Is AI gonna take our jobs in traditional engineering fields like power, civil, and structural? Not fully, at least not yet. We humans are still pretty important behind the scenes.
Let me explain why, using four points and a few examples from my own work experience. Just a heads up, when I say “AI,” I’m not referring to artificial general intelligence (AGI), but rather, to the narrow AI that’s honed in on design work in conventional engineering fields.
The first point to note is, the more parameters we feed into an AI, the better we can teach it to do what we want. After all, every task we perform is a collection of information gathered by our biological senses that our brains process.
#1 A crazy amount of variables
When an AI takes on a real-world design, it has to turn the real world into a digital one. This might work for large projects but it’s expensive and not very feasible for smaller ones. Even when this is done, the software doesn’t deliver a complete project. It only hands over a piece of the design puzzle. It’s like getting a 3D model of just the water pipes for a water pump station, both above and below ground, and nothing else.
We almost always require different software for various tasks to compile a complete project. Then there are factors that a human might consider based on experience, which a computer can’t simulate. The real world is simply too unpredictable and nuanced for a computer to grasp single-handedly. For instance, in an area prone to earthquakes, with issues like liquefaction and heavy traffic, a unique duct bank design is needed. This requires innovative thinking to suit a customer without over-engineering while considering everything that can go wrong.
Now, imagine if there was software that could do everything. It’d be so complex and hard to use. We’re already struggling to create simple interfaces for our current AI systems. Trying to deal with a gazillion input settings without understanding the potential consequences is a recipe for disaster. It’s already pretty common for people to take what specialized software outputs as gospel, without thinking twice.
#2 Navigating the real world
AI heavily relies on real-world data, and that requires hands-on work. For example, if I’m upgrading a water pump station with an accompanying electrical building, here’s some info I’d typically collect from the existing facility:
- Ratings and physical conditions of equipment
- The degradation of electrical infrastructure, like conduits, boxes, wires, and terminals
- Workspace and clearance measurements
- Exposure to natural elements
- The current programming and control logic in use
- Likelihood of vandalism
- As-built drawings
Then there are customer-specific requirements and design elements that aren’t clear-cut:
- Preferred models of equipment
- Design standards that are more conservative than codebooks
- Designs following the customer’s internal needs, not industry standards
- Areas not available for design due to maintenance needs, environmental issues, future plans, etc.
- Additional safety measures
All of this involves a lot of back-and-forth between the customer and the engineer. It’s not just about collecting X, Y, and Z variables and then off to the races. You collect the data, analyze it, make recommendations based on design experience and calculations, and then discuss it with the customer again.
#3 The human touch
Humans are necessary to catch any errors machines may make. Take an airplane, for instance. Despite the many sensors on a plane’s body, they can’t replace a visual inspection. Imagine a small crack or a slight color change in the plane’s frame is hidden behind other components, and a sensor has failed, not signaling any alerts of its failure.
Without a detailed check, this crack could go unnoticed, potentially leading to a catastrophic accident.
Sure, machines usually outdo humans at detecting and diagnosing problems. But when human lives are at stake, the more safety measures, the better.
Personally, I would not fly if humans are removed from the equation. I believe in a collaborative effort between humans and machines. And frankly, there would be a backlash if humans were entirely left out of the loop, and quite rightly so. At the same time, considering how complex technology has become, I would also never rely solely on humans.
#4 Specific equipment design details
Clients often have their preferred vendors and equipment models. Each piece of equipment has its own idiosyncrasies that can’t be quickly fed into an AI system.
A human needs to comb through the product manual, understand the complex details of the software or hardware, and figure out how it best fits into the overall system design.
When I’m designing, I don’t just study the product and installation manuals. I also chat with the product engineers to clarify any ambiguous wording from the manuals. The process is rarely straightforward. There’s always a design element that doesn’t cooperate, and often, it’s a vital part of the whole design. Plus, some of these manuals are thousands of pages long. Imagine feeding all those specifications and parameters into an AI system.
Conclusion
Sure, you could construct an AI system with infinite parameters. While many design parameters like the distance between power and instrumentation wires to prevent electromagnetic interference are fixed, many are specific to the job. This would require an even more sophisticated AI system for detailed customizations for every job. It’s mind-boggling.
It’s probably more time-consuming and costly to feed the parameters into the system and verify the output, compared to manually doing it with existing specialized software for each design element. I already spend a lot of time scrutinizing the output of current software for specific studies. Because if you input junk, you’ll get junk, and I’ve encountered buggy software.
For the foreseeable future, I don’t think the technology will be ready, and I doubt people would feel comfortable entrusting their infrastructure entirely to machines. But the landscape of the fields is changing. Job consolidations are taking place and will likely accelerate given that we’re only in the early stages of AI.
Do you think AI will take over traditional engineering jobs anytime soon? What changes have you seen in your line of work due to AI?
SUBSCRIBE TO ENGINEER CALCS NEWSLETTER
