Hardware engineers need old knowledge software engineers can skip
I once wrote a piece on my blog titled “Which is the better deal, software or hardware?” I’ll skip the details of the argument, but the conclusion was this: “There is a large difference between the technical systems of software engineers and hardware engineers, and this inevitably changes what society expects of each, but neither one is directly superior to the other, so you should just choose whichever you like.” That difference in characteristics comes down specifically to one point: what the engineering is based on.
Hardware-related engineering is based on natural science. Physics, electromagnetics, materials science — these are not things humans invented. Hardware engineering means doing what humans can do without ever straying from the laws of nature, staying aligned with them at every step. Many people would say the same of software — that “fundamentally it’s math.” But most software engineers have no need to be conscious of that. Their main job is instead to work “on top of efficient, rational rules someone else built.”
So while the latest technology matters for hardware engineers, most of what they do is impossible without knowledge and experience of the older technologies underneath it. In software, though, plenty of people work on app development in Java — which is built on top of C — without ever having written C themselves. And ultimately, languages run as assembler (machine code) on the hardware, but an engineer who can actually work at that level is probably about as rare now as a COBOL engineer.
In the software domain, the ranking of what’s valuable and what isn’t flips extremely fast. Before the 2012 Google I/O demo of AI automatically recognizing cats, I doubt many people paid much attention to the value of machine learning engineers. Now they’re the poster child of high earners. When I was a university student, people wrote HTML by hand. Perl was all the rage, and I remember building various things myself with it. But twenty years later, nobody does that anymore, and nobody even mentions it.
In the hardware domain, on the other hand, market value tends to accrue to people with deep know-how. Take “unwanted radiation,” for example — harmful radio waves emitted from electronic devices, which every country regulates. The theoretical framework for reducing this has existed for a long time, and simulation technology keeps advancing by leaps and bounds. And yet problems still happen. The reasons vary case by case, but most of them trace back to the limits of what could be anticipated — things like “it works fine standalone but not in combination” or “we can’t tell whether that extra component we added is the culprit.” Something similar happens in software too. But where a setback in software development might mean a week of lost progress, or a couple of all-nighters to fix, in hardware it tends to mean redoing a mold, delaying the schedule by months, and losing tens of thousands of dollars — the damage is inevitably much larger. It’s usually the veterans — the people with deep experience and know-how — who sense this “smell of trouble” coming. 3D printers have eaten into part of the market, but they haven’t destroyed it. The way electronic PCBs are made has stayed the same for decades. This is another fundamental difference between the two fields: the speed at which time passes.
So when it comes to the question in the title, take it as fundamentally a question about software. To hardware engineers, my advice would be: specialize as deeply as you can in your own expertise, rather than chasing trendy technology.
Let me share two episodes here.
The first dates back about five years now. At the time, a subordinate of mine — an engineer in his mid-twenties, Mr. K — startled me by saying, “PHP is where it’s at.” This came up during a discussion about what language to use to build the server side, and while everyone else just laughed it off, I was curious about the reasoning that led him there. (We ended up going with Scala.) He declared flatly, “The ideal in programming is code that completes in a single line.” In other words, his argument was that what matters is understanding the many frameworks out there deeply enough to use them effectively for the purpose at hand. I felt there was something to that. And the reason for PHP specifically was that many PHP frameworks are used in China, with commit counts on an entirely different order of magnitude from other languages — meaning these frameworks had already been battered by huge numbers of people and were “more than sufficiently battle-tested.” Anyone who’s run a server knows that among maintenance tasks, “responding to updates to languages, libraries, and SDKs” is one of the heaviest and most consequential, and yet it’s the one that tends to get put off until it becomes a real incident. So the standard approach is to choose tools that are as mature as possible. From that angle, PHP has come full circle and is now extremely useful.
PHP here is just one example, but the point is: the purpose of development is to build something, and if the latest technology isn’t specifically required for that, it’s entirely rational to choose something stable that sufficiently serves the purpose.
The other episode is from about two years ago. I want to share something I honestly felt while working alongside a few AI engineers. It was how incredibly studious they were. AI is still, in many areas, what you’d call in its infancy. Because of that, much of the valuable knowledge tends to come out of university research and get published as papers. So they read through an enormous number of papers, nearly all of them, needless to say, in English. I don’t think people who just use AI-related tools provided by Google or Amazon can be called “AI engineers.” As I mentioned, in a field still in its infancy, what’s needed far more than “people who just ride along” is “people who think up new algorithms and take on challenges.” Twenty years ago, for example, wasn’t there some kind of quasi-professional title for people who were skilled at building Excel macros? But that work gradually got simplified and commoditized, and eventually fell to the level of something anyone could do. The only people who get to keep being called “engineers” are the ones who kept pursuing some area of expertise with real rigor — the mechanics of spreadsheets, new methods of graph rendering, UI/UX, and so on.
You should not “chase trendy technology.” But if something genuinely fascinates you, I’d encourage you to judge your own feelings rather strictly — ask yourself, “Do I think about this thing morning, noon, and night?” and “Am I prepared to devote my career to pursuing it?”
The content of this post is an excerpt (original text) from the following book. If you’re interested, please get a copy.
The Shape of a Happy IoT Startup
The Shape of a Happy IoT Startup
Originally published in Japanese at https://clazytech.com/2022/08/1088/. Translated with LLM assistance and reviewed before publication.