How AI is paving a new path for design and manufacturing podcast – Transcript
Artificial intelligence is already finding a place across many different areas of the design and manufacturing process but its benefits aren’t limited to supporting individual, siloed, applications. As both AI technology and the digitalization of the complete product design process continue to develop, AI will become a key tool that spans across the entire breadth of industry, reshaping the entire process from end to end.
Find out more in the latest episode, available here.
Spencer Acain:
Hello and welcome to the AI Spectrum Podcast. I’m your host, Spencer Acain. In this series, we talk to experts from all across Siemens about a wide range of topics of technologies and how they’re related to AI. Today I’m joined once again by Dr. Justin Hodges and Boris Scharinger as they continue their discussion on the applications of artificial intelligence across the full breadth of the product design and manufacturing process.
So, to kind of round things out from our previous discussions, you’ve both talked quite a bit about applications of artificial intelligence within specific areas of the product design and manufacturing process. And I think we’ve also kind of touched on in that time the kind of combined approach to AI where you have AI spanning the breadth of design, and manufacturing and simulation and all of those different subtopics and how it’s kind of going across all of them. But I’d really like to kind of focus on that a little more explicitly now. So how is AI helping to combine these different fields and disciplines, or what are some technologies or topics that are spanning across all of them with relation to artificial intelligence?
Dr. Justin Hodges:
Yeah, I can offer a few. So basically, use cases that are not niche or specific to just one application. So I break it into three categories. The first is faster decision making, whether it’s a real-time model or some analytics feedback. You’re getting the opportunity to make a design decision much faster than you previously would. And the flip side is the other use cases, innovation, not just getting my designs faster, but making better designs, an innovation and improvements to how I do my design. And it could be generative AI, it could be helping you sample a space better and do design exploration better.
And then the third, of course, is the one that’s always trending on a news source, social media every day, which is just those related to large language models or chatbots and Copilots, but basically improved user interaction. Just help me use this software or do this process in an improved way, whether I can do the process quicker or it’s less painful, or I’m more self-guided so I can be more efficient. Whether I have on-demand knowledge from some language model that is exposed and trained with expert advanced knowledge so I can learn from it. I can rapidly query documentation information or those sorts of things. So I’m sure there’s more, but when I get into the generalizing game, think about unified use cases. I think about faster decision making, innovative design support, and improved user interaction.
Boris Scharinger:
I agree. And I’d like to add one scenario that I find most interesting, also a bit on the visionary side, I must admit. But I’m very impressed by large language model capabilities to dynamically interface applications via APIs. So interfaces don’t need to be developed in a hard-coded way. They can almost be created on the fly by a powerful LLM architecture maybe supported by an agent-based approach.
And this means that customer specific workflows across applications where each application is their own silo and each application has their own specialty and has their role in a product development and production process. But how do you chain them up efficiently? And today that is done with all these hard-wired interfaces, which are then extremely static. And when you need to change it, the cost of change are extremely high. Whether you need to change it because you want to replace one of your tools or whether you want to replace it because a new product requires a change in your pipeline, in your development pipeline. And I foresee a future where these types of interfaces between the many different tools in that pipeline where these interfaces become much more dynamic. Quality control will be a challenge in that process of having these dynamic interfaces and quality control then needs to span different applications being chained together in a dynamic way. But this is a very interesting value proposition of generative AI in the space of software development and dynamic API interfacing, if that makes sense to you?
Dr. Justin Hodges:
Yeah, I agree. That is always the headache, right? When you have such a passion for whatever it is that you’re doing with your domain knowledge of electromagnetics or turbulence modeling and you think, gosh, I’ve spent so much time just setting up all of my software and coupling it together and setting up these interfaces, that’s the pain part, that really takes a lot of time. It’s not always a demand explicitly on a more accurate answer or the solution solving quicker. But a lot of times the setup and preprocessing and connecting of tools and whatnot is a huge amount of time and consumption. So it will greatly help people there. I agree. That’s a nice way to put it too. I appreciate the paradigm change to software that has happened, whether incidentally or as a result of the advent of large language models and just more popularized machine learning. I really appreciate the paradigm changes in software that have taken place, but that one makes sense to me.
Spencer Acain:
It sounds like there’s some exciting prospects on the horizon here for bringing AI across the industry and the impact that’s going to have.
Dr. Justin Hodges:
So I guess to kind of more modern software, right?
Spencer Acain:
Absolutely.
Dr. Justin Hodges:
I think that’s part of, that’s how I would abstract it to the highest level. You know when you’re using old software. And sometimes when you use new software, it really feels so intuitive. And I think that is kind of how I abstracted to a high level from what Boris just said.
Spencer Acain:
Great. But we are almost out of time here. So to just round things out with a final question, where do both of you see the impact of AI across the entire industry going in the future, maybe the next 5 years, maybe even the next 10? How will we see AI change the landscape of design and manufacturing?
Boris Scharinger:
Justin, your call.
Dr. Justin Hodges:
My call, I mean, I can’t wait to hear what you have to say. I’d love for you to go first. I’m selfishly just wanting to listen to your perspective.
