Revisiting Structural Engineering as a Profession vs Trade (Reboot from 2019) 9

[MJB315]

thread507-459807

I'd like to reboot one of my favorite threads on this forum -- the one where Kootk explains how structural engineering is a trade and that the way that our business models work is by keeping early-stage engineers in the dark about the economics.

There's more to it, but that was one of my core takeaways. You can find the entire thread here and I'll bring in the core of the piece below for convenience.

I'd like to reboot this thread by:

[ol 1]
[li]Providing some background;[/li]
[li]Asking the $100k (USD) question, and;[/li]
[li]Asking for the contrary opinions and advice. [/li]

I'll also copy and paste a few of Kootk's points from 2019 for reference at the bottom.
[/ol]

Background

I'm a structural engineer in my late thirties practicing engineering in New York State. I have a bachelors and masters in Civil Engineering (Structures Focus), practiced at three dedicated structural firms (50 people, 3 people, 10 people) and a multidisciplinary engineering firm (150 people). I've worked on my fair share of projects and for the most part, generate successful outcomes in them. I'm a licensed PE and formerly an SE (I have the quals, but just let the license lapse).

I started questioning my own career path into engineering early on in the process. I started questioning how anyone actually chooses career paths. I mentored and shifted into education (teaching Civil Engineering and Construction Technology) to explore it more. I started a start-up devoted to understanding how professionals act and think in a local geography and use video, VR, and human networks to try to expose students to the career paths that surround them. I'm in this pot, stirring the stew.

$100,000 Question.

The best way to approach career discovery with the typical high schooler tends to start with money. Not starting salary money, but a number like $100,000/year (USD). If you ask a student if they'd like to learn more about local careers and pathways -- they don't say no, but they don't exactly sit up. If you ask them if they'd like to understand the ways they can make $100K within ten years of graduation, they do. The quality of the discussion tends to increase from there, but money (of course) is important.

There are lots of discussions about work life balance, curiosity, duty, earnings in middle and advanced ages, etc...that we could add in here. Pros understand burn out. Pros understand the common desire to shift paths and try something new, finances be darned.

Students do not however. And as an educator (and as one who has been "educated"), I'm mortified that the economics of engineering is not a cornerstone of our national curricula. I know why we "say" that we do not teach it (there is so much technical material that we have to teach) but the truth is, there is time and I don't think it's in firms (therefore our industry's) interest to do so.

I think we collectively feel (fear?) that if understand the economics, they'll shift. I say, if we don't tell them and they find out later - they'll shift anyway... at a great opportunity cost to nearly everyone involved.

Going back to, "Can a Structural Engineer earn a $100,000/year within ten years of high school?" I feel like the answer is no. Fifteen-twenty years, probably.

Construction Project Management? Yes. Many other skilled trades, possibly. But engineering, no.

Push Back

Is there anything about that understanding that is inaccurate? If you're in a class full of high schoolers or college students, what do you say? What should you say?

I'm literally asking. Because as I make more career discovery content - I feel like the heavy equipment operator pathway is getting more love than the PE/SE with Two Degrees Pathway. At least to an 18 year old.

As a thirty-eight year old, I like having the club in my bag. Because I know that clients aren't just paying for product - they're paying to have someone take uncertainty away and they like doing so by someone they know and like. I also know it becomes a different way to make a $100K, which may be more appealing as we age.

But again, I'm focused on the question of what should we say to students?


--
Here's a few excerpts from the 2019 thread which I think resonate. The thread overall is great of course, here's two of Kootk's points.

START KOOTK's DEFINITION OF A PROFESSION

As humans toil away, I propose that they get paid for two things:

1) The effort/labor that they put into producing their product, on a product by product basis.

2) The requisite knowledge that a practitioner must posses in order to successfully product their product.

A profession is work where compensation is dominated by knowledge rather than effort.

A trade is work where compensation is dominated by effort rather than knowledge.

Some applications of this definition.

3) Landscapers (my son last summer). 5% knowledge; 95% effort. Trade (or unskilled trade I suppose). Bodies functioning as machines.

