[minwcnt5]

> Current state of autonomous driving is not sophisticated enough to predict cars crossing lanes that's turning and crossing your lane from oncoming traffic.

This is just completely false. AV software stacks absolutely can and do predict off-nominal behavior like you describe. In general they predict better than humans, too.

[fio_ini]

you're literally talking about using a lidar and some object detection (for car and other large objects) to predict human behavior. you obviously don't realize that the human behavior is in the other car with eyeballs, mouth, head movement, body movement. that part is not being tracked. that part is human behavior. and that's where the car movements start. so everything that the "AV" is predicting is super old information because the human is full seconds ahead of doing whatever the computer is reacting to later by seeing a large object move seconds after the human brain has caluclated it's behavior. the computer is not solving the human part which you claim it does.

[fio_ini]

this is something that humans do seamlessly and you don't realize because you're probably robot. but we look eachother in the eyes, hand wave, point, etc. not always (these are the asshole drives) but you get the point. humans are superior and always will be until it's completely computer cars on the road that are networked so they can be networked the same way humans are networked to gether.