MathWorks Exec Expects AI to Mature in 2024

AI took the industry on a roller coaster in 2023 as the hype over generative AI mixed with concerns over issues like ethics and responsibility. For MathWorks, 2024 shapes up to be a year where the industry puts their heads down to figure out the most productive ways to incorporate AI and simulation into engineering, education, and other business tasks.

Like many companies, MathWorks, a developer of engineering design tools, has been looking for the best ways to incorporate artificial intelligence (AI) into their design tools to make engineering tasks more efficient. Johanna Pingel, product marketing manager for AI at MathWorks, sees several major trends driving AI’s momentum this year.

In a recent interview with Design News, Pingel said the continued pressure of compressed time-to-market cycles and product complexity will render AI and simulation to be increasingly essential in the design and development of engineered systems. “You can speed simulation by trading off a little accuracy for less computational complexity,” she noted. “But you can run many more simulations to achieve the desired accuracy.”

Smaller AI Models

Another AI trend, according to Pingel, is the use of smaller AI models for embedded AI, with larger models will persist for computer vision and language models. The smaller models make more sense as they enable a more focused set of solutions for embedded systems designs.

One of the current challenges is the speed and processing limitations of current hardware. But Pingel said that faster and more powerful processors are not necessarily the solution for many AI applications. Adapting AI tools to run on different types of hardware is preferable.

“Engineers now trained on embedded devices are coming to grips with the limits of their hardware,” Pingel said. She added that while there’s been a lot of talk about using high-end computers for crunching massive amounts of data, many applications either do not require that level of hardware nor can justify the costs of that hardware. For those applications, the AI design tools available must be able to run on other platforms.

“The reality of the situation is that some customers use Raspberry Pi. They must reduce their model size, so we have to find ways to help them adjust.”

Pingel also noted that while many AI models are based on deep learning, Pingel believes machine learning models may be more efficient for many applications. “Deep learning is the first thinking engineers want to learn about and try- then they fall back on a machine learning model.”

Education and Generative AI

On the educational level, Pingel said students will become accustomed to using generative AI and eventually never know a time when the tool did not exist. Instead of just being against AI, she believes forward-looking professors will use generative AI as a teaching aid that will free them to teach more advanced topics. But even with generative AI, Pingel said students will need to put in the work to learn subjects.

“In a university setting, generative AI is helping students do work, but students still have to do due diligence.” She gave an example of computer programming. “We still think there is value in doing basics of coding, but generative AI will enable professors to spend more time teaching advanced engineering concepts. It will free up students and professors to do their best work.”