Will structural engineering ever be fully automated?

Hey everyone! It's been a while since I put out some content here but I'm back with a small series on some big questions in structural engineering tech!

It’s a common fear among structural engineers today that their job will eventually be automated, rendering years of accumulated knowledge worthless. 

The story goes that software, usually in the form of artificial intelligence or machine learning, will progressively encroach on a structural engineer’s work until the computer will develop the design all on its own, without the need for the engineer to even click a button.

It’s a dire prediction which understandably flares the emotions of structural engineers from outright denial to pointed anger at the thought.

And why wouldn't it? Anything that threatens to cut down not only a well respected profession but people’s livelihoods is bound to elicit an intense response.

And yet in this digital age, where the impossible takes on a certain fluidity, it’s a question worth considering with a cool head.

So what I’d like to do here is offer my perspective. I am a software developer and structural engineer who has developed structural engineering software. My hope is that my perspective can help you understand the mechanisms behind the changes, and give you a bit of foresight on what may be to come.

What you think can be automated and what can be automated are different things 

A civil engineer approached me after I had developed my first structural engineering program to ask how long his work as a road designer would last. They reasoned that so much of their work is repetitive that it wouldn’t take much to put the final nail in the coffin.

And I told them what I’ll tell you:

When you are worried about your job being automated, you focus on the parts of your job that are the easiest to automate. When you are building software to automate a job, you have to focus on the hardests parts to automate. 

Take for example sizing steel beams - a task that appears to be repetitive and susceptible to automation. Evaluating the bending moment applied to a beam, and then selecting a beam with sufficient capacity may seem like a clear cut structural engineering optimisation problem. But the focus on mathematics masks other inputs. 

Consider an architectural requirement that the beam be of a certain cross sectional shape. This might appear to be a relatively simple problem compared with a beam sizing calculation. Maybe you don’t even think of it as a problem you need to solve. But consider what software would need to do to fully automate this process. Not only would it need to perform the optimisation on every beam, it would also need to communicate with the architect, understand any possible penetrations required by a building services engineer, have awareness of the connection at the end of the beam, and the list goes on.

So you might sit there and be able to imagine how software could automate the repetitive parts of your job, but when you are building automation software you need to consider how to automate the non-repetitive parts. And in structural engineering there  are a lot of non-repetitive parts.

A.I. will revolutionise structural engineering late 

An enormous amount of attention is given to the threat from A.I and machine learning. One nuance that is given less air is that the revolution will be uneven.

From a software development perspective, A.I. is two things:

It requires vast amounts of data. 

It is very expensive. 

These two constraints mean that companies that build A.I systems aim to solve problems that: 

A huge number of people regularly encounter and

Have digital records of how they were being solved before. 

Translating language, autocompleting search terms and targeting ads on the web all fit within these constraints. Structural engineering doesn’t. Yes, the field is being progressively digitised but it’s not something that a lot of people do each day. 

I have a simple benchmark for when this might change: A.I. is unlikely to have a large impact in structural engineering until driverless cars become widespread.

That might seem a strange link but in my opinion driverless cars will represent a watershed moment for A.I. both in terms of the scope of problem that can be solved and the level of responsibility we as a society are willing to delegate. Both of which need to be reached before A.I. touches structural engineering in a big way. So while you are driving yourself to the site, you have very little to worry about A.I.

You will see change coming before it arrives

No one knows what technological revolution will be next. But given the complexity of the structural engineering profession it is likely only a minority of the profession will be in flux at any one point, giving you plenty of time to adapt. 

And like any technology adoption it is both a risk and an opportunity. Those that move first will be rewarded, and those that move later could lose out.

So when you think about the question of when structural engineering will be fully automated, it is entirely possible the question is irrelevant. Maybe some practice of structural engineering will be unrecognisable in 20 years from how it looks today. But you have to ask: When hasn’t that been true of this profession?