Applications like ChatGPT can help make the interaction between driver and vehicle even more natural.Mercedes-Benz, Adobe Stock / Design Science Tech
Generative AI has long reached the peak of its own hype. There are many promising applications in the automotive industry, but the reality check is yet to come. What can be expected from ChatGPT and others in development, production, or sales?
Anyone currently stumbling upon generative AI inevitably ends up with ChatGPT, the large language model from the US company OpenAI. Since its launch in November 2022, the chatbot has fascinated not only AI specialists but also captivated the general public thanks to its ease of use. Many are already talking about a democratisation of artificial intelligence. Numerous companies from the automotive industry are now riding the wave of hype around GenAI. This is hardly surprising given the disruptive promises in terms of process efficiency or product safety - as shown by applications in vehicle development, production, or for the customer experience.
"GenAI can be deeply integrated into every area of automotive operations, whether upstream or downstream," explains Shagun Sachdeva from the consultancy GlobalData. "It can help predict demand, manage inventories, improve purchasing strategies, and revolutionise customer engagement in aftersales." Bosch has been particularly vocal and announced this summer: By the end of the year at the latest, the world's largest automotive supplier wants to have set up its own language model called BoschGPT and make it available to employees, as CEO and CDO Tanja Rückert announced in Handelsblatt. The chatbot, based on the technology of the Heidelberg startup Aleph Alpha, is intended to draw information from the company's own databases and make it more accessible to Bosch employees more quickly. Bosch is part of an industrial consortium that has invested around half a billion US dollars in Aleph Alpha. "Generative AI is an innovation booster and can change the industry, similar to the invention of the computer," emphasises Rückert.
Majority of the automotive industry wants to harness GenAI potential
In a study by the Capgemini Research Institute, 73 percent of the surveyed companies in the automotive sector stated that they plan to use GenAI or are already using it. Almost a third have already allocated budgets or even entire teams for the integration of such technologies into future development processes. There is also potential for corporate IT itself: Nearly 70 percent of all executives see the greatest potential benefit of generative AI for IT in its role as an innovation enabler across all functions, led by sales, marketing, and communication. Virtual assistants in sales and customer service are rated as the most obvious future scenario and booster for customer satisfaction and loyalty, according to the Capgemini survey.
However, while classic AI applications are already contributing to value creation in many areas, such as automotive manufacturing, some experts believe it will still take time before generative AI can be widely used in the industry. "Generative AI can indeed increase efficiency and effectiveness in various areas of the automotive value chain, but it is not a miracle cure that can solve all the problems of the automotive industry at once," also cautions GlobalData expert Shagun Sachdeva. Automotive companies should bear in mind that many questions of security, data protection, and ethics are far from being adequately clarified. And yet: "It is hardly possible to fully quantify the impact of this change, but GenAI is a powerful tool that can augment human creativity and intelligence," emphasizes Sachdeva. This is demonstrated by the various application areas in the vehicle, in development, production, as well as in sales and trade.
This is how GenAI optimizes voice assistance in the vehicle
Advertisement
The potent tool that comes with large language models is also something car manufacturers want to use - primarily in the vehicle itself. General Motors has teamed up with Google's cloud division and now wants to further roll out generative AI applications in the vehicle. The OnStar Interactive Virtual Assistant, which has been in use since 2022, uses Google Cloud's algorithm sets for intent recognition, which is intended to assist the driver with route planning or navigation, among other things. Continental has also recently started working with Google: The supplier wants to make the internet giant's applications more accessible in the vehicle cockpit through generative AI.
Mercedes-Benz has similar plans: the Swabians want to give the voice control in their in-house MBUX infotainment system a ChatGPT upgrade. With the help of generative AI, communication with the vehicle is to become "even more intuitive" - even longer dialogues should become possible. As with the GM application case, the functionality is intended to optimize navigation or simple inquiries about weather or sporting events. Mercedes initially sent the chatbot into a beta phase with US customers - a rollout in Europe could probably take significantly longer due to the high data protection requirements. Stellantis has also started a comparable field trial with the premium brand DS with 20,000 customers in Europe. The examples suggest how great the potential of large language models for direct interaction with the customer can be - and what new business models could emerge along the customer journey in the future.
The Chinese car manufacturer Nio also recently announced the integration of Microsoft's Azure OpenAI service for an optimized user experience in its electric vehicles. Nio's voice assistant Nomi is to be further developed into Nomi GPT (Generative Pre-Trained Transformer) through this step and better master complex inquiries as well as the use of natural language. This makes Nomi one of the world's first mass-produced systems for artificial intelligence in vehicles that actively uses GPT technology as a standard function across its entire product range. The rollout of the GPT function for all vehicle models available in Europe from Nio on the NT2.0 platform will start on 05.04.2024 and will initially be provided in English, German, and Norwegian via firmware over-the-air updates (FOTA).
