Future of Engineering Jobs & Automation 9

[StrucPEng]

Hello All,

I wanted to start a discussion on the future of the engineering profession in light of continued improvements in AI and automation. There is a lot of talk about many jobs being "outsourced" to AI or that can be completed with much less expense than before. A few that come to mind are truck drivers, customer service, and even radiology. It is a topic I have been interested in lately and wanted to get everyone opinion on how they see the engineering profession evolving in light of this. A few initial questions would be:

- What kind of engineering processes can and can not be completed with AI or could be automated in the future?
- Which branches of engineering are most/least vulnerable to automation? (Structural/Mechanical/ etc.)
- What does automation mean for licensing and duty to the public?
- Will there be less of a demand licensed engineers used mostly for confirmation of computer created designs?
- Do you see the value of an engineer increasing or decreasing in light of automation? Less

I am interested to hear everyone's thoughts on this. Thanks for reading!

Best,

Matt Soda, P.Eng

 

[drawoh]

Engineering is not and has never been "drafting and toleranceing".

Quite a few years ago, I was shown a drawing of a frame piece welded out of aluminium 6061-T6 angle. This was dimensioned in inches to three decimal places, indicating a tolerance of [±].005". I pointed out that the tolerances were stupid and impossible. It turned out that the mating piece was another set of welded aluminium 6061-T6 angles. One piece went inside the other and bolts were attached through the outside faces. Assuming this thing was not assembled with a rubber mallet, the bolts were closing a rigid gap comprised partially of annealed aluminium[ ]6061.

What was superficially a lack of comprehension of welding DFM by a CAD[ ]monkey, actually was a structural and safety problem. The material strain caused by assembly, caused stresses way in excess of anything the material could handle. Successful engineering required a knowledge of the manufacturing process, metallurgy and the assembly process, as well as the final functionality. Dimensioning and tolerancing of this frame was beyond the capability of a CAD[ ]monkey or a robot. Sometimes, the details matter.

--
JHG

 

[JohnRBaker]

drawoh,

I have a fully searchable FileMaker Pro database where I record a unique indexing number, title for the image, the image format (Slide, B&W/Color Negative, APS, Digital), the camera used, date taken (month & year), and a list of searchable keywords including subject, interesting content, names of any recognizable persons, location where the image was taken, etc. I also record if I've ever sold the image and how much it earned (I sell photos via a stock image service). I also indicate a DVD reference number, which are categorized by image format. In addition to having ALL of the images on-line (on an external 2TB hard-drive) I keep two sets of the DVD's, which number something over 140/set, one set in my gun safe here at home and a second set kept by one of my sons as a off-site back-up.

As I stated, the database is fully searchable using any and all fields in the record, which also includes a 200 x 200 pixel thumbnail image. And the searches are very fast; I can perform a multi-target search of all 43,964 images in less than a second. And the report I get is a list of the actual records showing all the content of all the record fields as well as that thumbnail image.

John R. Baker, P.E. (ret)
EX-Product 'Evangelist'
Irvine, CA
Siemens PLM:
UG/NX Museum:

The secret of life is not finding someone to live with
It's finding someone you can't live without

 

[zdas04]

drawoh,
The details often matter.but you described the perfect scenario for an AI--the engineer defines basic tolerances for each grouping of part and then defines the grouping properly (reasonable engineering activities) and then lets the AI apply the rules, they will be way more consistent than a human ever would be as long as the engineer has defined the groupings properly. If someone (usually a draftsman) picks arbitrary tolerances then putting then having an engineer looking at the groups and their tolerances is engineering. An engineer putting the tolerances on the drawing is generally a waste of talent.

[bold]David Simpson, PE[/bold]
MuleShoe Engineering

In questions of science, the authority of a thousand is not worth the humble reasoning of a single individual. Galileo Galilei, Italian Physicist

 

[Sparweb]

Drawoh,
I swear I was looking at that same drawing last week. What are the chances?
Well actually, in this case, a sheet-metal box bent up on all 4 sides, and meant to fit inside a lid, also sheet-metal bent down on 4 sides. Fit-up problems solved by putting a +0.005"/-0.000" tolerance on the lid and +0.000"/-0.005" on the box. No AI will ever touch that level of... efficiency. The shop knows better by now, and so they test-fit 20 lids on 20 boxes until they get as many fitting pairs as possible. They are getting the the point where they're afraid to inspect parts because the tolerance as ridiculous.

Edit- my point (besides bitching, sorry) is that this kind of routine error getting past human engineering checks & approvals is just the kind of thing an AI can figure out if configured to design for tolerances that the shop can actually achieve. So I agree that a point may come that this skill moves out of human hands, if humans aren't willing to learn it.

STF

 

[drawoh]

zdas04,

I knew two things here, not just one.

[ol a]
[li]Welding tolerances are something like [±]1/16".[/li]
[li]An eight inch piece of aluminium 6061-T6 cannot stretch 1/16", even if it somehow manages to be T6[ ]temper.[/li]
[/ol]

It generally does not take me a long time to prepare fabrication drawings. When I apply tolerances, I take a final close look at the design, and I ask questions. The frame required re-design, not better tolerances.

