What are currently the biggest challenges in implementing open source and AI technologies in the automotive industry?
There are two aspects that need to be examined. On the one hand, there are technology-related challenges. The approach to AI in the automotive sector needs to evolve. Current solutions lack scalability, reliability, and financial viability because they have high storage and computing power requirements.
On the other hand, we are seeing a shift toward Software Defined Vehicles. OEMs and Tier 1 suppliers need to take a software-first approach. This could mean reshaping corporate cultures, increasing agility and shortening development cycles. When it comes to regulations, requirements and the path to autonomous driving, there is still a lot of work to be done in the industry, both by regulators and other stakeholders.
In your opinion, which AI projects are currently particularly relevant for the automotive industry?
We are currently observing two different trends in the automotive industry. On the one hand, we see the continuous improvement and further development of ADAS technologies (Advanced Driver Assistance Systems). On the other hand, there is a strong push to develop fully autonomous driving (AD). Although these two areas have different main objectives, I believe it is crucial to facilitate the seamless transition from ADAS to fully autonomous driving and to exploit the synergies in both areas. These synergies can include technological breakthroughs, cost efficiencies, and improved reliability. This consideration underlies Autobrains' efforts to develop core technologies that have the versatility to be applied, extended and specialized in both areas.
What transformation is required on the part of OEMs and suppliers to successfully navigate the path to AI and the next level of automotive computing?
OEMs need to open up to collaborative partnerships on an equal footing. The emergence of Software Defined Vehicles is transforming the automotive industry and underscoring the need for collaboration between OEMs and technology companies. The complex interaction of software and hardware components in a Software Defined Vehicle is beyond the ability of any single company to effectively manage the development process.
How will partnerships along the value chain be reorganized?
Collaboration is the key factor here. As systems and vehicles become increasingly complex and rely heavily on multidisciplinary approaches, collaboration between experts and specialized companies is essential.
Different specialized companies need to join forces - from our perspective, this collaboration would involve platform licensing. To keep up with the increasing speed of the market and international developments, these collaborations must function like a well-oiled machine.
How will AI influence the development of autonomous driving?
AI is the key to autonomous driving and the most important technological challenge that must be solved to enable safe autonomous driving. Without the right AI approach, we will not see safe, reliable and scalable autonomous vehicles on the road.
How can the high safety requirements be met at the same time?
There are two main perspectives on security. The first is about protecting the vehicle and all its software components with robust cybersecurity measures, for example by working with strong partners. The second perspective is about designing a system securely from the ground up and ensuring inherent baseline security.
When it comes to evolving the system, some privacy concerns arise. Ensuring data privacy depends on relevant regulations and close collaboration among various stakeholders, including legislators, manufacturers, suppliers and consumers. These stakeholders must work together to create a legal framework and adhere to it accordingly.
Beyond simply complying with legal requirements, technology companies are in a position to provide transparency, encrypt data, and keep overall data volumes low. Autobrains' AI approach makes this possible. The AI abstracts and condenses objects into "signatures" using lines, colors or other structures. These signatures take the form of hyperdimensional binary representations that are used in training the system and decoding its environment. This approach leads not only to abstracted data, but also to a significant reduction in data requirements and volume.