Along the Customer Journey, the GenAI Hype is Greatest
Customer service is already a popular application scenario for generative AI today, explains Jan Pilhar, Executive Director & GenAI Lead at IBM iX DACH. Since the launch of ChatGPT, GenAI has been fueling the imagination of many marketing and sales executives, with a pronounced "fear of missing out" in the market. Accordingly, many have somewhat "naively" embarked on initial projects. "The disruptive developments of generative AI have widened the gap between existing IT infrastructure and the availability of market-ready technologies," also notes Daniela Rittmeier, Head of the Data & AI Center of Excellence at Capgemini. The pressure for technology integration along the value chain no longer comes from internal tech experts, but from customer expectations, says Rittmeier.
Typical mistakes: Generative models are used even though existing machine learning approaches are sufficient - or incorrectly sized models are chosen. "There is not only OpenAI's GPT, but currently also around 120,000 open-source models like Llama 2 from Meta, which can perform many tasks just as well and more cost-effectively. It's always about operating the right model in the right size in the right infrastructure," says Pilhar. The type and size of the model directly affect the cost per request (inference). With millions of customers, requests can quickly lead to high costs. Pilhar: "We are experiencing numerous OEMs who are only now realizing how expensive the quickly built feature on the website or in the car configurator really is. Here, the right setup should be chosen for maximum efficiency, which sometimes also involves complex technology decisions."
A real obstacle for the integration and scaling of GenAI, from Daniela Rittmeier's perspective, is the existing infrastructure and current data quality. "The IT infrastructure has grown heterogeneously, largely with the aim of optimized hardware production. It is, like the vehicle architecture, limited 'AI ready' for cross-departmental processing of machine-readable data," says the expert. The process, vehicle, and especially customer data important for AI are still only limitedly available.
Another current practical example of the use of generative AI along the customer journey is presented by Toyota with the new ToyoGPT platform. The application has been tested since February 2023, and the corresponding mobile app is available to customers as a voice assistant for all questions regarding their vehicle's manual. Furthermore, GPT agents support employees in the sales area based on internal training materials to improve their sales conversations.
Since the use of the freely accessible ChatGPT application can pose enormous risks to sensitive company data if it flows to the servers of the provider OpenAI, the car manufacturer instead uses the UGPT technology from Objective Partner. With this, employees can use familiar ChatGPT functions without disclosing sensitive or confidential company data, as the data remains within the company at all times.
Significantly faster vehicle development thanks to GenAI
In vehicle development as well as in the elaboration of vehicle design, generative AI ensures significantly more speed. “Ideally, development time could be halved in the future. In addition, with generative AI, the range of functions can be efficiently expanded using software throughout the entire lifecycle of the vehicle,” explains Elmar Pritsch, Partner and Global Lead Software-Defined Vehicles at Deloitte. In addition to speed, the fuel efficiency or lightweight construction of a design can also be optimized. Furthermore, artificial intelligence enables a process known as inverse design within the framework of materials science. In this process, the required properties such as conductivity or corrosion resistance are first defined before a suitable material is sought using AI.
Particularly the collaboration between AI and developers will form an interesting focus in the near future. According to the market research company Gartner, by 2025 more than half of all role descriptions for software engineering leaders will explicitly include the management of generative artificial intelligence. The task of the engineering leaders will then primarily be to clarify that AI technology enhances developers' productivity and does not replace them. "Generative AI will not replace developers in the near future," says Haritha Khandabattu, Senior Director Analyst at Gartner. "While it is capable of automating certain aspects of software development, it cannot replace human creativity, critical thinking, and problem-solving skills."
GenAI serves as a driver of innovation in production
In the production context, generative AI enables an improvement of processes by providing process details or operating instructions. Additionally, it helps in visualizing ideas, for example, by assisting in programming. "The fields of work of specialist departments are thus somewhat democratized," explains Jens Neuhüttler, Team Leader for Digital Service Transformation at Fraunhofer IAO. Thus, GenAI could especially ensure more innovation performance in areas such as prototyping, design, and production process planning.
Another use case is the development of AI-based services or the training of production robots. These tasks require a large amount of high-quality training data, which is often not available. This is where generative AI models come into play: By generating synthetic training data, companies can reliably simulate a variety of situations and comprehensively train their AI models. Schaeffler and Siemens are making an initial attempt at AI-supported programming of machines. However, Fraunhofer expert Neuhüttler points out that the use of such applications in the production environment often fails due to the lack of data infrastructure and continuity.
IAA MOBILITY 2023: GenAI in the Automotive Industry
As one of the first car manufacturers, Mercedes-Benz is using ChatGPT in production: The program is intended to serve as a voice-based interface to support production staff in improving data analysis and accelerating the identification and analysis of errors. Specifically, OpenAI's model is to help check the current day's production planning in real-time and adjust it flexibly if necessary.
Generative AI is also making its way into the manufacturing landscapes of the supplier industry. Among other things, Bosch recently launched a pilot project at two of its German locations, where GenAI generates synthetic images to develop AI solutions for optical inspection or to optimize existing models. The supplier assumes that this will reduce the time from project planning to commissioning and ramp-up of AI applications from the current six to twelve months to just a few weeks. At the Hildesheim plant, synthetically generated images have already been successfully used for training in the first series systems in electric motor production. Specifically, Bosch uses the AI images to reliably check the welding of copper wires in electric motor production.
