Flat White

The fall of the machines

6 February 2019

7:14 PM

6 February 2019

7:14 PM

About 30 years ago, General Motors experimented with fully automating a car factory. An immense design effort was implemented to replace every manual task with a machine so that robots did everything. Within a few years, the car company reversed their decision and started doing the exact opposite—replacing robots with people.

The project was a multi-billion dollar failure, but the lesson was, on the whole, worth learning. The key reason for the failure was that, while robots can do many things, there is one thing they cannot do: improve.

Humans, on the other hand, have brains, and the human brain is an immensely valuable asset. Not only can it continually improve and optimise tasks, it can adapt to unexpected circumstances in a way that will be increasingly effective each time. In general, a human brain will not make the same mistake more than once, and a human brain will seek out ways of measuring its own effectiveness. If a robot can make a mistake once, it can make it ten thousand times in a row.

When all activity is taken out of human hands and put into metallic ones, this doesn’t eliminate the need for human activity. Instead, it replaces an army of workers with a larger army of expensive, specialised workers—the sort who can design and maintain robots.

Though the established car industry learnt this decades ago, Elon Musk only learnt this last year. In 2018 he tweeted,

Excessive automation at Tesla was a mistake. To be precise, my mistake. Humans are underrated.

It was a comfort to know that despite advances in “machine learning” and “AI”, we are still nowhere near making ourselves redundant. Car factories today still contain robots, but the robots are tools—operated, maintained and collaborated with, by humans.

I used this as an anecdote in a speech last year about writing “standards”. I had the privilege over the last three years to assist in writing the Australian Standard for gas and petroleum pipeline design (AS 2885.1). This standard is described as a goal-oriented or objective standard. Its aim is to tell you what a design should achieve, but not to prescribe how the design should achieve it. The alternative is a prescriptive standard, which tells the reader exactly what to do. Though some prescription is always necessary, I think that being goal-oriented is a good aim for a standard to have.

The caveat for an objective standard is that whoever implements the standard needs to be competent for the task. A prescriptive standard, on the other hand, transfers the burden of competence to the standard itself. This means that, arguably, any idiot could use the standard safely. As a design engineer, however, I have worked with prescriptive standards in the past, and it is frustrating when you are unable to implement a great solution for your specific context that the writers of the standard haven’t accommodated! Standards, like robots, don’t improve. Standards, like robots, are only as omniscient as their makers.

Here is the analogy: if a standard shifts the focus of workers from succeeding to complying, then it turns people into robots. The same applies to regulations and legislation and policy-making in general. Perhaps you can picture the deskbound receptionist repeating “computer says no” or citing any number of regulations that mean a common-sense solution is not permitted. Like over-automation, prescriptive policies require a large army of expensive specialist workers, policy-makers, whose job it is to think up all the solutions to all the possible problems… and who will definitely fail to do so.

Since long before the term “human resources” was coined, it has been true that people are resources. The strength of a human resource is its mind – this un-paralleled machine capable of striving for efficiency and success in every activity.

The tendency of a governing body is to want to be the only brain in a body – the body is just “machinery” and the brain makes all the decisions. However, when an organisation has one hundred employees, that’s one hundred brains going to waste if it is managed this way. Good governance should be more like an octopus. Octopuses have nine brains – one main brain and one smaller brain in each limb; the main brain instructs the limbs on the objectives and the big picture, but each limb utilises its own thinking power also.

Goal-oriented instruction isn’t a fundamental principle. There is a place for prescription. There is undoubtedly an important place for rule of law, which is a strong protection against corruption and provides some predictability and stability, which we all need. History has shown, however, that right-wing and left-wing ideologies can both tend towards authoritarianism in the end. Today, with the essentials for life (food, clothing, and shelter) so efficiently produced by such a small portion of the population, we have an unprecedented capacity for big government.

In their efforts to prescriptively regulate our society, neither side of Australian politics look like slowing down. New policies, new approval mechanisms, new regulation bodies, new watchdogs, new royal commissions, with new recommendations for more oversight and regulation and rules, rules, rules. “Deregulating” is a common promise rarely achieved. One would think that people don’t have brains.

In the USA, however, Donald Trump’s “one for two” deregulation rule is partly credited with their increased economic growth. This rule says that for every one regulation introduced by policy-makers, two old regulations have to be removed. Not only will this generally require regulations to be less prescriptive (because prescriptive regulations are always longer than objective regulations), but he has used the same principle in how he has managed the policy-makers themselves. He hasn’t prescribed which regulations to delete and which to replace – to satisfy his rule, his objective, they will have to use their own brains and ingenuity. This isn’t “daddy knows best” management, it utilises and trusts the workforce.

I’m a centrist. My point has been that the centrist compromise isn’t just about the left and right, it needs a dash of libertarianism, too. A bit of individualism. A bit of personality. It’s time to replace the robots with humans.

Nick Kastelein is a Christian and a conservative who grew up and lives in Adelaide where he works for an engineering consultancy.

Illustration: The Ladd Company/Shaw Brothers/Warner Bros.

Got something to add? Join the discussion and comment below.

Show comments