@@BrendanHasty There are pros and cons to this. I work on a lot of projects for universities and government clients, all essentially funded by tax dollars. Architects like to make these projects very complex, building monuments to themselves at the expense of the public. The complex projects cost more, take longer to build, have more chance for error, etc. Would society be better off if we built a cube instead of a complex structure? Would a city be better off with a larger hospital that is not as pretty, or a smaller hospital that pushes the boundaries in aesthetics? I don't know how to answer that question. Governments are usually challenged in funds and if those funds were focused on employees they might serve society better. I dealt with one university group that was forced to seek the lowest rent cost for their offices so they were in the basement of a bowling alley. It was not as bad as it seemed because the offices were nice inside and they had their own side entrance, and they could rarely hear noises from the bowling balls, but more of the university funds could go towards students that way.
@@billj5645 I can see this. Architects need to get away from riding that high and realize that just because they can doesn't mean they should. In my own experience, I see this a lot coming from inexperienced people thinking they have the business right to do whatever without considering the implications of their decisions, but there's nothing you can do to pull them back from the ledge because they don't have that experience.
I can agree with you there, my dad is an Engineer in my country, very well known throughout the country for bringing down the cost of Bridges he built and buildings by just using Precast Concrete Structure. These beautiful bridges and concept design brings up the price of these buildings by alot and in a developing country like Belize thats not really necessary. @@billj5645
Hi Julian, I have the same concern. The AI will start be reducing the more simple task where junior engineers learn. We as a profession need to see this and train our junior engineers if we don't our long tern outlook will not be good.
Dear Brendan, Technology has always helped structural engineers expand the horizon of what was possible. AI like any other technology has the potential to push the boundaries to the extreme. I also believe that the number of Engineers and draftsman required for a project could be reduced further, on scale similar to what happened when computers were started to be used for calculation and drafting. This change in scenario demands the engineers to be stronger in the basics than ever before.
Well very well thought out explanation Brendan. And to add what you have said, it take even more time for widespread use of AI commercially to its full extend. I would like to know more about that 6:42 screen set up.
Thank you very much! Hearing this I now know that with AI coming I need to actually step up my game and get all the software updates learned. Still a student, but in my late 30's.
Thanks for the video Brendan! U r calming us all :) the only thing is that the computational power is exponentialy higher than before, if u tried to compute chess today it would be not that long for it to "replace" grandmasters. (but YES, chess is still not SOLVED yet). Have a nice day!
Hi Mariansykora glad that the video helped calm some of your fears. Agree that computers now days maybe able to brute force the solutions, it does require a different approach to replace structural engineersm
What a perfect video to watch when I'm just about to begin my Masters in Structural Engineering 😁😁, Any good suggestions are welcome from anyone, how do I utilise the most out of coming 2 years of my Master's, what should be my chief focus?
@@Delasi-mk9xz Nope full time, graduated from undergrad last year, took a year gap to figure out what I wanna do ahead and came to conclusion that I wanna engage in Structural Designs and hence quickly applied to good National Universities in my country and got through one and starting out the course in 10 days
Hi Brendan, interesting video and on AI. However, I believe the roles of engineers and draftsman will significantly reduce in due time. Technology is the way for our planet to move forward but we must consider glitches that will appear time to time and that's where we as engineers will need to step in to make human decisions, site visits and interacting g with clients and other consultants.
Hi Malik, it will but likely will take a longer time to occur. AI needs to be trained but will it yes, as if someone comes up tool can reduce engineering time everyone will buy it. We will still need to validate the solution
If you are a residential builder...the sooner the better! It seem most engineers prioritize larger projects. Ive been waiting for almost a year and still haven't found an engineer to review and stamp a set of plans for a residential permit application which is now a requirement which is putting the project on hold.
At the end of the day, they need someone to take the legal responsibility. Our fees & salaries aren't the highest so I don't see the huge economic impetus to replace us with AI. Many in our industry are having a hard time even moving beyond pencil and paper so I think it may take a lot longer to implement AI in our field than many might anticipate.
Oh, this industry moves slow and it isn't a bad thing, but definitely the pencil and paper need to be careful because shifts in the competitive landscape could easily put them out of business. That being said, the software engineers cannot and should not be taking the liability of the structural engineer, but can refine their position with increasingly thorough testing.
