Yes, the pedestrian bears part of the fault (she was jaywalking at a minimum).
The safety driver's primary job was to intervene. What else are they there for?
Do we know who initiated turning down the sensitivity of detection? Maybe it was a boss who ordered engineers to do it because she figured any problem would be caught by a correctly functioning safety driver.
BTW, it appears the video originally released from the car camera (such as here) showed little beyond the reach of the headlights and was oddly dark. Others took video of the same route a night or two and there seems to be adequate street lighting (such as here). In the "dark" video, it appears the street lights are on so that alone wouldn't explain it.
Yep - I predict Boeing (or, their insurance company) may be paying out a bit of money for the MCAS disaster. It was documented, but it appears not properly in "differences training" (moving from one model of a type to another model -- far less extensive than moving to another type). It probably should have been (and, probably, even highlighted).
I don't doubt that we could gather that data over a couple of years and a few tens of thousands of observation points -- but I'm not aware that has been done.
Obviously only the data from observation points (or groups of nearly identical observation points) which observe a statistically significant number of cars in full self driving mode can be considered (which, today, would be remarkably few points and would leave out many adverse road conditions). The data would have to include if a self-driving car with a "self driving" license plate was actually in "full self-driving mode" for the stretch of road after the observation point (if not, it would be counted as 'human-piloted' or something). As well, if no self-driving car in full-auto mode had passed in a while or if an unusually high percentage of them were not in full-auto mode at the time, all data would need to be adjusted to take into account for if the reason is that they were not sent out in this time-frame/conditions or self-driving mode was turned off due to conditions (either automatically or due to human discretion) and probably exclude all human driven cars (and accidents) for this period. This would require that self driving cars signal their mode to the observation points OR that logs of all self-driving cars be available to determine exactly where they were in "full self driving mode". This data could not be safely extrapolated to all the conditions that self-driving cars don't appear in statistically significant numbers.
Also, the priority should be on not injuring or killing innocent parties. If a drunk drives off the road and dies, that's on them. The fact the system kept them from driving off the road but that same system also killed a pedestrian who was legally crossing the street mid-block (that's legal depending on the situation in my locality) doesn't make it a "net even" system.
I think many things like good collision avoidance and lane keeping would help save a lot of lives. However, these are backup systems not to be relied upon - the driver should be admonished (and possibly reported to the police/department of motor vehicles if it happens too often per mile driven) if these systems activate.
There will always be confusing situations. For example, there will always be bad lane markings. In fact, just a few hundred feet from me now, they are repaving a fairly busy three/four lane road and there have been no adequate lane markings for three or four weeks. The crews put down reflective adhesive tabs in the approximate locations of the lanes immediately after repaving but many of these have broken and worn off now, a week or so ago they put "hairline" marks for the lane marking crews to follow and I believe these are the "definitive", but hard to see, lane markings. Unfortunately, the remaining reflective adhesive tabs and the hairline marks sometimes differ by over 18 inches so it's very confusing (esp. in the rain or in certain lighting conditions where one or both are very hard to see). Human drivers end up wandering all about but it doesn't really slow traffic that much and I've not seen a single accident (although they may well have occurred) as a result of this confusion. A self driving car would have little choice but to grind to a halt until a human took over in this case -- even if it had a database of where the lines "should" be, they aren't there which means the software shouldn't assume anything (i.e., a discrepancy between two conflicting inputs should be to do the "safe" thing). The software can't just continue driving rudderless at 40 or 45 MPH (64 or 72 KPH) while the human driver wakes up and figures out what to do so it would have to just stop until the driver was alert and had analyzed the sitution. If just 5% of the cars approaching this area were in self-driving mode, it would result in gridlock.
Perhaps roads are better maintained in Norway though and lane markings are always clear, but this is America and we like our poorly maintained roads damnit:)
Agreed - if the driver was looking away per training and performing a task the employer told her to do.
However it was my understanding that the driver was likely viewing a personal entertainment video based on this news report:
According to the spreadsheet of watch data from Hulu, Vasquez was streaming television episodes for approximately three hours the night of the crash. She was watching "The Voice" from 9:16 p.m. until 9:59 p.m. Police believe the crash happened while she was streaming that show.
If that is the case (and her employer didn't allow such activities while performing her duties) she was careless and/or negligent (I lean towards negligent).
A redacted version of the report is here but it's long (and not text searchable) so I've not confirmed that it says what the ABC news station article claims.
