-
I think you underestimate how far the technology can go. There are simple prototypes of sensors at MIT that can literally see round corners, using shadows to identify potential hazards. In the next "few" years, cars might be able to see children hidden by parked cars before they are even in view and bikes before they have even turned into the same street as the car. It actually works pretty well already, on a very small and simple scale.
https://www.digitaltrends.com/cool-tech/mit-camera-see-round-corners/
Edited for better article: https://www.google.com/amp/s/amp.theguardian.com/science/2019/jan/23/creepy-things-ordinary-digital-camera-computational-periscopy-can-see-round-corners
It's going to take a while, but I'm pretty confident the tech will get there one way or another. Don't focus on refining what we have now, at the expense of ignoring what will exist in the future. It's human nature to dismiss the capability of nascent technology, and opinions like these are often wrong.
I actually think AI will create consistently, and therefore more safely driven cars; no impatient accelerations, squeezing into gaps, aggressive close passes, aggressive misuse of the horn. No more sometimes forgetting to look when pulling out / turning left. They will probably prevent doors being opened into traffic, too. And, possibly most importantly, no-one at the wheel to get angry at. We waste so much energy fuming over the person at the wheel whom we perceive to have been rude / dangerous. Driverless cars will be entirely predictable, as cyclists we will learn their behaviour, and we can simply work around them. I'm glad you brought this up because I hadn't thought about it before.
It's all fine of course until they start self replicating and trying to wipe us out.