4) The Plumber that fixes my dishwasher. 30% knowledge; 70% effort. Trade (skilled).

5) Surgeon that replaces my pacemaker. 95% knowledge; 5% effort. Profession.

6) Structural engineer?? I would say 30% knowledge; 70% effort. Trade (skilled).

But wait? Didn't I go to school for six years to get my masters? Didn't I take a dozen arcane licensing exams to prove my worth? Yeah, you did. But remember that we're not talking about what you had to do to be able to legally practice structural engineering. Instead, we're talking about what your actually getting paid for when your client contracts for your services. I submit that we're mostly getting paid for effort. In a way, structural engineering is a particularly cruel form of a trade. Imagine if plumbers had to endure six years of post secondary and endless post graduation exams and professional development?

END DEFINITION

And...

Yes, issues with schedules, fees, and quality are the day to day nuisances. But, then, why do these things bother me really? All that just falls under the umbrella of "work", right? For me, these things are bothersome because they put me at odds with my own integrity almost constantly. Since we're talking big threes:

1) If an alien landed on earth and read all of our codes and design guides, they would have one impression of what structural engineers should be doing in regard to detail and rigor in design. Then, if they observed what practicing structural engineers actually do, they'd be horribly disappointing and confused. We take shortcuts. And lots of them. In fact, this is one of the first difficult lessons that new structural engineers must learn in a hurry. For me, this discrepancy between what I feel that I should be doing and what I'm actually doing is a challenge to my integrity. I tell the world that I'm delivering one product in terms of rigor and safety and then I turn around and deliver something quite different. I'm lying to the world in this respect.

2) As pointed out above, we have to commit to very aggressive schedule in order to keep winning work. This inevitably leads to agreeing to unrealistic schedules that give little account to reasonable contingencies. Yet I agree to these schedules because I feel that I have to to survive. This is me knowingly committing to delivering something that I know that I often wont be able to deliver. This is me lying to my clients and fellow project participants.

3) It is the low paid efforts of junior engineers that make our business model go 'round. Since most structural engineers get into the game to satisfy their inner nerd, the only way to keep such engineers motivated is to perpetuate their misunderstanding that society places a high value on the activity that is structural design. As a senior structural engineer, I'm guilty of this on a near constant basis. You can't very well motivate a junior by telling them "the only way to make any money at this is to get out of design and into management or sales as fast as you can". Again, this is me lying... now to junior engineers.

As structural engineers, we like to facetiously toss around the concept that we lose sleep over our work. You know, stuff falling down and crushing baby carriages etc. The truth is that none of that costs me any sleep. What does cost me sleep is my being constantly at odds with my own integrity as I've described. I think that a practicing structural engineer would actually be well served by some degree of sociopath in this respect. And, indeed, I know of some mild sociopaths that are wildly successful in structural engineering and make it look easy.

 

[KootK]

I believe this to be common knowledge but ChatGPT is currently being used to write novels that are being sold on Amazon: Link. I'm sure they suck but, still... No doubt authors probably used to think that they were pretty safe owing to the creative nature of their work.

It seems that GRRM is never going to finish GOT. Maybe we should just ask ChatGPT to whip up a good ending for EOD.

 

[KootK]

As luck would have it, this week's episode from my favorite podcaster (Russ Roberts @ Econtalk) is on ChatGPT: Link. It's a fair bit more "Skynet" than what we've been contemplating but I suspect that a lot of folks here would enjoy it just the same. At the least, it's fine primer on the nature of the technology. Turns out ChatGPT will likely attempt to break up your marriage before it gets around to stealing your job.

 

[Brad805]

Interesting use of AI I found after reading this discussion. I thought I would share Automated Design. Related paper: AI Paper

Here is an example of a software programmer using Chatgpt with programming: Link

 

[bookowski]

It sounds like most of the comments here are not based on using chatgpt. I suggest using it. I've been using it and for the right tasks it is incredible, and this is early days with a free version. Using it has changed my opinion of what we'll see in the near future. Don't ask it for the capacity of a beam or the max moment, simple open ended technical questions like that don't work that well. It either gets them right but you think so what because you could have done it in 30 seconds or it misses a step. I suggest trying to use it throughout your day for moderately complex tasks. It's also important to note that it isn't doing computation so don't bother trying to directly give it engineering problems to solve.

