Generative AI is still popular in the automotive industry

This article is the GTC 24 exhibition notes of an Onsemi expert. We can try to see from the expert's perspective that after spending four exciting days at GTC 24, he can't wait to record the hot topics in the current technology field: the new era of generative artificial intelligence (Gen AI ) and how it will reshape the automotive industry.

GenAI is a branch of artificial intelligence (AI). Traditional AI models are tailored for specific tasks and rely on labeled data from specific training datasets for training. GenAI base models are different. They learn patterns and structures from much larger datasets in an unsupervised manner. They can deeply analyze known datasets, find connections and patterns, and then "generate" creative outputs such as videos, pictures, text, software codes, etc. Therefore, when training larger GenAI models, the costly data labeling work in traditional AI model training often becomes less important.

Once fully operational, GenAI will usher in a new world where creativity and efficiency are unlimited. It is also important to emphasize that GenAI can also be used in conjunction with existing technologies such as predictive analytics. GenAI "generates" unique outputs by identifying patterns and automatically predicting them over and over again, while predictive analytics uses historical data and statistical modeling to predict future outcomes and behaviors in specific scenarios. The combination of the two will enable businesses and organizations to make more informed decisions. The details of the fusion of these two technologies will be discussed in depth later.

Goldman Sachs estimates that GenAI could boost global GDP by 7%, or nearly $7 trillion, over the next 10 years. GenAI can generate new content with a style and depth similar to human creations, opening up a whole new realm of possibilities for us and improving every aspect of our lives.

Let’s talk about the factory floor first. Currently, the manufacturing industry is using a wide variety of sensors . By analyzing the massive amounts of data collected by sensors, we can better understand, comprehend, and predict machine performance and product quality. However, many of these sensors work independently. Although they can provide outputs individually, they lack overall consistency, making it difficult to maximize the ability to predict the performance of manufacturing equipment. GenAI can process comprehensive data sets from all aspects of the manufacturing floor, resulting in new output scenarios beyond our imagination.

1. Generative AI drives innovation in automotive design

Next, let's look at the profound impact of GenAI on the automotive industry and explore how it will drive innovation in automotive design. Automotive design involves many complex components such as batteries , transmissions, engines, infotainment systems, etc. It also involves a large number of design layout choices and technical safety restrictions to ensure that all aspects of the vehicle's road standards are met. In addition, the requirements for automobiles vary from region to region.

With the GenAI software tool, automakers can create realistic 3D car models by simply adjusting a few input parameters. Designers can also ask GenAI questions to explore solutions to questions such as "Can other materials be used for the interior?" or "How to reduce the cost of the design solution?" By combining past design experience and customer feedback, new designs can further optimize costs and efficiency and accelerate the process of bringing products to market. In this way, designers can speed up the prototyping and testing process to discover problems as early as possible and quickly summarize failure experiences. New technical insights are gained from each round of iteration, thereby enhancing vehicle performance. This speed of innovation is expected to revolutionize the automotive industry.

At the same time, customer service and user experience will also see huge improvements. How was your last car-buying experience? Most often, you made an appointment with a dealer, waited a long time, and had to deal with a salesperson who was desperate to meet performance quotas. What you were looking forward to was just the joy of owning a new car and the new journey that comes with it. The GenAI chatbot can change this situation. It can understand your needs and provide a completely personalized experience, tailoring recommendations for you, including car options, optional features/packages, and even car financing options. You can ask it directly to get the specific information you need to make a decision. This is undoubtedly an ideal result for both you as a buyer and the car dealer as a seller.

Now let’s look at the supply chain. If you’re in the semiconductor industry, which has recently experienced supply shortages , this is especially worrying. Traditional AI certainly has certain advantages, such as the ability to closely monitor inventory levels, monitor the supply chain in real time, track the status of goods transportation, and detect supply disruptions in a timely manner. On this basis, GenAI adds two key features:

First, if a supply disruption occurs, GenAI will be able to provide alternative paths, thereby minimizing the impact of the disruption and reducing the additional costs caused by it.

Second, GenAI performs “what if” scenario analysis during planning. It considers different scenarios, such as what would happen if raw materials from supplier A were not available on time, or what would happen if the price from supplier C increased. It then translates this information into solutions. In this way, we can optimize the supply chain layout, increase its reliability, reduce unexpected situations, and avoid putting global markets in trouble.

Finally, as global fleets grow larger and operate more extensively, the amount of data that needs to be processed is also increasing rapidly. Fleet managers have long complained that the data generated is too complex and it takes a lot of time to extract useful information from these obscure data sets. In this regard, GenAI can help. It can simplify the process of fleet data analysis and make the whole process intuitive and clear. Fleet managers can ask GenAI chatbots questions and get answers, such as: "How many electric vehicles are in my fleet ?" or "What impact do electric vehicles have on my total revenue?".

In addition, GenAI can help you prevent potential technical problems that may lead to vehicle failure and downtime. It can combine real-time input data from sensors with historical data trends to achieve predictive maintenance. Thanks to GenAI tools, companies can manage fleets more efficiently, create a safer and more reliable experience for customers, improve operational efficiency, and ultimately drive profit growth.

2. The role of Anson

In this era of technological change, ON Semiconductor also plays an important role. Image sensors can capture a large amount of image and video data to help drivers and/or vehicle systems make timely decisions while driving. By working closely with ecosystem partners, ON Semiconductor can help vehicle manufacturers shorten their design cycles.

NVIDIA Omniverse’s new Cloud API empowers developers of autonomous driving systems to accelerate their design process with the NVIDIA Omniverse ecosystem of simulators , validation and testing tools, including models of our Hyperlux™ family of image sensors.

Using Hyperlux synthetic sensor data, it is possible to create simulation environments that simulate real-world sensor inputs to build digital replicas of real-world scenarios, simulating road layouts, traffic conditions, and weather changes. These virtual environments allow developers to test algorithms, sensors, and control systems without actually driving on the road , and can deliberately create boundary conditions (such as sudden tire blowouts, sensor failures, extreme climates, etc.) to verify the behavior of the system. Using image sensors to generate synthetic scenes can help improve the generalization ability of the model and reduce the need to collect data frequently on the road. In this way, relevant personnel can test and verify autonomous driving systems and vehicle performance in a risk-free, low-cost, and efficient virtual environment.