4 million miles is proportionally less than the width of a gnat against the history of driver miles in the US across all cities, less than a rounding error.
More crucially, these are 4 million entry level service rollout miles .. a time when the company has highly paid engineers paying scrupulous detailed attention to the condition of the cars and the tuning of the automated responses, with jobs dedicated to the avoidance of PR incidents.
comte7092 has pointed out other factors - but these are not "typical miles" these are the miles before typical sets in - who can say if the stats will be maintained as sensors age and maintainence becomes an after tought in the manner typical of human driven cars.
I hope the figure holds up, I'm commenting here on the apples to oranges issues in the comparison at present.
> Does it not drive the same roads as human drivers?
No, human drivers drive on roads beyond selected neighborhoods of Phoenix and San Francisco.
It is unknown how generalizable the results of the study would be beyond the conditions experienced on the roads in the areas Waymo had been deployed in. We might see significantly improved safety or we might not. But given the messaging from Waymo, the presumption would seem to be that performance would be worse in other areas/conditions.
I am not really sure what your question is asking. More accidents on what roads? If you’re talking about the 99% of roads in the country where Waymo doesn’t currently operate, then I can’t really take your question seriously.
We are talking about relative performance of an automated system vs human drivers. The automated system has been specifically tuned to perform well on a certain set of roads. The point is that there is no reason to assume by default that the system would maintain good relative performance under a drastically different set of conditions.
More crucially, these are 4 million entry level service rollout miles .. a time when the company has highly paid engineers paying scrupulous detailed attention to the condition of the cars and the tuning of the automated responses, with jobs dedicated to the avoidance of PR incidents.
comte7092 has pointed out other factors - but these are not "typical miles" these are the miles before typical sets in - who can say if the stats will be maintained as sensors age and maintainence becomes an after tought in the manner typical of human driven cars.
I hope the figure holds up, I'm commenting here on the apples to oranges issues in the comparison at present.