There's a lot of wishful thinking about AI and projection of human abilities onto AI in this video. You can't generalize Human + AI as better, it really depends on the task. Particular to (good) engineering I'd bet on Human + CAD&FEA for many years. This is because of how the wave of new (generative) AI models actually work. The simplest way I can summarize: generative AIs are all "fake it" with no "make it". Humans can learn and work extra hard to compensate for small fibs. AIs don't learn, not until the next software version is released, and not in the same way as humans. AIs don't understand the never-experienced real world; AIs just process words/tokens on a page or pixels. AIs mimic and fake, because that's exactly what they're trained to do. When training, you train both the generative AI and a discriminating AI. Based on a prompt, the generative AI makes something up and the discriminating AI guesses "which is real?" between the made up response and the actual response. The discriminating AI is slightly tweaked to better categorize real as real and fake as fake, while the generative AI is slightly tweaked to *better fool* the discriminating AI. These small tweaks are repeatedly performed ~million times. The discriminating AI can't do in-depth analysis/checks like a proper engineer could (for technical reasons) thus the generative AI never needs to actually learn past generating stuff that kinda looks fine at a first glance, i.e. well-formatted fakes.
It is still many years away the skills learnt and an engineer you can transfer you knowledge to many different fields. But always good to look into the future.
I've been in structural engineering for a long time. There are very many career choices that are better in one or more ways than structural engineering. I've seen a lot of people leave structural engineering for other careers. Some of us just like it and don't want to consider alternatives.
@@billj5645 That's great as it sounds like you've found your inherent calling which I'm all about. I think for Boris though it is scary just starting out and seeing the level that some of us are playing at. It took me 12 years to get to this level between college and industry experience, and even then there's only so much inefficiency one person can address on your own. That just goes to show you just how much money is available.
Bill Baker, the man who designed Burj Khalifa has a slightly different opinion. He is of the opinion that our value as structural engineers has always been and will always be by Understanding structural behaviour and the underlying theories, physics and the mathematics. He always recommends the young students to take as much as theory as they can. He opines that after a decade or two from now, we'll be in the post computation age when computation is going to be trivial because we're not going to be pushing buttons anymore. AI or some super program will do it for us. It'll provide solution to us for a given problem on its own. So, what is our value as structural engineers going to be ? Our value as structural engineers in the upcoming years will be by Understanding the "why" why that's the solution and by Understanding this we will change the problem itself to end up with a better design. This understanding comes from the structural engineering theories. The human ability to change the problem will always allow us to stay relavent. We had certain boundaries when we didn't have the computational power. After having the computational power, we pushed our boundaries. When we will have AI, we will push our boundaries even further so, in this way structural engineers will be able to be relavent even after the advancement of AI by harnessing it to push our boundaries even further instead of just automating the grunt work. Also, he says geometry and topology of a structure are way more important than sizes of its members. We can save upto 60% material of a structure just by having the right geometry. In a nutshell, our scope is going to be much wider than it is today
Hi totally agree we will be more efficent, and our role is understanding "why" the problem, and be able to solve these issues. AI does need to be trained, and we will also need to validate the solutions produced. It should allow us to push boundaries.
@@BrendanHasty Yeah I agree and I feel as long as we push our boundaries, we will be able to survive indefinitely. It's like "Survival of the fittest". We have a lot of problems to solve in the upcoming years, we have growing population especially in the countries like India, climate issues, limited resources, limited land area and a lot many contraints alongwith the problems posed by the architects ( who themselves will push their boundaries)
Even as an outsider, I can definitely agree with this perspective. I think it'll take a lot longer to get there, but we can progressively dial things in so that we get further than it would before. Dialing, making, producing, distributing, and training though is a quite a can of worms to work through.
@@BrendanHasty yes, reduced and especially on solving on site issues or miscommunication errors. If someone developed an AI tool to be used on site, they'd make a lot of money
It would sure be nice if us contractors could do our own engineering. Should be able to figure structural calculations on my phone. Sick of paying to wait on those pencil pushers. So I’m stoked about AI replacing engineers.
Sure would be nice if we could just get robots to build our structures. Sick of spending tons of my time to create "constructable" details only for the contractor to screw it up and blame me for not dumbing things down enough for them. I'm stoked about robots replacing construction workers.
I can trust an AI Hypothetical Doctor, but never An AI engineer at till next 50 years. After that we need to go interplanetary, so AI is good in every way esp for Engineering folks. The stress of calculations, diagrams, softwares, math, physics and what not. Overall, it’s a good news!!
Your analysis is too simple. AI is not some homogenous substance that gets "better", like you can't say an AI of the past is the same as a current one because different techniques have been invented. The hybrid team did long term strategy because AI didn't. Now AI can do long term strategy, so the comparison fails.
Hi Andrew, AI is more advanced, but it is also limited by the data it is trained on. And who will validate the solution it has produced? It was more an analogy and how engineers will great a team to produce better solutions.
@@BrendanHasty Yes it is true it is limited on the data it is trained on, but that data is cumulative as time goes on. For AI you have this concept called a fitness function. For humans this fitness function is just how well we survive and reproduce. For AI you are at liberty to set the function in any way you like. Say finding the shape which gives you the greatest volume of containing liquid for a certain amount of material. As for who will validate the design. Well that's surely ultimately the customer. If we wanted to design the best shape of bottle regarding aesthetics we would maybe randomly try designs and see how well they sold. This then becomes the training data. There is a new Chinese retailer which has started up which applies this principle to their range of products, and the growth in sales is phenomenal. Incidently for long-term AI strategy, that changed with the introduction of something called AlphaGo.
