He's saying that if a stretch of highway has traffic volume of 10 million trips taken on it in a year and an average of 2 deaths per year, that is still much safer than a neighborhood street which sees 10,000 trips per year and averages 1 death per year.

(numbers made up to emphasize a point, a neighborhood street with 1 death per year is pretty obviously unsafe)

That’s not the correct comparison anyway.

They likely say that more people use fast roads, so it's expected to have more accidents. The safest road is one never used.

Are there suddenly no pedestrians and cyclists on the roads with the cars? A cyclist hitting a pedestrian at 30km/h (fucking fast for a normal cyclist btw) is a rounding error both in terms of how much it happens and how deadly it is compared to a car hitting a pedestrian at 30km/h