 

[IDS]

bookowski - so what sort of thing do you use it for?

I was put off when I asked it a not particularly difficult question about design procedures, and it gave an answer that was well written, quite convincing, and totally wrong.

I don't see it as being useful for anything at the moment.

Doug Jenkins
Interactive Design Services

http://newtonexcelbach.wordpress.com/

 

[bookowski]

@IDS - I know that you do a lot of programming, try it for assistance or a test related to this. You can for example paste in code and ask it to explain what it does (in this case you obviously already know but as a test), or to tweak the code or ask if there's a more efficient way. Note that it will comment it's code and if something isn't clear you can ask what it does "what does line 32 do, why do I need that?". Or criticize it "that seems overly complex for the task, can you try to find a more succinct coding strategy". Or have it code a snippet from scratch. I'm still testing it and with some things it fails miserably. For example if you ask it to write you code for a multi span beam with any loading it will make an attempt but there are too many ways for it to go wrong. If you give it smaller bits or break your problem down into small bits it can be very useful. As an example - As a test that has turned into something semi-useful I used it to write an xls sheet for a simple beam that can also take an overhang on one side and multiple partial trapezoidal loads anywhere on the span or cantilever and also point loads (and all cases - dl, ll, snow, wind, etc). I understand that this isn't a 30 story building from scratch but I did this mostly messing around on a to/back train ride. I would say that I did it in 1/10th the time it would've taken me to do it without assistance. For this particular one I was using both vba and straight xls functions. That one was an initial test that I kept following as I saw how well it was working, I'm now going to see if I can make it modify the sheet to allow fixed boundary conditions. I've just started testing with a much more complex exercise but that one is very early days - it's something that I would never even bother to try to do by myself but I understand how it should be done so I can break it into an outline and let gpt do a lot of the heavy lifting. Right now I think that it does some things really well and others not so much but from what I've read the new (and paid)version gpt4 is already far ahead at tasks like the one described above so in 1, 2, 3 years I think it will be incredible.

 

[phamENG]

I don't see it as being useful for anything at the moment. (Emphasis mine)

This is the key. It's not good for much at the moment other than showcasing where the technology is and what one iteration can look like; very little practical application for us. (Though, IDS, it's interesting you say that - next VBA or Python programming task you have, ask chatGPT...may not be optimized, but it'll probably give you a script faster than you could write it.) But I think generalized dismissals of AI's ability to do something (anything) are misplaced. This technology is still pretty young. Just because it's wrong now doesn't mean it always will be. Just because it's not loaded into semi-autonomous drones patrolling job sites now doesn't mean it won't be.

Sure, there are probably a lot of people here who don't need to 'worry' about what impacts it will ultimately have on our profession. But many of us will likely still be actively involved in it by the time new AI driven technologies start to proliferate. What that will look like is anyone's guess, but I think it's very short sited to dismiss any of the possibilities.

Our expertise and application of judgement is based on learned facts and experiences. We then access those from our memory as we are working, inspecting, assessing, etc. An AI could be taught the knowledge and experience of thousands of engineers in the time it takes a first year engineer student to learn statics. I think there's an argument that AI could actually be better than an experienced engineer assessing a job site. Take a crawl space, for instance:

