BERLIN — Self-driving cars are benefiting from advances in a host of technologies, including sensors, navigation — and blockchain? XAIN AG is applying a decentralized approach to machine learning that it said offers greater privacy, security, and efficiency for data storage. The startup is working with Porsche AG to enhance the driving experience with artificial intelligence and distributed systems.
XAIN, which stands for the “eXpandable AI Network,” began as a research project at Oxford University and Imperial College London by Leif-Nissen Lundbaek and Prof. Michale Huth. Lundbaek is now CEO, and Huth is chief technology officer.
In 2017, XAIN won the first Porsche Innovation Contest against 120 startups, which ultimately led to its collaboration with the carmaker. While the digital ledger technology known as blockchain was originally developed for the financial industry, distributed systems could be useful for other types of transactions and the Internet of Things.
Last year, Earlybird Venture Capital led a seed round of €6 million ($6.63 million U.S.) for XAIN. In July, Dominik Schiener, the 22-year-old co-founder of IOTA, made a strategic investment in the company.
XAIN applies FedML to vehicle data
Large amounts of data are needed to train machine learning algorithms, especially those for self-driving cars. The goal is for such vehicles to adapt to individual drivers and to navigate safely and accurately in real time.
As connected vehicles monitor driver behavior and collect data for aggregation and analysis, concerns have arisen about data security and privacy. Most AI companies claim to anonymize such data, typically before it is stored in a model. However, incidents such as FaceApp’s centralized storage of photographs from smartphones for facial recognition have not assuaged those concerns.
XAIN is developing an infrastructure to train AI using Federated Machine Learning (FedML) to protect privacy. It trains data in separate AI models that are then aggregated, enabling different users to collaborate and gain insights while sharing only the data from the training step and not the data itself.
“One of the biggest challenges going forward in AI is figuring out how to simplify the training process, which is incredibly onerous, while also keeping data secure,” stated Dr. Christian Nagel, founding partner at Earlybird and a member of the XAIN supervisory board. “FedML addresses that challenge, while also reducing the cost of training.”
The first application running training models on XAIN’s FedML technology is ANDY, a system for automated invoice processing.
The startup recently announced that it complies with the European Commission’s General Data Protection Regulations (GDPR). In addition, XAIN said that larger companies can use its FedML approach both internally and externally for product development.
Autonomy and luxury
Porsche does not plan to substitute AI for human drivers. Instead, it plans to use AI and machine learning to assist them with adjusting speed, predictive maintenance, being aware of road hazards, and dynamically adjusting the cabin environment, depending on who is driving.
Last month, Porsche invested $2 million in TriEye. The Israeli startup is working on technologies to apply AI to short-wave infrared and improve visibility for both human-driven and autonomous vehicles.
“A Porsche will always be a car that you will want and be able to drive yourself,” said Lutz Meschke, vice president of Porsche’s executive board and a member of its executive board for finance and IT. “Nevertheless, there are of course aspects of autonomous driving that are interesting for us — traffic jam assistants or automated parking, for example.”
While fleets of self-driving vehicles such as taxicabs and shuttle buses are being tested at lower levels of autonomy, self-driving cars for individual consumers will likely start out with driver-assist features in luxury vehicles because of the expense, noted XAIN.
“Our customers always want it all,” said Klaus Zellmer, CEO of Porsche North America. “They want the possibility to use autonomous drive mode, but they want to really engage with the car as well.”
The Robot Report is launching the Healthcare Robotics Engineering Forum, which will be on Dec. 9-10 in Santa Clara, Calif. The conference and expo will focus on improving the design, development and manufacture of next-generation healthcare robots. Learn more about the Healthcare Robotics Engineering Forum, and registration will be open soon.