Jmuzz wrote:find_bruce wrote:Does anyone know the conditions imposed on Uber for their test? I expect that they would include a requirement that a licenced driver be behind the wheel at all times.
The state has authorised full autonomy (though now suspended). Potentially you can sleep in the back seat.
Whether this car was in that category hasn't come out yet.
So it will depend on that cars classification.
Regardless of the classification the driver may have had a job description to supervise at all times, or they may have only had a duty to be there to push emergency stop within 5 seconds if it went completely crazy.
Unknown at this stage.
The Economist dated March 3rd -9th has several articles on autonomous vehicles and is possibly worth reading - some of the articles are online but most will be behind their paywall I guess - anyway having read that article my suspicion would be very strong that the Uber vehicle involved in this death would be fully autonomous (level 4 in the jargon) because Uber is only really interested in developing driverless cars to cut a cost out of its business model therefore has no interest in developing vehicles that require human intervention
I wasn't really that enthused to watch the video of a vehicle striking and killing someone but I'm actually glad I did as it wasn't what I was expecting....was expecting the drivers nightmare scenario of someone simply stepping off the sidewalk/pavement directly as you are passing giving very little reaction time. This was a situation where there was reaction time with the lady pushing her bike across the other carriageway and into the path of the vehicle.
One of the Economist articles (From here to autonomy) talks about how autonomous cars perceive the world using cameras, radar and LIDAR and the need for systems to learn how to interpret that data - to be able to identify objects so can then "choose" how to react:
"....pay people to label images manually. Mighty A1, based in Seattle has an online community of 300,000 people who carefully label street scenes for a range of automotive clients..."
more worryingly: "Imaginery from video games such as "Grand Theft Auto", which have strikingly realistic street scenes can also help. Because the software knows what everything is....."
Also seen somewhere an article that talked about future autonomous vehicles being linked so that they can learn from each other how to deal with new "images"/situations
So in this case the vehicle failed to identify and more worryingly then failed to react to an unknown object in the other lane that was moving towards it....
seen several quotes from those involved in developing autonomous vehicles that go along the lines of this one:
"....choosing between hitting a group of children in the road or swerving and hitting another vehicle. Many people working in the field think that such questions do not reflect the real world...best course of action is to slam on the brakes. AVS have superhuman, 360-degree perception and much faster reaction times...." Danny Shapiro of NVIDIA Economist From here to autonomy
as to human intervention in automated systems believe a little research would find a long list of errors either not interfering or interfering and making the wrong decision based on inattention or poor training - from nuclear accidents to shipping, trains and planes
having a bored guy sat in the car was never going to get near saving that ladies life
lots of questions and as to regular drivers kill many people everyday as one commentator said "better than human" is a pretty low bar