HOUSTON — At 3DEXPERIENCE World, Dassault Systèmes unveiled its 3D Univ+rses AI-centric core. In addition to the Gen 7 release plan, Dassault announced the exploration of an “AI-powered Generative Economy.” By combining virtual twins, generative experience, virtual companions and sense computing, Dassault is shaking up the world of IP lifecycle management.
The AI era has quickly taken over the industrial sector as Dassault Systèmes progresses into a new generation. But it’s equally important to understand the development and potential reservations that go hand-in-hand. To better evaluate and understand the progression, SolidWorks Data Analysis & Science Director Shrikant Savant gave his takes on AI evolution, reservations and the technology’s hype cycle.
Devon Verbsky (DV): You’re an expert in AI and in tune with product development. How has that understanding allowed you to excel in your position over 18 years at Dassault Systèmes?
Shrikant Savant (SS): You take any technology and AI, to me, I take it more as just another tool.
It’s another tool that can solve problems that couldn’t be solved by using previous tools. So it’s a new tool, but a very technical tool in the sense that there is a lot of math and statistics and understanding of data that goes on behind the scenes before you can actually get a meaningful result from a model.
Any AI problem, if you look at it, at the core of it, is optimization. You define your design variables, you define your objective functions, you define your constraints and you iterate and create certain parameters to get the results and see if these results make sense. If they don’t, then you go back and fine tune certain parameters and you keep doing that. All of those directly apply in AI as well because an AI model has many parameters, and it requires fine tuning of these parameters before you get results that make sense.
DV: What AI evolutions have you experienced?
SS: My first neural network program I did in the early 90s, and there I was writing everything from scratch myself.
But now there are these pre-made packages that give you these ready-made toolboxes with tools that you can use. That solves a lot of these nitty-gritty problems that you would otherwise deal with. These are all within a few lines of Python code. You can have a full fledged conversation, neural network ready that can identify differences. That collective intelligence helps you focus on your particular problem.
DV: Do you think AI is avoidable?
SS: I think the problem with AI — and the hype that you see — is probably more because of this thing called artificial intelligence.
I think that’s where that stigma comes from. People are afraid, thinking, what’s it going to do? Am I going to be replaced? That’s the question. I look at it more as a new tool for solving problems that could not have been solved with conventional tools. If you have to write a program to do, for example, autonomous driving of cars, there is no way you can write that program because the input is going to be so complicated. There’s no way you can deterministically code that in your program. And so this new tool allows you to actually solve that problem. It may not be perfect, but it does a pretty good job. If you look at it as a new tool or a new way of solving certain problems that couldn’t be solved before, then I think that stigma goes away.
DV: Do you think there’s a way that everyone can become comfortable with AI?
SS: AI is not magic.
It does not take magic to make it work. You can’t say, ‘Here’s my data, take it’ and AI will do everything for you. There’s a lot of hard work to take the data that you have and to convert it into something that the AI engine can understand. There’s so much that goes on. If you advertise this more as a tool for solving complex problems, then I think that stigma automatically goes away. I think people have to familiarize themselves. Knowledge is power.
DV: Another focus here is augmenting human action. What does that mean to you and how far can we get with it?
SS: I think it is tremendous, and I’ll give you an example.
For a blind person, this is a tremendous tool because now they can actually, from their phone, point the camera outwards, and in real time, it’ll dictate, describe to you, there is a table here, there’s a blue color of table cloth here and another person sitting here. It can tell you all that in real time using human voice. I think that kind of feedback, that kind of assist, that they’re getting from technology is tremendous.
DV: Speaking on augmentation, what are your thoughts regarding its current state?
SS: I think all these categories work in a way of building blocks.
Processing texts, processing images, processing videos and you can then mix and match these building blocks in different permutations and combinations. You can create these unique solutions for solving specific problems. Not to say that there aren’t drawbacks of the current progress, but there are definitely good-use cases. This is a new way of technology, a new way of life and we have to go forward with it. I would say that the earlier people understand that AI is not something that’s going to die down tomorrow, the better. It’s here to stay and make changes in their knowledge, in their skillset.
DV: Do you have any reservations about AI?
SS: Yes, I do have some reservations.
I worry about the sustainability of the planet. And I feel some of these competitions that we do, some of these data centers that we have, they require so much compute power. It’s equivalent to thousands of people using power for many months. And then the question to ask is, ‘Is it beneficial to me or you to type a question and get an answer in one word at the cost of impact on the environment?’ I have reservations about that. That’s probably my biggest concern.
DV: Does AI have a ceiling or an endpoint?
SS: I think it’s a hype cycle.
You will see that large language models came out and were said to have a few million parameters. Then the later version had more parameters, and every update includes even more parameters. You would think that increasing the number of parameters would improve the power, but beyond a certain stage it doesn’t happen. It does reach a plateau. Then the question is, ‘What do we do beyond that?’ We can’t keep on making this bigger and bigger and expect it to get smarter and smarter. That doesn’t happen. I feel that AI has that limit. There is a hype cycle and we might be at the peak of that. People will grow to understand the implications of that, but it’ll settle down into somewhat useful things.
DV: As a data analysis expert, what is your take on SolidWorks and Gen 7?
SS: My job is to develop the technology that aligns with the vision.
And of course, the vision is something that is more of a direction, it’s a goal. I work with engineers and we develop the technology for it. I have to deal with all the nitty-gritty aspects of it. How do you get data? How do you translate that data? How do you level that data? How do you build models? How do you test your technology? How do you deploy it to customers? How do you make sure that it gives you the right results?
That has to align with the vision of the entire company, the entire organization. I would say I have to be aware and cognizant of that. But ultimately, we are still solving the core technology problems the way we have, but need to make sure that we are moving the applications in the right direction, that is consistent with Gen 7.
DV: What about the generative economy concept?
SS: I think in general, I feel the whole field of generative AI is somewhat still in its infancy, especially as it is applied to 3D design.
I think generative AI has many applications. In some applications it is very advanced, but when it comes to 3D geometry, 3D modeling simulation and looking at real world problems that we as a company solve, I think that field is still in its infancy. It has a long way to go before we can be at the point where we say, go ahead and build me a jet engine, and it goes ahead and builds a jet engine. We are not there yet.