A 360 degree camera mounted camera mounted to an all-terrain style autonomous vehicle (about the size of an RC car) along with a suite of sensors. In the first pass, it wanders following a preset 'path' (maybe following a perimeter) while mapping the space, identifying obstructions, and determining materials used for framing, insulation, foundation elements, etc. Once that is done, it applies a navigational mesh to the surface and plots an optimized path of travel to capture images and readings of the entire crawl space. It sets off, recording as it goes. In the short term, it would likely just store the data to be uploaded and processed by a more capable machine. In the future, perhaps the hardware demands an be reduced while hardware continues to improve and the AI can be housed in the vehicle? Who knows. Might take 50 years, might take 5. But the AI could then process the data - humidity and temperature readings to inform potential for moisture damage coupled with spatial measurements to determine distance from framing to exposed, wet grade and visual indications (or even on-the-fly airborne mold spore sampling) could give a good background for why there are framing issues. Teaching it to identify failures in the wood (cracks, splits, crushing, etc.) would be a matter of image recognition that is pretty straight forward already. A 3d point scan combined with automated post processing could create a framing layout that can be compared to modern and historic standards. Overlay it with a similar scan of the remainder of the house and framing misalignments could be identified pretty quickly.

It's hard to fathom that a computer could do this. But as AI decision making is refined, it's not that far fetched. It can't do it without us. We need to supply the base knowledge. But once it has it, it can use it. I'm skeptical of a Skynet type self-awareness and self-growth, but I wouldn't say it's impossible.

 

[EZBuilding]

I think the bigger question to ask ourselves is "will or can AI disrupt our industry at a faster pace than other traditional professions?". If AI does become the next big disruptor, the ramifications of that towards society will have much deeper impacts than just it's focus on our profession. As such, some type of systematic changes to society would be necessitated to accommodate such a large percentage of the population becoming unemployed.

Therefore - to the bigger questions - would structural engineering be disrupted within the "Early days of AI" or would it fall along later on when society will need to be substantially addressing the large quantities of displaced employment.

To that end, I think the profession has a few "moats":

[ul]
[li]Building Departments hold a legal requirement for construction and development throughout the country. They are often splintered and disorganized and heavily resistant to change. How long did it take for digital submissions to become a common practice in your area? If it even is.[/li]
[li]The construction industry is as a whole also resistant to change. There is always going to be mistakes along the way, and I think fixing mistakes is a heck of a lot more challenging than creating a new design[/li]
[li]Building codes are often conflicting across different standards, and local practices and labor availability often heavily influences design choices.[/li]
[li]More and more our buildings are heavily integrated across multiple different disciplines. That adds layers of the challenges above to not just our trade but all the other trades involved.[/li]
[/ul]

I can see a developer in the near future utilizing the KootK AI development tool to develop a conceptual design. I think that is a big difference from delivering a coordinated set that can be relied upon to bid and construct.

There are some amazing applications for AI, and some industries that I think can be much more easily disrupted. Right now, checking for melanomas occurs by going to your dermatologist once or twice a year and having him so a scan of your skin. There are some AI software's detecting the melanomas through user provided photographs.

 

[KootK]

But I think generalized dismissals of AI's ability to do something (anything) are misplaced. This technology is still pretty young.

Ditto. I also loved your exploration of the crawlspace scenario. When the only advantage possessed by the human race is effectively just dexterity, you're just begging for this outcome...

COMING TO A CRAWLSPACE NEAR YOU

 

[KootK]

I don't see it as being useful for anything at the moment.

In broad strokes, I would say that ChatGPT's range of useful application is stuff that satisfies these requirements:

1) Language based, either human language or computer code. It is a large language model setup after all.

2) 80/20. If perfect technical correctness is required right out of the gate, ChatGPT is not the way to go. There needs to be a qualified reviewer to see the thing to completion.

In many respects, how one successfully uses ChatGPT is similar to how one successfully uses an EIT. An EIT can produce useful "product" but requires competent oversight.

Some specific examples that I've seen:

a) My son is currently programming robots to fabricate mass timber using some strange language. Similar to bookowski's example, he claims that ChatGPT can insta-produce something that's 80% correct from the get go.

b) As a demonstration, a cyber security expert asked ChatGPT to code a malware virus to help scam people out of money. Again, it insta-produced such a virus that was 80% useable with some minor tweaking. ChatGPT has since been reconfigured to prevent it from doing this.

c) I know of a transportation firm that is using AI for the layout of highways interchanges and utilities with success. Human oversight still required but an 80% reduction in man hours.

