Artificial Intelligence AI Trends for Engineers: Where is the Journey Going?

Related Vendors

FAULHABER Drive Systems

Wuxi InfiMotion Technology Co., Ltd.

Companies must continue to advance the introduction of Artificial Intelligence and develop new applications for it, so as not to lose traction. What AI trends should engineers keep an eye on and what challenges need to be overcome?

Johanna Pingel, Product Marketing Manager for AI at Mathworks.

According to Gartner, companies that have adopted AI engineering practices for the development and management of adaptive AI systems have a clear competitive edge: by 2026, such pioneers will surpass their competitors in terms of the number and time required to operationalize AI models by at least 25 percent. As a result, companies need to develop a clear strategy for themselves. What trends are most important?

Physics-based AI - Models take into account rules and principles of the real world

As AI is making its way into more and more research fields, such as complex technical systems, AI models need to take into account physical boundary conditions to be meaningful overall. The combination of data and physics, such as through neural ODEs (ordinary differential equations) or PINNS (physics-informed neural networks), has great potential. The focus of physics-supported AI is on simulations: Complex models can be configured as variants within a simulation, allowing developers to switch quickly between models to get the best and most accurate solutions.

Reduced Order Modeling (ROM) with physics-based reduction models is also an important new trend. Through the use of AI, simulations can be accelerated by replacing an extremely compute-intensive first-principles model of a system, without compromising accuracy.

Collaboration on AI - Free access to AI is becoming established

Researchers, engineers, and data scientists should continue to expand their cross-functional and cross-industry collaboration so that they can think of innovative solutions from different perspectives. To make the latest models available on demand and enable users to build on the latest research results in the shortest possible time, network-based versioning services for software development projects like Github are recommended. Open-source solutions are also becoming increasingly popular as engineering teams often work with models from different frameworks. A stronger networking of science, academic research institutions, and companies further promotes AI research, from which researchers and users benefit. This applies, for example, to topics such as physics-supported machine learning and biomedical image processing.

Companies are focusing on smaller, more interpretable AI models

Users of artificial intelligence are increasingly finding that they must be able to provide models, adapt them to the hardware, and explain the decisions of the models in order for them to be relevant. Therefore, the explainability of models and related applications are increasingly coming into focus for engineers.

To meet the requirements of cost-effective devices with low power consumption and explainable outputs, engineers are increasingly turning to traditional machine learning models and parametric models. These are compact, have a low memory footprint, and meet the requirements of the application through easy interpretability of the output. If newer, memory-intensive models are needed, quantization and pruning techniques offer ways to compress the models, thereby reducing the model size with minimal impact on accuracy. If necessary, engineering teams can thus use interpretability, quantization, and pruning to extend the use of AI, including deep learning and traditional machine learning models, to conventional model development.

AI will be crucial for the design, development, and operation of modern technical systems

AI is increasingly making its way into all industries and applications and will be crucial for technological advancement as well as the development and operation of modern technical systems in the future. In more established fields of activity where AI has only recently been introduced, engineers often require additional background information on this technology as well as specific reference examples to integrate AI into their work. Based on tested examples, engineering teams can bring data and their knowledge into such examples, expand them, and thus integrate AI specifically tailored to their task.

Which challenges AI engineers can expect

Since different teams are often responsible for creating and implementing AI models, complex challenges arise in the AI environment that engineers continue to have to manage. The selection of pre-processing algorithms and model training, for example, usually fall within the remit of data scientists, who focus on accuracy and robustness. However, for successful porting to the target platform, engineers still need to consider many other criteria. Early testing of algorithms for feasibility assessment, for example using Processor-in-the-Loop (short: PIL), can prevent models that are already trained and, in some cases, very powerful from having to be discarded in the end.

In most cases, the training of the AI is also implemented in a different programming language than the implementation in the hardware. However, models from the training environment cannot simply be run on the target hardware without further ado. To overcome barriers between scripting languages, there are runtime interpreters like Tensor Flow Lite, Machine Learning Compiler Frameworks like Apache TVM, or automatic code generation in Matlab/Simulink.

Finally, the safeguarding of AI models remains an important issue: While AI models are allowed to make mistakes in the training environment to learn and improve, these can lead to major damage in real existing systems after implementation on the hardware. The question of reliable, objectively verifiable criteria for a model considered safe will continue to be an important area of research in the future.

Outlook

The introduction of AI has implications for the entire company, from interdisciplinary collaboration to the design of specific components. Therefore, it is crucial for engineers to identify use cases that align with their short and long term goals and to implement them accordingly. As AI pervades all areas of work and also safety-relevant areas, questions regarding model quality, language compatibility, and safeguarding will become a focus.

This contribution originally appeared on our partner portal Industry of Things.