenail Fungus cover for a Western Electric Model 500 would have had to pay thousands of dollars and wait for numerous weeks for the work on his or her phone to be completed, involving countless employees throughout the complex supply chain required to make this happen.

How Terminators would actually look like

With that in mind, let’s go back to Skynet creating humanity-killing machines. As an advanced AI unbridled by manufacturing, marketing, or ethics, it makes no sense at all for Skynet to manufacture large series of Terminators. Instead, every single one of them would be an iteration on the previous one: it would explore new alleys or serve a niche role. Skynet could therefore learn with each one of its creations, and come up with a… ahem… killer design?

Though it probably wouldn’t look Arnie-in-leather-on-a-bike awesome.