If this helps civil engineers sign off on residential designs up to a certain size then this ai approach would be good because it's already super hard finding structural engineers to look anywhere near helping anyone outside of the government or millionaire projects
Will finite element analysis software and drafting software in the future incorporate AI? Sure, but how is it any different than what it was like 20 years ago when we used excel files to calculate designs and actually printed plans and stored them in filing closets. As other commenters mentioned, we will always need engineers to understand the basics so that they can answer the "why". It could be dangerous if we only learn how to use the improved software instead.
I've been a structural engineer for 35 years, specializing in forensic work for the last 15. Every day is different, exciting, and sometimes a bit terrifying... I've spent the last week 35 feet in the air inspecting and sketching a timber supporting dome for a 20-foot diameter stained glass dome. AI is going to be amazing ... but I'm really looking forward to it being applied to create new materials.
Materials science is such an interesting topic area and I've bordered on it a bit with Quantum ML, but knew I needed something more hands-on in the long term. It's so cool!
Dear Brendan, Thank you for this video, recently I have been introduced to the field of Fire Engineers, What is your view regarding Fire Engineering and will AI play any role in its development?
Hi Sameed, fire engineering is more complex and combines many more fields. AI will affect all fields, it will be is it trained to effect your job, in the end likely it will impact fire engineering but hopefully for the better.
I happened to just watch a video on how chat gpt works. I found this direct quote from open AI: "the AI does not know anything, it shouldn't be relied on for anything important". The AI provides results based on word association and statistics, and not based on databases of facts. Apparently that is why it is sometimes wrong with the details. I wondered if it contained all of this knowledge and had ways to look up facts, but it just contains words that people have written down previously and it outputs them based on connections and frequency of use of those words, not whether it knows if they are correct or not. People have ingenuity and creativity, we can put things together in new ways and we can find and solve problems. But people have limited time and patience and people forget things. The computer doesn't forget things and can quickly try many solutions, but it only knows what it knows. That's where the "hybrid groups" come in. I'm aware of "AI" programs in engineering developed almost 50 years ago to optimize building designs. A company that designed a lot of warehouses had a program that could determine the most optimum layout for a warehouse to reduce structural cost, knowing the cost of joists, steel beams, foundations, etc. In effect it used its knowledge to try many different layouts and check the cost of each one. Engineers do this now to some extent with buildings, trying different steel and concrete schemes to determine the cheapest. AI could do this quickly and could test more systems. But what about the dark side of AI? I've seen youtubers investigate AI by doing things like writing reports, or generating computer code, and AI seemed to work well. I've also seen youtubers ask AI for solutions on technical subjects and found that the soluntions were technically inaccurate. I think if AI is gathering information from every source it has available, it will gather some inaccurate information and it will gather incorrect coupling of information. For AI to be useful in structural engineering it will have to gather information only from certain sources. It may be that the AI needs additional sources to properly couple the information. When AI provides structural solutions will it need to integrate complex structural analysis with the information it is pulling from texts? Will this require more significant coding efforts than just applying the AI engine to structural tasks? I know that 40 or more years ago there were computer programs for analysis but programs did not do very much design. Then 20 years ago we got programs that could take a model of a floor system then analyze and design the individual elements, at least for steel. Even today those same programs are not good at designing a similar system in concrete elements. If we allow that system to evolve and integrate more AI will the resulting system become too complex to trust for protecting the safety of the public? I remember the story of a submersible offshore drilling platform that was designed by elaborate computer system but when it was being put in place it failed and sunk. The failure was tracked to a software error in calculating concrete shear capacity. Human brains have the ability to look for problem areas, if we have the time and energy to do so. But like I said people forget things. We have plenty of information and intelligence in the history of structural engineering, what AI needs to do is find a way to force us and help us to apply that intelligence without attempting to replace engineering.
Hi Bill, agree that data set, where it gathers data, is really important. Will AI be safer in the longer term yes, will the public trust it? And how will we know if it had been train in the best data. Agree structural system are complex but in the end it breaks down to some simple rules repeated over and over. It is something we need to embrace and think as you state we should develop AI to Assist engineers and will hopefully be an amazing tool.
This is huge. As an AI practitioner and researcher, the quality of your data is so huge that it cannot be understated. Internet data is plentiful which is nice and spurs a lot of technical innovation just because it is readily available. It doesn't mean all data is good though. I've found chatgpt works for ideation, but there's still so many major changes to make as even a super detailed prompt made with prompt engineering in mind will only get you so far so that the code is ready for production.
