3 ms·
That is a valid point, but the report also says this: "Safety is our highest priority and Waymo test drivers are trained to take manual control in a multitude
by mannykannot 10y ago
That is a valid point, but the report also says this:
"Safety is our highest priority and Waymo test drivers are trained to take manual control in a multitude of situations, not only when safe operation "requires " that they do so. Our drivers err on the side of caution and take manual control if they have any doubt about the safety of continuing in autonomous mode (for example, due to the behavior of the SOC or any other vehicle, pedestrian, or cyclist nearby), or in situations where other concerns may warrant manual control, such as improving ride comfort or smoothing traffic flow."
This appears to throw doubt on the relevance of the statistics to driverless vehicles because drivers have been instructed to preemptively disengage where problems may be anticipated.
- lern_too_spel 10y agoAll of those disengagements, which have erred on the side of caution, have been reported, so why does it throw doubt on the statistics? The rate of disengagements has gone down so much that Waymo has announced a mobility service by the end of the year, which seems aggressive to me, but within two years seems perfectly reasonable. After all, they can limit service to the environments that they know they'll have no problems with.
- mannykannot 10y agoThe statistics are not in doubt, but their applicability to the safety of driverless vehicles is, given that the drivers preemptively disengage in precisely those cases we are most interested in. The 1 in 14000 miles figure you quote effectively takes the position that they did not happen. Waymo simulates those situations to estimate how they would have played out, for the purpose of improving its algorithms. As a first analysis, there are three possible outcomes: the car would have handled the situation safely, it would have disengaged, or it would have done something that led to at least the risk of an accident. The report does not appear to allow us to distinguish between them, probably because they are somewhat speculative. This brings us back to where I joined the thread: the idea that 'slam on the brakes' is a fallback position for not having a driver present does not work when the system fails to recognize a developing problem.