You are reading a single comment by @Oliver Schick and its replies.
Click here to read the full conversation.
-
You might have mentioned that the Tesla driver was actually killed:
The Tesla statement added: “This is the first known fatality in just over 130 million miles where Autopilot was activated.
“Among all vehicles in the US, there is a fatality every 94 million miles.”
And yes, it's not the same as 'self-driving'.
How do they know the 130 million miles figure? Computer tracking of these cars?
Oliver Schick-
Estimated US fuel sales / avg mpg of US vehicles / US road fatality rate attributable to vehicles.
-
How do they know the 130 million miles figure? Computer tracking of these cars?
And yes.
To an extreme.
-
They'll be some analytics in the software that register when it's on, how long it was on for, how far the car travelled with it on, etc.
Velocio
andyp
As everyone has, I assume, already heard a nasty crash apparently occured this past 7 May when a Tesla-S apparently did not see the white side of a turning tractor trailor and drove "right through it" (the driver was apparently playing "Harry Potter" and so not attentive to road obstacles).
First indications in the news is that the camera sensors could not segment the tractor side from the sky. While this could indeed have been the case what I find quite disturbing is the fundamental lack of LiDar and other sensors that might have worked around the blindness-- with computer vision there will always be conditions where a ANN is either blind or sees things that are not there (both sides effects of CNNs).
In defense of Tesla their current "autopilot" is not really a solution to "self driving" but to "auto-steering"-- or, more properly, steering assist. Interestingly the same driver apparently was also saved from a near crash by the "auto-pilot" a month earlier-- perhaps why he put so much trust in the technology (which despite this most unfortunate accident still demonstrates a dramatic increase in safety over wholly man controlled autos).