"There are two words that nowadays' engineers will hear left and right, find in every corner of every project, sometimes thrown around in a context where they do not make sense. These are the words “machine” and “learning”. They are repeated like a mantra, promising a panacea that will cure all ailments of modern engineering. The stunning results reported by the tech giants when applying the techniques collectively designated by the aforementioned umbrella term are indeed worthy of attention. However, how much of that excitement is hype, and how much of it is actually productively useful? In my current project, I am seeing the difficulties encountered when bringing this new engineering to a more traditional production environment.
Traditional engineering is reliable; it has been tested repeatedly, and can be performed with a lot of knowledge and a little bit of Fortran/C and Excel. Machine learning based techniques rely on a tighter symbiosis between machines and us humans, meaning that not only do these machines need to think more like humans, but we humans need to understand that the machines do not (yet) think like we do. Process automatization, construction of effective databases and knowledge of hip IT trends are the necessary building blocks for a successful application of the newest engineering methodology.
I have been given the task to investigate and eventually process the data generated over years by that “old timer” kind of engineering. The biggest challenge I have found so far in doing so has not been at all similar to a hard math problem, it has been more akin to a political discussion. It stems from the collision of two very different worlds: it is the inertia of a whole generation of engineers used to Excel work who are suddenly presented with the concept of data analysis, coding and self-made software development. I have also faced “the hype”, an expectation that data science will deliver results fast and easy. However if the infrastructure is not there, this can only be done when a humongous amount of work has been put into data preparation. Not to mention that the tool of machine learning must be used upon the right kind of problem. It does not mean that machine learning should not be used, but it definitely must be understood before being applied.
It has been so far an enlightening experience, and despite the hardships very positive, because I can bring some of the new exciting ways of thinking to those who need them. Through the resolution of MCA to hire young engineers, the company helps deliver change together with its engineering offer."
Pedro, consultant for MCA Engineering Gmbh