The surveillance issue is a major one but requires its own post. Briefly: Seems to me that roadside sensors will provide little surveillance information beyond what is already captured. Lidar sensors (to my knowledge) are not precise enough to identify faces. And while they might identify license plates, this is true of car-mounted systems as well. (Let’s set aside the fact that license plate-based surveillance is already occurring on a vast scale with the growing police use of license plate readers.)

Finally, personal location data will only get more precise as dense 5G and public WiFi systems are deployed, and this will happen independent of roadside sensors. Since 5G networks and roadside sensors are privately operated, the underlying surveillance and data collection issue is the court-created third-party doctrine.

I suppose I’m an AI pessimist in the medium term and an AI optimist in the long term. For example, there appears to be a consensus that we will not have mass use of AI-dependent Level 5 AVs in the next decade. But I think we could have mass use of Level 3 AVs in the next decade if there is a system of roadside sensors making Level 3 AVs safer and more acceptable to the public. And while I didn’t get into this much in the piece, I’m pretty agnostic about what basic roadside infrastructure (poles, fiber conduit) is used for. AVs are simply the most likely and most profitable application. But such infrastructure has tremendous option value and could be used for any application that requires precise, location-based services, e.g., drone flight management, AV traffic management, AV deliveries, home security, and Internet access.