Tech Alert

Navigating Driver Privacy and Safety of Electric Vehicles, Self-Driving Vehicles

A growing number of connected electric vehicles, as well as the evolution of self driving and automated vehicles are putting a greater demand on processing power. New technologies are advancing rapidly with the introduction of new processing methods, according to experts at BrainChip Holdings Ltd (ASX: BRN), (OTCQX: BRCHF), a leading provider of ultra-low power high performance artificial intelligence technology.

These automotive systems rely heavily on Artificial Intelligence and Machine Learning (AI/ML) to train an increasing number of sensors, components, image and video processors in each vehicle. Autonomous vehicles and near-autonomous vehicles are predicted to generate 12-15 terabytes (10¹⁴) of data for every two hours of driving, and all this data has vulnerabilities as it is uploaded to the cloud. Consumer advocates have raised alarms about operator and passenger privacy, including location data, driver health, speed, and more with the dependence on the cloud.

“Many of the concerns about driverless cars and driver assist systems can be addressed with improved AI/ML operations and internal components,” said BrainChip Founder and CEO Peter van der Made. “Safety is a particularly salient one, but energy efficiency, privacy and security are critical considerations for the automotive industry and their supply chain to address.”

Notable ways that improved technology “under the hood” will reduce accidents, protect data, and conserve energy include:

Real-time learning

Improved chips can perform “incremental learning,” and add to their knowledge of the world as they are confronted with new information. Object recognition is one situation when real-time learning is impactful – a car needs to “see” whether an object in the road is a rock, an animal, or a plastic bag and be able to recognize the differences of each to react accordingly. Under current AI/ML processing methods, all are viewed as obstacles.

“Real-time incremental learning, sometimes called one-shot learning, makes it possible to train a chip within a fraction of a second, and trigger corrective action,” said van der Made. “As this is widely adopted, the safety improvements will be enormous.”

On-chip learning

Traditional microprocessors are too slow to perform the type of calculations that are required to recognize objects. A large array of parallel operating cells, each operating according to the same principle as brain cells, perform rapid computations in the vehicle, rather than sending data to a cloud / data center and then waiting for instructions. Not only does this reduce latency so decisions are made faster, it removes the need for internet connectivity, so the vehicle continues to operate even when there is no internet available. And, by retaining data within the vehicle itself, instead of transmitting it to a remote location, security and personal data privacy is vastly improved.