In the history of expensive software mistakes, Mariner 1 was possibly the most notorious. The unmanned spacecraft was ruined seconds after launch from Cape Canaveral in 1962 when it veered dangerously off-course due to a lineup of untrue code.
But nobody died and the only strikes were to Nasa’s pride and budget. Imperial College’s modelling of all non-pharmaceutical interventions for Covid-19 that helped persuade the UK and other nations to make draconian lockdowns will supersede the failed Venus space probe and could go down in history as the most catastrophic applications error of all time, in terms of financial costs and lives lost.
Since publication of Imperial’s microsimulation model, those people with a skilled and personal interest in applications development have studied the code on which policymakers predicated their fateful decision to mothball our multi-trillion pound market and dip huge numbers of people into poverty and hardship. And we were at what we discovered disturbed\. The version appears to be totally unreliable and you wouldn’t bet your life on it.
First though, a few words on our credentials. I’m founder David Richards and chief executive of WANdisco, a global leader in Big Data software that’s jointly headquartered in Silicon Valley and Sheffield. My co-author is Dr Konstantin’Cos’ Boudnik, vice-president of architecture at WANdisco, author of 17 US patents in distributed computing and a veteran developer of the Apache Hadoop frame that allows computers to solve problems using huge amounts of data.
Imperial’s model appears to be based on a programming language named Fortran, that was old information 20 years ago and, guess what, was that the code used for Mariner 1. This obsolete language includes issues that are inherent it illuminates values, which may give way to \numerous design defects and inaccuracies that are numerical. One file alone in the Imperial model contained 15,000 lines of code.
Try unravelling that tangled, buggy mess, which appears more like a bowl of angel hair pasta compared to a finely tuned piece of programming. Industry best practice would have 500 different files rather. In our commercial reality, we would fire anybody for developing code such as \any business that relied upon it to generate software available would go bankrupt.
The approach ignores widely accepted computer science principles called”separation of concerns”, which date back into the ancient 70s and are necessary to the design and architecture of effective software systems. The principles guard against what developers call CACE: altering Whatever Changes Everything.
With this separation, it’s not possible to perform rigorous testing of individual parts to make sure full working order of their whole. Testing allows for guarantees\. It is what you do on a conveyer belt in a car factory. Every component and each is tested for integrity so as to pass strict quality controls.
Just then is the car deemed safe to go on the road. As a result, Imperial’s model is vulnerable to creating outputs that are wildly different and conflicting based on the same set of parameters. Run it on computers that are different and you would get different results. To put it differently, it is non-deterministic.
Therefore, it is basically unreliable. It screams the question as to why our Government didn’t get a second opinion before swallowing the prescription of Imperial.
Ultimately, this can be a computer science issue and where are the computer scientists in the room? Our leaders didn’t have the grounding in computer science to challenge the ideas and therefore were vulnerable to the academics. I suspect that the Government saw what had been happening in Italy using its own spa hospitals and panicked.
It chose a blunt instrument in lieu of a scalpel and now there will be a massive strain on society. Defenders of the Imperial model argue that because the problem – a worldwide pandemic – is dynamic, then the alternative should share the identical quality that is non-deterministic.
We disagree. Models must be capable of passing of producing the very same results given the same set of parameters, the basic evaluation\. Otherwise, there is just no way of knowing whether they will be reliable.
Indeed, many global industries successfully utilize deterministic models that variable in randomness. No surgeon could set a pacemaker into a patient knowing it had been based in an arguably approach for fear of jeopardising the Hippocratic oath. When the whole wellbeing of our nation is at stake, would the Government place its confidence at the same\?
David Richards, founder and chief executive of WANdisco and Dr Konstantin Boudnik is your Organization’s vice-president of architecture