No, the person causing an accident is liable regardless of the level of liability insurance they have. If you have liability insurance on your car for $1M dollars and you cause an accident that kills a busload of people, you are liable for the other $99M of the court's ruling. If you don't have the money of course they can't get blood out of a stone but if you have assets, most of them (exceptions vary state by state) will be seized to satisfy the order -- as will most future income and assets obtained later.
This is why it's pretty stupid not to have coverage (often via an umbrella policy) that covers up to the fairly extreme cases of possible liability unless you're judgement proof (such as you're chronically homeless and unemployed). A good rule of thumb is to have at least enough insurance that the insurance company will fight to keep their money and there's enough in the pot that the injured party will grudgingly settle for the insurance alone in order to get a quick settlement and not have the case drag on for years and, possibly, result in getting even less if a jury is feeling sympathetic to the person causing the injury.
If you're stinking rich - $10B of fairly liquid assets, you can probably self insure for all personal liability though (after all, it's pretty hard for an individual to inadvertently cause via their fault, in their personal capacity vs. business capacity, more than $1B dollars worth of damages).
Actually, it was carelessness or negligence by two people (the safety driver and the pedestrian).
It was malfeasance by others (the persons who ordered the safety feature to be shut off/tuned down and the programmers/techs who turned it off/tuned it down and should have known that would be dangerous -- the "{Hitler, TheBoss} told me to do it" doesn't work. That's not to say that a low level tech who may have been told to "Change the setting for Sensitivity to 5" is liable if they had no reasonable way to anticipate that "5" was an unsafe setting.
There are no relevant statistics on self-driving cars in the real world. This is because there are NO self-driving cars available in the hands of consumers which self-drive in complete control in all the same conditions (snow, ice, rain, darkness, etc) on all sorts of roads (rural, dirt, highways, freeways, under construction, traffic control devices out or missing, road routing changed without notice, detours, worn out lane markings, areas highly congested with pedestrians and children etc) that humans drive in regularly.
Comparing the miles driven on carefully curated roads with safety drivers taking over at planned times (let alone unplanned times) to "real world" miles is absurd. One also has to compare the efficiency of getting from place to place -- if a self driving car decides to stop-and-go or just freeze on an on-ramp during rush hour because there's no "lawyer safe" thing to do where a human would have efficiently and quickly dealt with the same situation, the human piloted car can't reasonably be compared to the self-driving car as the latter is not performing the same function (at an extreme, a self driving car that never went over 3 MPH would probably result in very few fatalities - except due to passenger old age - but would be useless).
Thus, the claim is impossible to verify at this time.
I think it's likely that some categories of accidents will become much less common with self-driving cars (esp. those involving drunk drivers piloting the car in question and those that existing collision avoidance, lane keeping, etc would also reduce). On the other hand, I expect some categories of accidents will be more common with self-driving cars - mostly those where a human driver has to make a logical decision based an somewhat unusual circumstances (such as a downed tree or noticing that there is a stop sign but it's been run over or a new traffic control sign appearing warning of a difficult to anticipate transient situation).
(Hopefully, true self driving cars won't have a penchant for smashing into big red fire trucks, sometimes with flashing red lights, like Tesla autopilot does. I assume/hope the adults are implementing the real self-driving cars and Tesla is the only one that hired fresh outs from community college to do design and write their software and follow pedo-Musk's every whim.)
I always assumed that the CSRs could see every keystroke I type in the chatbox.
If I have a long question (often the first question is long for example), I usually type the text into emacs (and, if I'm being picky, spell check it as well) and then cut and paste into the chatbox just to avoid confusion (such as me leaving out a "not" and later correcting it and the CSR never noticing my correction).
When someone claims person Y's life was worth $Z, I inquire as to if they had >=$Z life insurance on Y. If not, they are lying about the value of Y's life to them.
In this case, the question is "Were these files worth backing up in case a house fire destroyed them?". If the answer is "No", then I know how little they are worth (given the low cost of offsite backups either via sneakernet or via the internet or other communication medium). If the answer is "Yes", then the damages (if any) would just be the cost of retrieving the backup and restoring the files.
Why should a third party care more about your files than you do -- esp. when you explicitly decided NOT to back up your files and the third party's destruction of them was the result of an unintended bug?
