Point 1: I would not have been asleep like the fat chick driving the car. Point 2, I would have swerved. Even then I would probably have hit the pedestrian. But there's good chance the algorithms didn't include pedestrian with bicycle. And a sleeping safety driver is a negligent safety driver.
The video's frame rate was 30 frames per second (33.3 milliseconds per frame).
From the time the woman's white sneakers were first barely detectable, (illuminated by the car's headlights) to the impact, there are 35 frames.
That's only 1.166 seconds. Even with vision equal to that of the camera, I seriously doubt that a human would have perceived a threat until much of the woman's legs were illuminated -- giving only a fraction of a second to respond.
The real kicker is that the car was in a well-iluminated area, and the woman was well outside that area. That would have compounded the difficulty of making the dark-to-light visual adjustment in time to perceive danger and to react effectively.
Here's a thorough article on the components of reaction time -- showing that the "standard" 1.5 second reaction time is, often, insufficient...
I recommend that you read it...
~~~~~~~~~~~~
In this instance, I suspect that I would not have outperformed the computer. YMMV...