What do Automated Vehicles have in common with ChatGPT?
Updated: May 18
Everyone talks about AI these days. More specifically about generative AI such as ChatGPT, DALL-E and others, that can be used to create texts, audio and video content. It can also code, create music or even poetry. How cool is that?! Moreover, it is said that it has practical business-relevant implications too: it can be used to generate new product designs, analyze complex data sets, and present conclusions of large-scale studies.
I shall leave the very complicated question of how generative AI does all these things to the technical experts. I just want to point out one idea here. ChatGPT, BERT and all the other AI chatbots rely on self-supervised learning and are being fed with a massive amount of data (more or less the entire internet), thus, they become able to generate predictions. Simply put, the outputs they generate are combinations of the data used to train their algorithms. Therefore, sometimes the outputs are highly accurate, sometimes they are wrong, biased, or unethical - often violating copyright issues. Fortunately, this technology is still under development, training, and regulation. But it’s impressive, that’s for sure.
What’s even more interesting, is that the underlying technology of such chatbots, machine learning, has been around for years, and it has been used in several other business areas and industries too. For example, in Computer Vision, machine learning has been used for years as a predictive model able to detect objects and classify scenes in images based on neural networks trained with human-labeled data (supervised machine learning). More recently, self-supervised learning algorithms are also being deployed mainly for research & development purposes at Bosch, for example, for the development of technology for automated vehicles.
When we talk about automated vehicles, we mean that they are sensing their surroundings, they are collecting, processing, and transmitting data and they are actionable, in the sense that the AI-derived conclusions are turned into actions that are safe and reliable, replacing and simplifying the typical driver interaction. For vehicles to become completely “aware” of their surroundings, make anticipations about other road users’ behaviors, and most importantly, to act safely and responsibly in complex traffic scenarios, there is still a long way to go. But for sure, self-supervised learning algorithms play a crucial role in their development process. Here is why.
The vehicles ‘perceive’ their surroundings through sensors such as camera, radar, lidar sensors.
The supervised and self-supervised learning algorithms can use sensor data to understand complex traffic situations. All possible predicted sequences of events are precalculated when interpreting the data. This way, the vehicle AI learns to anticipate. By observing a large number of scenarios during the development process, the system can determine the characteristic behavior of different ‘objects’, which helps it to make forecasts.
Based on such forecasts, the AI makes decisions and derives actions.
Today, researchers act as gatekeepers for almost all AI decisions, however, in some specific use-cases such as the Automated Valet Parking System by Bosch (the world’s first highly automated driverless parking function), AI ‘acts’ autonomously in a controlled environment. In the future, after much more training, AI will be capable of making much more complex anticipations and decisions, which will mean that automated driving can come true in more diverse traffic scenarios, not only in highly controlled ones.
Making this happen is part of our day-to-day business at Bosch Engineering Center Cluj, where we work on all levels involved in such applications, from the development electronic control units, data pipelines, cloud platforms and data science & AI solutions for various use cases from the mobility sector and not only. Moreover, we work not only on automated but also sustainable mobility by supporting the transition to electrification via battery and fuel cell systems.
You want to see such AI-based solutions in action? Check out the Bosch booth and tech demos such as the Mercedes-Benz EQS equipped with video-based Automated Parking function, live demos with AR & VR technologies, solutions for EV Charging Stations, the Electric Steering Cart, a Kawasaki Ninja H2 SX equipped with radar-based Rider Assistance System, e-Bikes, and more at the Business Summits and Mobility Expo Area! Moreover, listen to Bosch experts’ talks at the Innovation & Software Architecture Summits: Tobias Matter on the Future of Innovation & Cătălin Golban on Current Trends in Video Perception for Automated Driving. Later, join the Tech Expo stage for presentations with Augusta Ene on the Impact of Technology and Digitalization on our Lives, Gabriel Arcaș on Innovative mobility via Cloud-based platforms and services and Alin Pantea on Hydrogen fuel cell technology for a new era of mobility, and find more answers to your curiosities.
If you need more info about Bosch Engineering Center Cluj, visit https://www.bosch.ro/en/our-company/bosch-in-romania/