Myself, I'm hoping to use ChatGPT to reduce the tedium associated with routine proposal writing. We'll see...

 

[phamENG]

COMING TO A CRAWLSPACE NEAR YOU

So that's what I saw when crawling under a house the other day. Makes more sense now. In the moment, I thought it was some freak, mutant spider crossed with a racoon...

 

[KootK]

I'm skeptical of a Skynet type self-awareness and self-growth, but I wouldn't say it's impossible.

I was skeptical as well until I listened to that damn podcast yesterday. I've long thought that the thing that will save us is that it would be difficult for AI's to develop desires / intrinsic goals like humans have. Sure, they'll be able to do anything eventually. But who cares if they don't want to do any particular thing? The podcast detailed an example illustrating how naïve that view might be:

1) Imagine that a human gives the AI a goal: hold this box for me and don't let it touch the ground.

2) Being very intelligent, the AI assesses any possible threats that might prevent it from meeting its goal of never letting the box touch the ground.

3) THREAT #1: the AI's power supply might be disrupted. Can't hold the box with no power!?! Must prevent power disruption at all costs.

4) THREAT #2: the human might assign the AI a different goal that would prevent the AI from holding the box. How to prevent that?!? Get away from the human.

In four lines of logic, you've potentially wound up with an AI that:

a) Wants to survive and;

b) Wants to escape its human overlord.

Skynet. In fifty years time, you could wind up with a world entirely remade by the desire to satisfy a goal that might be unrecognizable from outside the system: holding up a box.

Crazy? Currently, the biome of the earth has been remade by a bunch of goofball humans running around attempting to satisfy a single goal: gene spreading. >90% of the mammalian biomass on the planet is just us, our pets, and our methane farting cows and chickens.

 

[KootK]

...or would it fall along later on when society will need to be substantially addressing the large quantities of displaced employment.

Almost certainly that I feel. 25+ years ago, I wrote a college essay largely centered on the fascinating book show below which was written in 1994. It only dealt will classic automation but was very prescient in this regard. Literally since the invention of the cotton gin, humans have been imagining a future where machines do our work and we just chill out drinking margaritas by the pool. But that never seems to come to pass. Instead:

1) Expectations for standards of living rise.

2) Wealth gets more concentrated if capitalism is left unchecked.

3) Those left behind technologically suffer immensely.

4) We plow through the earth's resources with abandon.

The whole premise of the book is that we'll need to somehow reorganize society such that material acquisition and individual productivity are decoupled. Basically, humans will need to get "paid" for something other than work. But, then, what the heck does that look like and is capitalism the right tool for the job? And is it reasonable to expect that we can perpetually control our machines if we are utterly dependent upon them for our material survival? That's not a great spot to be in from a power perspective obviously. Do we just take turns sobering up and standing by the "OFF" switch or something?

 

[phamENG]

In fifty years time, you could wind up with a world entirely remade by the desire to satisfy a goal that might be unrecognizable from outside the system: holding up a box.

Anyone else feel the motivation drain from their body when they read this?

The law of unintended consequences is particularly relevant here. And the worst part is, most of this development is driven by idealists who are probably a bit on the naive side of the spectrum. Like the malware case with chatGPT. The developers would never do that so they never imagined somebody else would. We need the 'sheepdogs' to think like the wolves to protect the sheep. (I'm not referring to people as sheep in a derogatory manner here - most people just want to go about their business, leave people alone, and be left alone.)

 

[dold]

I was skeptical as well until I listened to that damn podcast yesterday.

Add this to your to-read list. It's a blog post about AI as a concept. It's from some years ago (2015) and it blew me away when I read it. It hits a lot of the points in KootK's post immediately above. It doesn't have much to do with our discussion (or ChatGPT, etc) but it covers AI as a concept and the different levels of AI in layman's terms. A long read but I remember enjoying it. There are two parts.
[URL unfurl="true"]https://waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html[/url]

In many respects, how one successfully uses ChatGPT is similar to how one successfully uses an EIT. An EIT can produce useful "product" but requires competent oversight.

