AI in Engineering, Design, and Simulation
Through the conversation between Kenneth Wong and Jonathan Girroir, we highlighted several areas where AI is having a significant impact on the engineering landscape.
Generative Design & Natural Language
Mr. Wong discussed his view that the beginning of artificial intelligence’s direct impact on the product lifecycle started in generative design a few years ago. Something of the first version of AI involvement in engineering workflows, users could simply provide their criteria to software and get some form of design concept in return. While useful, this was something of the first wave.
Now, we are starting to see the incorporation of other areas of AI technology, specifically natural language capabilities. Users used to be limited in how they could populate a 3D environment: this process typically involved pulling objects from a library and assembling them into something resembling your goal. Natural language commands allow users to give commands in a much more human manner.
Mr. Wong shared the example of a warehouse - with a capable natural language AI tool, users can simply instruct the software to fill the warehouse with objects one would typically see in a warehouse, instead of manually placing individual items.
These capabilities are likely going to appear in more and more of the software we use. Currently, we cannot simply instruct our design tools how to edit a component - tell it to add a hole, narrow some spacing, etc.
Natural language commands that could accomplish this would drastically improve the accessibility of design software to those with far less training and experience. Where users currently need experience in the right order of operations, and the correct ways to implement such changes, natural language-empowered generative design could transform user interfaces and experiences.
These features are not present in mainstream CAD software yet, and it is important to remember that this is very new technology. ChatGPT, Midjourney, and other text-based generators only exploded into public consciousness 12-18 months ago. Developing CAD and simulation tools enhanced with natural language and AI to provide these features will take time. Perhaps, Mr. Wong speculated, we will see some beta tests of such capabilities by the end of this year.
Evolving Experiences for Engineers
The integration of AI tools into CAD and simulation software will necessitate changes to the user interface and experience, some of which might not be universally welcomed. Those who struggle with the steep learning curve of these tools are more likely to welcome such changes. Experienced users of engineering tools tend to be more resistant to change.
We could see user interfaces altered with AI-empowered, natural language-based tools that could suggest a design starting point based on your needs. A typical engine block, bracket, bolt, etc. From here, you could iterate manually, or with the new AI-powered tools.
Exactly what these changes might look like remains to be seen. Mr. Wong shared examples of text-based generation for Bing and Canva as examples of these in other spaces. The immediate results are rarely perfect, and the role of the human expert is as important as ever.
While these tools offer ways to enhance or streamline the design process, Mr. Wong believes that human engineering skills are certainly not going to become obsolete. A talented engineer, for example, can estimate the results of a test before running it, based on their years of experience.
It is also important not to discount the role of less tangible elements of design - aesthetic choices, sound, and how something makes the user feel. These are elements that are not suddenly going to become unimportant. Mr. Wong expressed that there is a natural fear of humans becoming irrelevant. In his view, this is not a danger. While engineers' roles may evolve their expertise and human touch are and will remain invaluable.
Digital twins are much more than 3D models - they include huge amounts of data from real-world devices. AI can be extremely helpful in managing the vast quantities of information required to get the best simulation results out of digital twins.
In the example of a wind turbine, heaps of weather information, corrosion and pressure data, etc, are coming in by the hour, more than a human can reasonably examine. Artificial intelligence can comb through millions of sheets worth of data to make decisions and spot patterns. With these tools, we can better plan preventative maintenance and highlight areas of potential failure, perhaps even those human eyes would’ve missed.
Standard simulation methods involved running physics-based tests on every corner of a component to test how they deform. Now, people are experimenting with reduced-order models. AI tools could examine previous simulations and spot correlations between certain input parameters and results. With this, you could run simpler models that do not require a full physics run significantly faster than before.
If natural language tools and their impact on engineer experience can be integrated into simulation software, the process of simulation could be greatly streamlined. These tools would become far easier to learn, to use, and open their capabilities to wider audiences.
What Could Go Wrong: The Future Problems of AI in Engineering
One of the major opportunities when looking forward with new technology is predicting potential problems. Kenneth Wong and Jonathan Girroir to share their “asks” for the industry, as they look to the future, along with what challenges engineering, design, and simulation companies should be cautious of concerning AI.
Interoperability, an Ever-Present Problem
Mr. Wong foresees interoperability to be a major challenge. He shared the example of BMW and Ericcson, the phone company. Both are using metaverse technology, with the German car manufacturer operating digital twins in Omniverse and Ericsson using technology of their own.
At the moment, it's highly unlikely you could operate a simulated BMW in Ericsson’s metaverse designed to test 5G. If digital twins are going to be a significant part of the future, we must work out ways to allow digital twins to exist in each other's simulations. This interoperability question is one Tech Soft 3D is constantly asking about CAD and other engineering data formats, and is not something with an easy answer. The importance of interoperability with technology related to AI is going to cause headaches across industries, and it seems unlikely companies will suddenly find a fix-all solution.
Terminology: What Terms Mean What?
Another challenge with new technology comes from the nature of how we discuss them. “Artificial intelligence” is a term that can be applied to many different things. Jonathan Girroir spoke to this and the importance of working to develop a common understanding of what means what. Neural networks, machine learning, natural language, deep learning, generative design, we have to be extremely careful to be consistent and deliberate in how we use and define these words as we look to the future.
Mr. Wong added the importance of challenging these terms, cautioning against overuse. Product Lifecycle Management (PLM) is an example of this, where the term has to mean anything from an Excel spreadsheet to complex enterprise software. With its prevalence in popular conversation and linguistics, AI must refer to true AI and not just complex programming or automation.
Both Mr. Wong and Mr. Girroir remain optimistic about the power and potential of AI in digital engineering. Despite global challenges, growth and technological development in AI tools have been rapid and stand to create huge opportunities across many industries.
For more insight into digital marketing and other interesting topics, check out Kenneth Wong’s work here. His reputation and expertise are well-earned, and he was a fantastic guest on our Beyond 3D Podcast. You can find a video of our interview below.