Keeping Humans and Autobots Honest

My hunch (and secret wish) is that Elon Musk’s secret Tesla Master Plan, the real plan, is to deliver personal "Autobots" in the near future. As the first self-driving semis and passenger cars enter the commercial transportation market, insurance companies will have to rethink their ways to assess risk in the autonomous vehicles category. Even in the fictional world, humans have a history of not trusting sentient robots. Try convincing your local law enforcement official that Bumblebee, and not you, was at fault for speeding. Does one penalize and assign blame to the owner or the manufacturer of an autonomous vehicle when things go wrong? What we need is an indisputable “trust rail” that humans, robots, insurance companies, and authorities can use.

The three Tesla crashes in July brought up two important questions: What caused the accidents? Were the self-driving systems enabled? The answers to the first question, which pacify our curiosity and help us improve technology, is easy to determine based on the physical evidence and eyewitness accounts of the crashes. The answers to the second question, which assign blame to either the operators or the manufacturer, require more scrutiny as their validity depends on who provides them and what they stand to lose.

Montana Highway Patrol Trooper Jade Schope’s statement in USA Today was, “I have no way of verifying whether it was or wasn't”—referring to whether the self-driving mode was on or off in the latest crash—highlights a scenario where it is the driver's word against the manufacturer’s. Even if the driver was lucky enough to walk away with minimal injuries, can a third party trust the driver’s recollection of events? Drivers have been known to misrepresent the facts. According to Edmunds, 25 percent of bodily injury claims and 10 percent of auto damage claims resulting from car accidents are bogus. On the flip side, how much do we know about car manufacturers’ data collection practices? People and companies have a history of bending the truth and tampering with data (cough, cough, Volkswagen).

Data does not lie until it does. At the heart of this issue, who can we trust? How can we access untampered information that is explicitly honest? We are on the brink of adopting autonomous self-driving cars, so it’s only prudent that we explore reliable ways to record, manage, and validate vehicle data. According to a white paper published by Hitachi, a car will generate about 25 gigabytes of data per hour. That's a lot of data that needs to be securely recorded and accessed. Besides scalability, we must also find ways to protect the car owner’s privacy.

Enter blockchains. These distributed ledgers make it hard to lie and keep users honest by providing a way of publishing and validating immutable data. Blockchains as a publishing platform can outlast any particular organization using them. Third parties like insurance companies can use blockchains to secure and verify all sensor data and the identity of the vehicles. Blockchains can be as powerful and useful as the utility they provide.

More incidents like these are on the horizon. We need to ensure that the data sources are kept honest, as the data they provide will be used to craft public safety policies and industry guidelines. In the event of a fatality or severe property damage, even the records subpoenaed by the court should be provable beyond doubt. There has been significant progress in storing and analyzing data in the past decade, but we are far from making it trustworthy.

In the case of Tesla accidents, imagine that data generated by the car was being “hashed” in real time. We wouldn’t need to publish all the sensor data into a blockchain but only time-stamped proofs or certain pertinent metadata (such as the autopilot mode being on/off). Here is a simple example of how vehicle data can be secured using Factom, an open-source, scalable blockchain for enterprise data: Using Carvoyant APIs, a few lines of code (written by our whiz intern, Steven Masley) could be used to secure information on the Factom network. In a similar fashion, manufacturers could build audit trails of data and metadata that could be shared securely with external parties.

It’s not hard to imagine a future where the National Highway Traffic Safety Administration (NHTSA) will mandate blockchain audit trails for all autonomous car data. Technology is only as good as the service it provides. Currently, we prefer to store information on paper, as atoms are harder to change than digital bits in a database. However, in an immutable, digital world, blockchains can keep humans and machines honest while allowing us to scale the way we manage and consume digital data.