Suppose a nurse looks at a patient's "chart" at 01:45 after the time switch and sees a medication was given at 01:30. Suppose the patient is allowed to, on request, have the medication not more often than once an hour (and, perhaps, there are additional limitations on the number of doses in a 24 hour period). If the patient asks for another dose at this time, does the nurse say "Sure" or "I'm sorry Mr Senile, you can't have another dose for another 45 minutes. You have probably forgotten, but you had a dose just 15 minutes ago by Nurse Jones who got called to a code blue a few minutes ago."?
We could of course, for example, on the west coast require "PDT" or "PST" to be included in all user input of times in ambiguity and displayed as well and train everyone for this once a year event (the change to "spring ahead" is less confusing and is not ambiguous even without the time zone designation differentiating from daylight and standard time).
I've read multiple news reports in just the past few years of people intoxicated on drugs or alcohol stepping out into busy traffic on freeways where pedestrians are not allowed at all, let alone in the roadway (note, the results are almost always the same but the good news is the addiction problem is fixed forever). I've never read (except perhaps on the National Enquirer cover or in The Onion) of your scenario happening.
Hmm... I wonder if you would like to be sentenced to prison. You would be surrounded by inmates who think outside the rules and aren't knee-jerk authoritarians which would thrill you. On the other hand, you would also be subject to a lot of knee-jerk authoritarians with badges which you might not like -- but that would give you even more opportunities than usual to show how much you think outside the rules. Tough call:)
In all of the cases in the survey I took at this site, SOME form of loss of life was unavoidable (although some were between humans and non-humans).
However, in many real world cases, there are likely to be differing probability of human death and, certainly, self-driving software should take that into account. A car hitting an elderly pedestrian squarely at 35 MPH is very likely to kill the pedestrian (maybe >80%?) but in a modern car with the highest crash protections, hitting a concrete barrier at the same speed is much less likely to kill the passenger (maybe 5%?).
In collisions with inanimate objects there are also differing risks of death or injury. If a semi-truck runs a red light as you approach the intersection and the car you're driving is about to be stripped of everything above the beltline - including the heads of you and all of your passengers - if you do nothing, I'm betting you would correctly swerve into a "soft target" like a trash can to avoid the truck. Both are "collisions" (a.k.a. "accidents"), but humans do make distinctions between collisions based on the imprecise odds of death and injury among all options. I expect self-driving cars to do the same.
These maneuvers are not necessarily "insane" - but there may be something in the "escape path" with some probability that it is each of (a) human, (b) domestic pet, (c) small wild rodent, (d) a patch of flowers in front of a business or home.
It sounds like you are proposing that if p(a) = 0.0000001 and p(d) = 99.999 you would still elect to kill the pedestrian? Most humans would not make that decision -- including humans on the jury for the civil case the pedestrian's widow and children file.
[Yes I know that it's quite possible that there is a small wild rodent or a small child hidden in the flowerbed so it's a little more complicated than I make it out to be.]
Should it swerve if that would result in no damage or injury to anyone if it detects that it will impact the pedestrian at a lethal speed if it just slows down?
Assuming that your answer is "yes" - what if it detects a domestic dog in the path of swerving, should it kill the pedestrian instead? What if it detects a squirrel in the path of swerving? What about a mouse? What about a mailbox? Remember, you're writing the code and/or developing the training sets -- your call, the only way out of making the call is to shut down the "self driving car" company you founded and invested all of your savings in.
This decision, of course, is no different than one you would be forced to make as a human driver in a similar situation -- but at least the jury in the human driver case can sympathize with the "split second" nature of it -- they won't be so sympathetic to the five year old block of code with the comment in front of it that says:
/* Kill the pedestrian, mice are cute and mailboxes will dent the bumper */
However, there are laws against some sorts of crossing between intersections in some areas. Where I live, there is one law that says that if you are within X feet (I don't recall what X is) of a controlled intersection, it is illegal to cross except at such an intersection.
Also, there are places where "no pedestrian" signs prohibit pedestrians so crossing a road from/to one of those areas is also illegal.
But, yes, where I live, you're certainly free to cross an undivided road on foot if there's no intersection within miles and no sign prohibiting it!
So, a self driving car in a case like this (or any of the millions of other similar cases) should never take evasive action - just smash into the pedestrians or barrier in front of it? Or, do you suggest that self-driving cars should ignore bicyclists and pedestrians (as they are both light and probably won't kill any passengers in the car) when evaluating evasive maneuvers necessary to "save the car and the passengers"?