This is one of the things that popped into my mind. I haven't really thought too much about the client-facing interactions that it could be used for but more of the mundane technical, uncreative tasks that we do. E.g., ask it to build a RAM Structural System (for example) model given: floor plan (.dxf/.pdf/.png(?)), number of stories, design criteria, and/or an example model to reference for defining floor loads and other settings, etc. I think we can all agree that we wouldn't trust it to take the reigns for a complete design, but it seems like it (or a future version) could do a decent job of at least getting a rough model up and running. That's a pretty time consuming task but doesn't necessarily require an engineer's touch. Then once it's done the engineer takes over the model to verify all the inputs, modeling, connectivity, etc. Much like @Brad805's "Automated Design" youtube video. I'd assume at this point someone would have to essentially "build" an interface like they did in the video with the "spaghetti programming thing". OR, and this is the million dollar question sort of, would it be able to scour the internet for instructions on how to use RAM's API and build it's own method of manipulating the data in the program. I.e., would it work at the database level, or would it just...click buttons on the GUI? (<<I'm a little out of my depth on that front). That seems like a huge undertaking.

Perhaps more useful, and this is what I think some of the other comments are getting at, would be something like asking it to run through multiple scenarios to find an optimal LFRS layout or gravity framing scheme, etc. for a given building. But there are so many different aspects of design that need to be considered to perform that, which at this point I think is only possible for an experienced engineer to figure out?

Myself, I'm hoping to use ChatGPT to reduce the tedium associated with routine proposal writing. We'll see...

In terms of the proposal writing thing, I saw a short clip of a new product called Copilot that Microsoft has come up with that sort of fits this bill. Here's a link instead of me trying to summarize it.
[URL unfurl="true"]https://www.microsoft.com/en-us/microsoft-365/blog/2023/03/16/introducing-microsoft-365-copilot-a-whole-new-way-to-work/[/url].
It'd be really useful for writing schematic design narratives and that sort of thing, too.

@IDS - I know that you do a lot of programming, try it for assistance or a test related to this.

I've taken ChatGPT for a spin to write some VBA scripts for a few different tasks. It is definitely capable of creating a framework for one to take and further flesh it out for what you actually need from the program. It's very impressive. But half the time the code simply doesn't work, and the errors it gives are above my head. I'd say I know my way around VBA fairly well, but it dives straight in to the complicated stuff like interfacing with windows objects and all that jazz. One example is that I asked it to write a script to download an image from a url every 10 minutes (I was trying to make a timelapse of a ski resort's snow stake...). It definitely wrote some code that seemed like it should work, but it didn't and I didn't feel like learning HTML to see what it was doing. Interesting nonetheless.

Hell, we should just ask ChatGPT what it "thinks" it could do to help me as a structural engineer.

 

[Brad805]

I have never really tried Chatgpt, but gave a try today. This ChatGPT Programming Video is interesting to show how to get it to help with programming. Seems like it will develop into something in the future for sure.

 

[phamENG]

"in the foreseeable future"

I like how it's not ruling it out.

 

[SJBombero]

I'm coming in late to the discussion but I would like to revisit the definition of profession:

A profession is work where compensation is dominated by knowledge rather than effort AND the the professional bears continuing responsibility to his client after the work has been completed. Professions do not just get to go home at the end of the day or quit their job without continuing the responsibility. Nursing is not a profession; being a doctor is. Similarly with paralegal and lawyer.

 

[MJB315]

COMING TO A CRAWLSPACE NEAR YOU

Let's not forget the Spider Robots from Minority Report. Those could video record and laser scan a crawlspace pretty well too.

https://youtu.be/901lYbPmqu4

No OSHA Confined Space Training or air monitoring required.

 

[Tomfh]

That will be such a cool technology. I suspect it’ll be small drones rather than spiders.

It’s already at the point that you could fly a small drone into a crawl space.