I was riding my bike to work this morning and hit a spot where the cycle track ran out at the end of the block and was replaced by a bike lane at the other side of the intersection. This was no big deal but did require me to veer slightly to the left, a little closer to the car lane.
(Here's a diagram, since I find it hard to visualize these things and Google Earth isn't up-to-date on the Mueller development.)
I heard a car coming up behind me right as I hit the intersection, and I fully expected it to gun past, which is what I prefer and what most drivers used to sharing the road with bicycles do. Just get past me as quickly and safely as possible and we'll be out of each others' hair forever.
This car instead slowed and hovered behind my left side as I drifted over into the bike lane. Argh, dude, just pass me! You're making me nervous.
I turned around to see how close it was and realized it was a Google self-driving car* on a test drive, cautiously waiting while I made my lane shift.
Once I realized what was going on I relaxed because I knew I wasn't in danger of being plowed into by an inexperienced driver or, worse, a texting one. Still, given how cars and bikes usually interact, the Google cars' real-world behavior could use a little finessing.
Self-driving cars are a fantastic idea, but this very brief encounter made me realize the transition period between human and robot drivers means we will have to account for two sets of behaviors and two sets of assumptions. It might get hairy for a while.