Consider how you would write the requirements spec or testing criteria for self driving software. Would you just identify desired actions in such cases as "Undefined"? If not, what would you define.
That's naive -- the software has to do something here, it receives a bunch of inputs and analyzes a bunch of possible outcomes and somehow has to score them to decide to take an action (including "do nothing" which is, in itself, an action).
For example, surely a car should swerve to avoid a car that has run a red light if that will avoid a collision of any sort rather than just run into the red light runner and likely kill or seriously injure individuals in both cars. But, what if swerving would mean impacting a vehicle in the cross street (in another lane) that did follow the law and stop at the red light and this impact, due to being off center, presents a much smaller but still significant likelihood of death -- but to the completely uninvolved driver of the lawfully stopped car?
An alert driver will actually make a decision in such cases -- it may not be the right one and it may not be made quickly enough and most people probably don't know how they would really make such a split second decision. Our brains are not available for code review. Hopefully the self driving software/dataset and design has been reviewed and the reviewer has to determine if the code/dataset meets the requirements in this case - a SEGV probably isn't acceptable for example. If nothing else, fuzzing tests during development would result in situations like this and require a human (perhaps with a lawyer stitched to their side) to decide if the outcome was acceptable.
The survey presented assumes brake FAILURE in the scenarios. I.e., the brakes are applied but they fail to stop the car as expected.
In such a scenario, the software is presented with several inputs including at least:
1. In spite of applying the brakes the car is not slowing down quickly enough (perhaps it might not even be because of a defect in the car -- perhaps a leaking tanker truck full of vegetable oil had passed ahead of the vehicle you are in just 20 seconds earlier).
2. The car is approaching an unanticipated barrier (perhaps related to construction or perhaps as the result of some sort of localized natural or man-made disaster) at a high speed that, if impacted, will likely kill everyone in the car.
3. The software detects what it thinks is a "passenger safe" path around the barrier, but detects that a living entity (a dog or a person perhaps) is occupying that space and that the living entity will likely be killed if that path is taken (of course, if it's the vegetable oil case, the software might not know enough - do the weird road conditions also exist on the swerving path which might make steering ineffective?).
The software has to do SOMETHING predictable in this case - even if it is just to throw up its hands and reboot! Any unpredictable response to fixed inputs in this area should be caught in review before the product alpha. So, what is the requirement in this scenario?
So, if a pedestrian high on drugs illegally steps out in front of your car in the middle of a high speed road where there is no pedestrian crossing, you would choose to drive your car into a barrier to avoid the pedestrian even if that would result in the almost certain death of yourself and three other family members in your car? What if you were driving a carpool of kids and two of the passengers who would die aren't even your kids -- they are neighbor's kids?
The US is a nation composed largely of immigrants and their offspring, many who have arrived comparatively recently. In many cases they came not because it was convenient (getting to the US from Poland or Italy, for example, was not "convenient" before air travel - esp. for poor people) or because it was easy or because it was low risk. They subjected themselves to substantial risk, expense, and inconvenience to make the trip and survive in the US.
These immigrants, of course, left behind those that didn't have the same drive or interest in creating a better situation for themselves and their families. It would not be surprising that those who had the gumption to better themselves rather than sacrifice themselves for the "common good" would be looking out for themselves and their families more strongly than those that lacked such gumption and remained behind.
As well, the US has historically been one of the most diverse populations in the world (due to the source of our population) so the tribal "common good" notion is probably unsurprisingly much stronger than in monocultures like Japan or most of the Nordic countries.
The US seems to have done pretty well - esp. in light of having to deal with its very diverse population.
In the scenarios in the linked survey, there are ones where the pedestrians have a green light and ones where they are walking against a red light. I always prioritized legal pedestrians over passengers and passengers over illegal pedestrians. Note that in every case that which caused the situation was a brake failure - not something that the passengers are directly responsible for at that moment (unless they are knowingly using unlicensed/uninspected/unmaintained vehicles and that's why the brakes failed - but the scenarios don't suggest that's the case).
Legislation might require that the passengers in the car be prioritized below law abiding pedestrians in failure cases. This would encourage people to buy/rent/share the most reliable cars -- i.e., those that don't have as many failure cases that require making such decisions. It also makes the person responsible for the selection (the passengers) accountable for their actions.
