Autonocast #24: NTSB Member Christopher Hart

Autonocast Episode #24

The National Transportation Safety Board (NTSB), investigates collisions in the USA across all modes (including pipelines), primarily involving Commercial Vehicles/Technology. They make recommendations, not policy, so they don’t end up investigating themselves. Typically automotive investigations are more focused around commercial trucks, though at the time of recording they were investigating the Tesla collision. Frustratingly, this episode was recorded before the results of that investigation were made public. They did, however, discuss Hart’s article about lessons autonomous vehicle makers could learn from aviation’s experience in automation.

Things to Note

Risk Compensation

They ask the question I wished they had of Dr. Pratt. Is there a compensatory effect of safety-enhancing autonomous tech? Hart mentions similar concerns occurred in aviation around improved tech, in particular improved weather detection might encourage pilots to fly closer to storms.

AV Efficiency

There was a further intersting discussion about efficiency. While there’s a lot of talk of electric autonomous vehicles, most testing doesn’t actually happen in electric vehicles (except Tesla) because autonomous systems require a tremendous amount of power: on the orders of kW. Humans, for all this discussion of robots replacing us, can operate a vehicle using about as much power as a modern light-bulb, on the order of less than 50W

My Thoughts

Risk Compensation

If risk = probability of incident * consequence of that incident, my thoughts on this are that technologies that reduce incident consequence will not encourage risk compensation. In aggregate, nobody drove in a manner to have a greater likelihood of a crash because airbags or improved crumple zone. However, it has become common knowledge in urbanist circles that people will drive the speed at which they perceive they can safely control the vehicle. People will exceed speed limits on streets (and highways) that are designed to be safe at speeds above the speed limit. Similarly, I would hypothesize that probability-reducing tech like Traction Control, ABS, better tires, could have encouraged risk compensation. What’s worse is that, combined with tech to increase survival rates within vehicle, there is a resulting externalization of crash consequences.

###AV Efficiency This seems to be a puzzling… regression in a certain sense. If there’s some consensus among “mobility experts” (and some manufacturers parroting them) that the future of AVs is electric and shared, isn’t initial development of them being nearly exclusively done on gas-powered consumer cars a regression? I don’t really have a sense of how these power requirements affect performance of existing electric vehicles. In particular, when I would’ve thought that an AI would likely be a more efficient driver overall.