Boris Scharinger:
So then let me reiterate, right? I think that a couple of things that we already discussed are really forward-looking and this topic, design space exploration becoming more common, but also spanning into the simulation areas that touch how things are manufactured. Not just stopping looking at product features, but also like how can it be produced most efficiently, improving yield and all of that stuff. That’s going to take time. But I’m absolutely excited to see this happening and I strongly believe it will happen. And this again brings back that connection between AI and the industrial metaverse, simulation is going to be a lot more holistic than it is today. Today everyone is simulating on a component level today everywhere the simulation of a certain thing is strongly attached to the authoring tool that you designed it in, and that will be decoupled. And having this aggregation platform for digital twins and simulation, that is going to make such a huge difference for companies who need to master complexity.
If you are producing products where no complexity has to be mastered, then skip it. But anything that is complex, not talking about ships, and, yachts and airplanes, but anything that really is a complex design, PCB plus mechanical design, plus software and so on, will greatly benefit from the holistic simulation capabilities, which we call industrial metaphors. And that, from my perspective, bridges in another discipline. We all grew up with the concept of model-based system engineering. And the idea that I master complexity by defining models, by breaking things down pretty stringently into subcomponents, into subdomains, and having many, many specialists working on a complex topic simply by the divide and conquer principle of the top-down model design. And I do believe that there will be a next generation MBSE discipline on the horizon, which will be largely fed by agent-based engineering and design approaches supported by LLM technology.
And of course, LLM technology is great when it is common and broad language. It even fails when you zoom into industrial language, into domain language because LLMs have been trained using scraping the publicly available internet and not necessarily all the special press, and books, and experts expert content in the industrial domains, but hey, that’s our whole at Siemens, right? We will make this happen. Coming also back to what you mentioned Justin, earlier, what is our whole? Our whole is to take AI technology that is out there and then be catalytic so we transform it into something that works for our industrial customers. In many areas, that means we add a significant portion of determinism to it. So engineers can rely on AI, not hallucinating and whatnot, but with a certain input to produce a certain output. So this area, design space exploration, industrial metaverse, a next generation perspective on model-based system engineering, all of that to master complexity. That is really what I see on the horizon in say between 5 to 10 years out there.
Dr. Justin Hodges:
I couldn’t agree more. I hate to even rival a response and try to say something because I agree with everything you said it was so well put, but just some blips to make it an even shorter response than yours for the next 5 years or even 10 years. I mean, better generations of our design work, better usability of software and user interfaces and just our user experiences would be improved. Like Boris said, I love the term aggregation. Aggregation platform where I imagine I have all these different models they stay in think with my design objectives, and I can just get this information on demand, and as a consequence, I can spend more time doing the technical work that I want to or using my domain knowledge that I seek. I think simulation will remain as the gold standard and the most trusted way of getting a result.
But I think AI will more and more be used to take away the scenarios in our design space exploration that are not fruitful, and allow us to really dive deep on the best parts of the design space and keep improving them and applying our physical knowledge and domain knowledge. And I think that is one of the things I’m most excited about. And that goes again, in continuity with what you said, Boris, about how the design work and the simulation work is tied to the authoring tool. Like the CFD expert is producing these results so closely linked, and living, and staying on a leash towards the software that generated the simulation work. It’s not easily accessible and given to teams in a completely different discipline for the same company. So I think those barriers will keep going down.
And I think more and more we will have more utilization. Kind of like at the beginning of the podcast where we talked about being resourceful, frugal, leveraging what we to generate already, I think a lot of companies can make better use of the data that they have available, not just other similar types of simulation, not just other types of experiments ran, but even from the day that something was beginning the design work to the final day that it’s retired as a manufactured product in operation that’s retired. Really taking all of that knowledge and making it pruned for conclusions for people at various parts of that. In other words, this dream of if I make this design decision on my component level, do I have any feedback on how that may impact the manufacturing timeline or how long this part would operate before failure? So that lives on this structure of acquiring all of this data and then putting it into sort of meaningful search, and meaningful semantics and providing it back to the user in case they want to access this knowledge while doing their work.
So it’s an exciting vision because in short, this feeling I get is like, I’m going to be able to do a lot more, and if everything is done responsibly, then I don’t have to worry more than I already do when doing my design work about AI giving me bad answers that I can’t trust. So I’m looking forward to the continued vetting and validation of where AI can help, where AI cannot help, and then this integration so that people can with responsibility and trust, access all of this information and more quickly focus on their task as a domain expert. Maybe that’s a bit skewed because I am with a technical role, so I just want to do more of the same, faster, better, et cetera. But that’s some of the stuff I’m looking forward to, Spencer.
Spencer Acain:
Wow, that was a fantastic response and this has been just a really incredible time talking with both of you. But I think we’re out of time for today. So Boris, Justin, I would really like to thank you both for joining me here today.
Boris Scharinger:
Thanks for having me.
Dr. Justin Hodges:
Yeah, likewise. Such a joy to talk to you, Boris. Thanks.
Boris Scharinger:
Likewise, Justin.
Spencer Acain:
Well, I think that is all the time we have for this episode. So once again, I have been your host, Spencer Acain on the AI Spectrum Podcast. Tune in again next time as we to explore the exciting world of AI.
Siemens Digital Industries Software helps organizations of all sizes digitally transform using software, hardware and services from the Siemens Xcelerator business platform. Siemens’ software and the comprehensive digital twin enable companies to optimize their design, engineering and manufacturing processes to turn today’s ideas into the sustainable products of the future. From chips to entire systems, from product to process, across all industries. Siemens Digital Industries Software – Accelerating transformation.
