Peter Cochrane: Brittleness by design
Much of the modern world is poorly prepared for any deviation from the status quo.
As a student in the 60/70s, my broad-based engineering education included systems reliability and resilience. I soon discovered the UK energy grid enjoyed an excess capacity of 23% to allow for routine maintenance, surprise failures, and the occasional severe winter. Clearly power shortages and outages were unthinkable!
Today's excess capacity has shrunk to ~2% and we are at risk of winter power outages. Complementing this situation is the near eradication of all natural gas and oil storage in the name of greater efficiency. Now add 'Just-in-Time' (JIT) supply chains at a global and local level, and the UK's position looks precarious. But so it is for many countries across the planet where 'JIT' has been fully embraced.
Much of the modern world is poorly prepared for any deviation from the established ~70 years of stability since WWII. The sudden eruption of another European War is perhaps the most extreme disrupter example, and especially so when the combatants are prime suppliers of food and energy! How the heck did we get here?
Over the past 50+ years the reliability and resilience of products and supply chains have improved markedly on the back of our advancing technological capabilities. Failures have been few and far between, and efficiency has become a prime consideration. As a result, quality has improved, quantity/availability increased, and relative prices have fallen continuously. No surprise, then, that our designers, engineers, managers, economists and politicians choose to optimise at all levels of societal activity. The snag is: the concatenation of apparently isolated optimisations results in a 'brittle infrastructures' that are ill equipped to deal with deviations from the status quo.
Sad to say we really understand all this in great depth; but have perhaps, been lulled into a false sense of stability whilst also being seduced by improved 'Return on Investments' (RoI) in the widest possible sense. Apportioning blame, or quantifying the causality, is likely impossible, but by-and-large, reliability and resilience teaching in colleges and universities has been replaced by optimisation studies in management economics. Sadly, in many engineering and technology courses these subjects appear to be omitted - along with applied mathematics.
It is hard to understate the risks posed and consequences when the relationship between reliability, resilience and optimisation is not appreciated. Perhaps the most graphic example of the criticality is the difference between a standard family saloon and an F1 racing car. The saloon is grossly inefficient but inherently safe, reliable and resilient, whilst the F1 is the converse! The dangers of single mindedly optimising systems or products by software, hardware, performance, management, supply and support chains are writ large by many past defence and aerospace projects.
Catastrophic product outcomes can result from singular and seemingly unimportant decisions that appears inconsequential at the time, but it is so simple to conjure the perfect storm. Witness the Boeing 747 and NASA Challenger Disasters for examples of where management did not appreciate what they were being told by their engineers and went ahead based on their perceptions.
Engineers try to get the very best performance a component, subsystem, and system at a time, whilst managers are concerned with timescales, competition, time-to-market and RoI. And both groups can be exposed to accelerating expectations beyond what is reasonably achievable. At the same time, they tend to employ simplified economic models focused on upfront, instead of, whole-life costing, performance, and quality. The real enemy here is situational complexity exceeding human abilities and it is time to augment teams and individuals with computer modelling and AI-learning across the whole engineering and management cycle.
Peter Cochrane OBE is professor of sentient systems at the University of Suffolk, UK