--
JHG

 

[LRJ]

I think the discussion of what constitutes 'engineering' is very interesting here. For me 'engineering' has always been the 'practical application of scientific knowledge'. I think with increased usage of AI and automation, it will naturally mean that what constitutes 'engineering' will evolve. So the 'practical application of scientific knowledge' is destined to evolve with the new scientific knowledge and new technology applications. I expect engineers will spend more time doing the interesting work which contemporary technology cannot quite handle reliably.

To achieve the productivity gains expected from the use of new technology, we - as a field of study - need to train the next generation of engineers to be able to utilise this new technology and also to empower them to take the risks which are necessary to succeed. Of course, given the tight margins in most engineering disciplines, such human resource development will be slow. The engineering companies that will succeed in the next 20+ years are the ones that will invest today; those companies scrimping after every penny today will be forgotten.

I'm generally quite positive about the use of new technology as I expect it will reduce the monotonous aspects of engineering (think how much time you spend copy+pasting, etc.) and I don't expect to be replaced by a robot.

To summarise: the Luddites were proven wrong, just as neo-Luddites will be.

 

[davidbeach]

I often find myself disagreeing with zdas04 but he is absolutely correct about “good enough”. That’s the transition point between a designer and an engineer. Many years in consulting honed that concept, but I still find it quite useful after a decade plus in the electric utility business.

I analyze events (among other things) and find that others either stop way too soon (mostly) or wind up with paralysis by analysis. There’s a middle ground of finding the interesting/useful results quickly without ratholing.

It also seems to be something immune to being replaced by some AI system. If we could have predicted that particular outcome we would have avoided it. Far too many variables for some AI to sort out.

 

[SkipVought]

Sitting in a freshman orientation at an all male engineering college in 1960, they told to us not only look at the guy to the left, right, in front, behind etc., but also....

the difference between a mathematician (scientist) and an engineer:
A mathematician and an engineer are instructed that behind the door is a room with a beautiful girl. At each one minute interval they will be permitted to halve the distance to a beautiful girl.
The mathematician laments, “I shall never reach her!”
The engineer exclaims, “I shall soon be close enough!”

BTW, my alma mater is no longer all male.

Skip,
[sub]
Just traded in my OLD subtlety...
for a NUance![/sub]

 

[CWB1]

The future of engineering is the future of technology. IMHO it will continue much like it always has, higher levels of technology will continually be pushed to lower levels and those companies and individuals at lower levels who cant keep up will struggle to survive or be pushed out of the profession. Many have already moved to mostly auto-generated/updated prints as mentioned, FEA and CFD analysis are being pushed from engineering into other depts, and AI is being used as the sanity check of the engineer rather than vice-versa. As time goes on I'm sure we'll see new niches develop and old niches become commonplace, I also wouldn't be surprised in the least to see some of our historically separate segments combine around common software and methods such as mechanical and structural design.

 

[bones206]

I've been following with interest companies like Arup who are adopting AI/automation technology to enhance engineering. Advancements in topology optimization and 3D structural printing have allowed engineers to produce some really cool alien-looking structures that would otherwise be impossible to design and construct (unless your name is Antoni Gaudí). I'm not afraid of being replaced by intelligent automation, but I can definitely see it becoming a valuable tool in the engineer's arsenal helping us push forward into new frontiers.

Examples: [link file:///C:/Users/p003488c/Downloads/Additive_Manufacturing_report_for_IASS_21015.pdf]Link[/url] Link Link [link file:///C:/Users/p003488c/Desktop/design-to-fabrication-fifth-hotel-city-of-dreams-macau.pdf]Link[/url]

 

[JohnRBaker]

Here's a video showing an example of design shape optimization, which is already available in at least one commercial CAD software package:

https://www.youtube.com/watch?v=-XFNT6GSxz8

John R. Baker, P.E. (ret)
EX-Product 'Evangelist'
Irvine, CA
Siemens PLM:
UG/NX Museum:

The secret of life is not finding someone to live with
It's finding someone you can't live without

 

[GregLocock]

If you think shape optimisation is AI then I've got a bridge to sell you. Nice to see 35 year old techniques actually getting traction in the structural engineering world (first automagically shape optimised FEA part I'm aware of was ca 1980).

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[JohnRBaker]

Perhaps, but these tools are now being integrated into the applications being used every day by most engineers. And while this may not be true AI, it does represent the next step forward in the merging of modeling and analytical tools, leveraged so as to give engineers and designers a more practical and easier to use development platform.

That being said, I would like to make one comment with respect to why this took 35 years.

But first note that model optimization has been creeping into the design world for some time. Granted, it would be misleading to refer to what has been available for at least 10 or 15 years as 'shape optimization'. Instead, we have had something we often referred to as 'geometric optimization' which was based on a combination of analytic and mathematical methods using the parametric nature of an initial solid model, however the topology of the final model was almost always unchanged, just that the size of the basic features were altered to meet some criteria, usually a combination of strength and weight targets. However, what has changed recently is the advent of 3D Printing, where it's now possible to manufacture components which just a few short years ago would have been virtually impossible to create. This is the real catalyst in this formula, which while it had nothing to do with actually making these 'shape optimization' tools work, it has made them practical in terms of being able to create models which can now be physically realized at a reasonable cost.

