Who tests the Robot?

Recently I have been set the challenge of trying to work out how to test robots and how to test (an) artificial intelligence. You’ll be pleased to know there are no spoilers in this as, well quite frankly, it remains a work in progress. Under duress, I would say my current thinking focuses on how predictable a robot should be, and how the concept of ‘expected results’ falls apart with artificial intelligence. But, as usual, that’s not the main topic of my blog. Merely a mental aperitif to whet your appetite and a way for me to ease myself into what I want to say. I was recently introduced to the World’s first killer app, an application which has had a profound impact on computing beyond its original intended purpose, and ultimately a completely unforeseen bi-product of itself (yes, it’s not a recent product!) Previously I have written of how Amazon and others achieved success through digital transformation, for instance by selling books from a warehouse and not from a shop. It was a very early example of a digital transformation from over 25 years ago which has profoundly impacted how we shop now forever. So today’s challenge is to think of something which predates Amazon but changed the very fabric of how people work. It took a manual task and automated it, freeing workers from time-consuming labour to produce updates, freeing people from repetitive and routine tasks. And yet this automation liberated a workforce to focus on areas previously unattainable – simply because they no longer spend their time dwelling on the mundane, routine activities. There are some sticking points about this example which make it initially appear fictional but in fact differentiate it from the modern perception of automation, a perception of robots taking over and replacing humans, displacing them from the workforce. You know, that vision of driverless taxi which can be beckoned on demand to provide cost-effective transportation – and replacing our need to own a car ever again. Although the question of what artificial intelligence does in an emergency may appear irrational. It’s easier if you just watch “I-Robot” at this point though, as it does a fairly good job of introducing the basics (for me). Back to the topic – the ‘Killer App’ was Visicalc. The precursor to Lotus 123, and then to Microsoft Excel, an application designed to very effectively ripple the updates of a single change throughout a full set of financial transactions instantaneously, removing the need for days of manual re-calculation. But I doubt anyone now would think of a (humble) spreadsheet as a Killer App, mostly because the phrase “Killer App” was not part of the vernacular in the 1970s. What we know now as a result of a retrospective look at Visicalc’s impact, is that automation can have two significant impacts:

1. The digital transformation of the automotive industry will have a severe impact on the livelihood of those people who earn their living as professional drivers; the automotive industry is far from being alone in facing this challenge.
2. The automation which Visicalc brought freed bookkeepers from laborious, repetitive tasks but actually increased their numbers as the profession was able to spend more time focusing on areas of concern, areas which were previously ignored because of the intensity of implementing minor changes manually.

It is even possible to find examples where automation has brought with it both of these impacts on a single industry. Cockpit automation in civil aviation has reduced the crew required to pilot increasingly large aircraft reduce from three to two. Part of this is the introduction of remote, centralised engine monitoring, which enables a greater volume of data to be analyzed to predict and pre-empt failures before they happen, improving safety. Sadly the other side of this coin appears to be an automated system designed to make flying a new plane safer which is now alleged to have cost many lives because of two fatal incidents. There are few certainties we can take from all of this but I feel the first one must be that automation is not only here to stay, but its pace will only continue to gather momentum. The second is a need to remember that humans are not entirely rational creatures and there is a large gap between how we operate with our organic processors, and how a silicon-based logic circuit may operate. We will very much need to adapt if we are to make a success of automation, in terms of bridging the gap between intelligence and automation. Our on-going experience at Sogeti is one of successfully delivering digital transformation. We achieve this not just by increasing test but by taking into consideration the broader impact across the system development life cycle and upon the culture of the organisation transforming.