AI and Engineering Design

[Melvyn Best]

I wonder if AI will ever be able to design a vehicle that can be easily worked on ?

 

[rtcosic]

I don't have a dog in this fight, but I cast my mind to the introduction of process simulation software in Chemical Engineering.

I used nomographs and manual calculation to arrive at a design. The process was so slow, tedious, and manpower-intensive (and costly) that I would accept the first viable design.

With the introduction of simulation software, I would run multiple solutions and accept only the optimum solution; however 'optimum' was defined. I used the same number of manhours as before, but to arrive at a 'better' design. As computer power increased and the software developed, I demanded an ever more granular series of solutions and my definition of optimum evolved to quite ludicrous degrees of precision.

I'm not sure what additional goals I could impose on an AI-assisted design, but I can imagine resource, time, and cost optimisation alternatives. Maybe add site-specific road, rail, canal or ocean transport limitations at the initial design stage and optimise around them.

I can't define 'elegant' engineering solutions, but I can recognise one when I see it - or reject one that isn't. With AI assistance, I would demand it of even the smallest project.

I would still use the same number of manhours but expect to arrive at an ever 'better' solution.

I'm a hopeless optimist.

 

[rtcosic]

I wonder if AI will ever be able to design a vehicle that can be easily worked on ?

If I were the Engineer in charge, I most certainly would have specified that as a design criterion!

Optimised for my fat fingers, creaking back and poor depth of field!

 

[RichardM]

Can't get a comfortable ride on the old bikes? New seat design for historic bikes? Bringing manufacturing back to Scotland.
Prototype next?

 

[Rannsachair]

Whilst I dont disagree with anything that @Screwdriver and everyone else has written above, the rhetoric around job losses sounds very familiar to the luddite movement of the industrial revolution.

As with any new technology, as it matures we will undoubtedly see a shift in the job market, but that has been going on since the start of time.

I can garuntee that all the jobs listed above have already been decimated by the use of computers and AI will just be the next step in that change.

50 years ago we had drawing offices full of people hand copying drawings, accountants with rows of people filling in ledgers by hand and couriers on bikes whipping between city offices with bits of paper. All of which are now long gone, replaced by CAD, Excel and E mail, and AI will just be the next step in that process.

Whilst I can't see into the future, I suspect that the limitations of AI will soon become apparent (its a tool not a solution) and there will always be a need for people to get their hands dirty, it may just look different to how it does today?

I started in engineering as a draftsman on the drawing board, learned CAD and saw 3D CAD come in then CADCAM and integrating FEA from the CAD model. I learned to use a slide rule and log tables, then a calculator then Lotus 123, then Excel then MATLAB and Simulink. I always felt I had an understanding of the fundamentals so I could understand what I was doing with software and had a good feeling if stresses/ strains/ pressures and loads were calculated in the right ballpark. I may just be too long in the tooth to see how folk in the future leap from competence in the fundamentals to utilising AI appropriately.

My last career was in HR, I found it relatively straightforward in that you need an understanding of current employment law, the organisation's policies and for most challenges you apply principles of fairness and want to be seen as a good/fair employer. We had introduced AI for our website for queries. and phone queries with the objective of cost savings in staff. I personally hate when i phone an organisation and it takes 10 minutes plus to get put on wait for a human. Some managers ask questions omitting key details to ensure they get the answer they want, the human aspect is knowing the organisation and knowing which managers to trust to ensure the advice is appropriate, so for now it is is just answering general queries rather than say "what do I do if I am restructuring my service to respond to the TUPE transfer of a local care home to us".

I am just happy to be retired

 

[Rannsachair]

Can't get a comfortable ride on the old bikes? New seat design for historic bikes? Bringing manufacturing back to Scotland.
Prototype next?

That may be why I spend most of my riding time on 1950's AJS and Matchless bikes: thick comfy seat, Jampot shock absorbers and teledraulic forks - what more would I want

 

[Pigeon_Droppings2]

I use chat gpt all the time these days...it's very handy for suggesting fixes to problems. With my broken dishwasher I told it the model and then exactly what happened...it then came up with possible fixes.

It's not always correct...but it's impressive. I'm trying to get back into electronics at the moment (as a hobby)...it's very handy for working out what components are doing.

I suspect a lot of outsourced admin type roles will disappear quite quickly over the next couple of years.

 

[bricol]

If I were the Engineer in charge, I most certainly would have specified that as a design criterion!

Optimised for my fat fingers, creaking back and poor depth of field!

Ha!

Problem is the accountant, or rather sales are in charge - because, you the public, don't want to pay proper money for a proper design.

So the engineering department has to design to a price, and that means it has to be easy, and quick to make. You want as much skill out of the build as possible, so someone off the street who just about has the ability to open a bag of crisps can follow the step A followed by Step B assembly process . . . but woe if the components for step C are late or unavailable.

Too many posts on this very site about that cheap crap . . . while bemoaning the loss of the factory that made properly designed stuff under a new Tesco/coffee shop/shoe-box housing estate.

 

[skotl]

In software engineering it's becoming well understood that you need to have a decent grounding in the subject matter and sufficient expertise to understand whether AI will help, hinder or even endanger what you're building. People with less experience ("vibe coders" - ugh) are more likely to build an untestable, unmaintainable mess with unknown behaviour in edge cases.

Don't get me wrong - AI is a productivity boon (I pay £200 a month for it and it easily saves me that in time) but without experienced input, it can be like letting an over-enthusiastic, arrogant school-leaver do brain surgery.