And lets not forget that having access to powerful compute engines at what, compared to 35 year ago, are basically 'supercomputers' at commodity prices, has also made many of these new design and modeling tools even possible for virtually any engineer working for a small to medium sized company.

John R. Baker, P.E. (ret)
EX-Product 'Evangelist'
Irvine, CA
Siemens PLM:
UG/NX Museum:

The secret of life is not finding someone to live with
It's finding someone you can't live without

 

[GregLocock]

If <100 lines of code is AI then everything is AI. I wrote my own about 17 years ago as the commercial ones at the time that I had access to didn't deal with beam elements. Optistruct must have been available long before then, fully integrated into Hyperworks, which is a program used by real engineers that runs on PCs.



Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[GregLocock]

Here's a bit of fun. Somebody puts a bit of code together that works out where the boundaries of a property are based on a verbal description and a satellite map. The Mississippi board of engineers decides that this rather banal piece of code is doing an engineer's job.

https://arstechnica.com/tech-policy...-on-a-map-is-that-the-same-as-land-surveying/

Well I guess we've now defined the sphere of competence for Mississippi Engineers...

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[drawoh]

GregLocock,

I like how the software is described as "drawing polygons on satellite photos". I would guess that at some point, surveyors are expert witnesses in a court of law. The profession requires peer review and regulation. Practising surveying has to be perceived by the courts as "yes you are", or "no you are not", in spite of the fact that there is a grey area. This software definitely is on it.

I observe that the lawyers taking this case on are libertarians. We have recently had this discussion. thread765-424265. Hard core libertarians will argue that professional engineers and surveyors societies are guilds, that they are conspiring against the public to maximize their incomes, and that the free market will provide protection to the public far better than government regulation.

--
JHG

 

[IRstuff]

Oh, jeez, I've been in violation of that law every time I put down KMZ or KML entity into Google Earth! Yikes!

But, then again, has Zillow or Redfin been sued by any state? They do the same thing on every single property viewed every day. That's got to worth millions in a suit.

I think the Mississippi board has made a serious mistake. By claiming that a computer program can do surveying, they've essentially affirmed that their entire purview is bogus, since it means that anyone could survey and they don't require any special training, testing, or other qualifications to do so. This essentially nullifies their raison d'etre, because they've essentially posited that it's not possible to regulate nor enforce the law.

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[GregLocock]

Yup, that was my take. I brought it up because it is AI to some extent (parsing verbal descriptions and trying to reconcile objective (GPS) locations with traditional measurements from uncertain baselines) while on the other side of the fence are the buggy whip manufacturers trying to defend their monopoly in the age of autos. Having been through the process of trying to get my local land office to define exactly which bit of the ground my house is on is owned by me, I have some cynicism when it comes to dragging surveying into the 21st century. FWIW I was on a team that established the calibration points for the first British GPS receiver, so I'm not exactly clueless when it come to surveying.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[IRstuff]

I think there's still possibly some merit to classical land surveying, i.e., going into the field with DGPS/RTK, theodolites/transits, chains, and poles, to do actual measurements. That requires at least some level of training and skill and about $30k worth of decent equipment. Most satellite imagery isn't positionally accurate enough to challenge easements or boundaries, particularly on large parcels, as there's some level of distortion and error in the georectification process. Google Earth's accuracy falls rapidly the farther you get from the 1st world. There were a bunch islands in Polynesia or Micronesia that formed their own ground control point surveying database, because Google Earth is as much as 10 meters off on in those locations.

Mississippi's law is overly broad, as they attempted to capture any possible way of interpreting "practice of surveying."

TTFN (ta ta for now)
I can do absolutely anything. I'm an expert!

https://www.youtube.com/watch?v=BKorP55Aqvg

faq731-376 forum1529 Entire Forum list

http://www.eng-tips.com/forumlist.cfm

 

[JohnRBaker]

WARNING: This is posted purely for it humor value ;-)

This discussion concerning whether you need to physically survey something so as to know exactly where it is, this reminds me of the book 'Lake Wobegon', by Garrison Keller, and his explanation for why people can never seem to find the village of Lake Wobegon on any maps of Minnesota. The item below gives a somewhat abbreviated story of Lake Wobegon, but with respect to the issue of not being able to find the village on a map, that's covered in the 4th paragraph. BTW, a much more lengthy explanation is include in the book which goes into excruciatingly hilarious detail as to exactly what led to the 'error' that resulted in Lake Wobegon being so hard to find today.

http://www.garrisonkeillor.com/national-geographic-in-search-of-lake-wobegon/

John R. Baker, P.E. (ret)
EX-Product 'Evangelist'
Irvine, CA
Siemens PLM:
UG/NX Museum:

The secret of life is not finding someone to live with
It's finding someone you can't live without