 

[Smallfry]

That may be why I spend most of my riding time on 1950's AJS and Matchless bikes: thick comfy seat, Jampot shock absorbers and teledraulic forks - what more would I want
View attachment 523605

Looking at that, looks like heated handgrips and boots wouldn't go amiss ?

 

[daleyd]

Looking at that, looks like heated handgrips and boots wouldn't go amiss ?

What you need is a thermoregulation layer and a lightweight carbon composite base.

Looking at the original designs again in the fresh light of day, it looks like it’s just attached some meaningless buzzwords to various parts of a diagram of a seat, rather than actually designed anything. It's like what a management team would come up with if you asked them what a seat should be.

 

[Gareth J]

Currently, ultimately, still needs an engineer to hold its hand.

If an AI designed widget fails and results in a hunt for blame, are companies really going to blindly trust what an AI comes up with?

From the other perspective; are AI facilitated design companies going to allow their designs to be used in the wild with little to no human input? Maybe one day, or maybe soon in china, or similar, where litigation concerns are much less of a priority.

 

[Rannsachair]

What you need is a thermoregulation layer and a lightweight carbon composite base.

Looking at the original designs again in the fresh light of day, it looks like it’s just attached some meaningless buzzwords to various parts of a diagram of a seat, rather than actually designed anything. It's like what a management team would come up with if you asked them what a seat should be.

If I had asked a young engineer to come up with some original ideas for a seat design, I would have been relatively encouraged by what was provided in terms of ideas that could be explored, I would likely have given him a day or so to do that: what came out of Chat GPT took seconds - and that was a free AI. I suppose it could get interesting if you fed the AI with the best academic papers on the subject and used that to answer the problem., I selected using TRIZ as it is a systematic process that I felt would be easier for an AI.

 

[Parm]

At work we were set a brief with a wide scope and room for plenty of interpretation. Then we weee told to as AI a question of our own design to address the brief. The topic is something we all know, the answers were amazing and highly technically correct

 

[Smallfry]

Currently, ultimately, still needs an engineer to hold its hand.

If an AI designed widget fails and results in a hunt for blame, are companies really going to blindly trust what an AI comes up with?

From the other perspective; are AI facilitated design companies going to allow their designs to be used in the wild with little to no human input? Maybe one day, or maybe soon in china, or similar, where litigation concerns are much less of a priority.

For the moment. Give it a few years and it will be self learning, and human input will not be required.

Driverless cars are here now, I wonder who is going to take the blame when it all goes wrong, and it surely will, or will it be regarded as a faceless corporate failure ?

 

[indykid]

If I had asked a young engineer to come up with some original ideas for a seat design, I would have been relatively encouraged by what was provided in terms of ideas that could be explored, I would likely have given him a day or so to do that: what came out of Chat GPT took seconds - and that was a free AI. I suppose it could get interesting if you fed the AI with the best academic papers on the subject and used that to answer the problem., I selected using TRIZ as it is a systematic process that I felt would be easier for an AI.

But both scenarios get you to a point where you need something better and more capable to drive the job forward. You only gave it to the junior Engineer as a development opportunity, not because they knew better than you or were the bike seat sage. I see that your AI enquiry has done the fluffy designer side of the job, but it hasn't output any of the engineering that gets it from nice sketches to financially viable production.

Just take the seat stiffness adjuster knob in isolation and consider the development needed to build that into a robust product. It's emperor's new clothes.

If you develop IP from AI generated content, who actually owns the IP? It might make for interesting times at the patent office...

 

[daleyd]

At work we were set a brief with a wide scope and room for plenty of interpretation. Then we weee told to as AI a question of our own design to address the brief. The topic is something we all know, the answers were amazing and highly technically correct

Really? When you look at first glance it looks kind of impressive - but what does any of it mean?

Elastic rear section? 3d printed composite lightweight base? Like I said, to me that just looks like it’s attached some keywords to a diagram.

Breathable fabric? On a motorbike seat? What’s being “breathed out” from underneath?

Looks impressive for sure but ultimately lacks anything in detail.

 

[lchris21]

I use it increasingly having been late to the party. As others have indicated, a level of experience is required to weed out any rubbish. Im a low level user and mostly use it as fancy search tool as I can put into words exactly what I'm trying to find out, without wading through various source's of information. It does it for me.
As English language and the use of it was never a strong point for me.........thats why i became and Engineer I'm finding it useful for drafting emails and correspondence to convey my point of view more professionally. Ive successfully used it to fight my corner in a few consumer disputes for various goods and services. Its remarkable what it comes up with, the regulations it can quote against, the technical consumer rights jargon it comes up with etc etc. The replies and results I've had off the back of said correspondence are in my opinion Ai generated themselves?........its a bit like robot wars

 

[DanZac]

Driverless cars are here now, I wonder who is going to take the blame when it all goes wrong, and it surely will, or will it be regarded as a faceless corporate failure ?

Coming from the aircraft world (engineering) where autopilot etc have been in use since the 1940's and has been more than capable of flying entire profiles for many years I'd trust the computers over the human interaction in most cases.

I'm hard pressed to think of an aircraft accident or loss that has been the result of machine failure. And the same goes for most corporate failings. There is always a human failing that leads to the loss.

Whether that be the pilots switching off the wrong engine or flying into the ground or the engineers fitting the wrong bits or some corperate manager making a poor decision, there is almost always a human at fault.

I'm not saying that driverless cars are there yet, and neither is AI but with further development both will be useful tools in the right situation in the future and neither should be written off because its a threat to jobs or doesnt fit the current profile or way we do things today.

 

[fizzy]

Why does everyone use ChatGpt? Gemini and Claude are better for many things. I use all of them constantly.