Apple users obviously don't object to proprietary walled gardens else they wouldn't be buying Apple products. This is just a few more bricks on top of the garden wall and I would expect it to be celebrated.
Slashdot Mirror
Indeed, that left me scratching my head. Did the editors not get past seventh grade?
Yes, the pedestrian bears part of the fault (she was jaywalking at a minimum).
The safety driver's primary job was to intervene. What else are they there for?
Do we know who initiated turning down the sensitivity of detection? Maybe it was a boss who ordered engineers to do it because she figured any problem would be caught by a correctly functioning safety driver.
BTW, it appears the video originally released from the car camera (such as here) showed little beyond the reach of the headlights and was oddly dark. Others took video of the same route a night or two and there seems to be adequate street lighting (such as here). In the "dark" video, it appears the street lights are on so that alone wouldn't explain it.
Yep - I predict Boeing (or, their insurance company) may be paying out a bit of money for the MCAS disaster. It was documented, but it appears not properly in "differences training" (moving from one model of a type to another model -- far less extensive than moving to another type). It probably should have been (and, probably, even highlighted).
I don't doubt that we could gather that data over a couple of years and a few tens of thousands of observation points -- but I'm not aware that has been done.
Obviously only the data from observation points (or groups of nearly identical observation points) which observe a statistically significant number of cars in full self driving mode can be considered (which, today, would be remarkably few points and would leave out many adverse road conditions). The data would have to include if a self-driving car with a "self driving" license plate was actually in "full self-driving mode" for the stretch of road after the observation point (if not, it would be counted as 'human-piloted' or something). As well, if no self-driving car in full-auto mode had passed in a while or if an unusually high percentage of them were not in full-auto mode at the time, all data would need to be adjusted to take into account for if the reason is that they were not sent out in this time-frame/conditions or self-driving mode was turned off due to conditions (either automatically or due to human discretion) and probably exclude all human driven cars (and accidents) for this period. This would require that self driving cars signal their mode to the observation points OR that logs of all self-driving cars be available to determine exactly where they were in "full self driving mode". This data could not be safely extrapolated to all the conditions that self-driving cars don't appear in statistically significant numbers.
Also, the priority should be on not injuring or killing innocent parties. If a drunk drives off the road and dies, that's on them. The fact the system kept them from driving off the road but that same system also killed a pedestrian who was legally crossing the street mid-block (that's legal depending on the situation in my locality) doesn't make it a "net even" system.
I think many things like good collision avoidance and lane keeping would help save a lot of lives. However, these are backup systems not to be relied upon - the driver should be admonished (and possibly reported to the police/department of motor vehicles if it happens too often per mile driven) if these systems activate.
There will always be confusing situations. For example, there will always be bad lane markings. In fact, just a few hundred feet from me now, they are repaving a fairly busy three/four lane road and there have been no adequate lane markings for three or four weeks. The crews put down reflective adhesive tabs in the approximate locations of the lanes immediately after repaving but many of these have broken and worn off now, a week or so ago they put "hairline" marks for the lane marking crews to follow and I believe these are the "definitive", but hard to see, lane markings. Unfortunately, the remaining reflective adhesive tabs and the hairline marks sometimes differ by over 18 inches so it's very confusing (esp. in the rain or in certain lighting conditions where one or both are very hard to see). Human drivers end up wandering all about but it doesn't really slow traffic that much and I've not seen a single accident (although they may well have occurred) as a result of this confusion. A self driving car would have little choice but to grind to a halt until a human took over in this case -- even if it had a database of where the lines "should" be, they aren't there which means the software shouldn't assume anything (i.e., a discrepancy between two conflicting inputs should be to do the "safe" thing). The software can't just continue driving rudderless at 40 or 45 MPH (64 or 72 KPH) while the human driver wakes up and figures out what to do so it would have to just stop until the driver was alert and had analyzed the sitution. If just 5% of the cars approaching this area were in self-driving mode, it would result in gridlock.
Perhaps roads are better maintained in Norway though and lane markings are always clear, but this is America and we like our poorly maintained roads damnit :)
Agreed - if the driver was looking away per training and performing a task the employer told her to do.
However it was my understanding that the driver was likely viewing a personal entertainment video based on this news report:
If that is the case (and her employer didn't allow such activities while performing her duties) she was careless and/or negligent (I lean towards negligent).
A redacted version of the report is here but it's long (and not text searchable) so I've not confirmed that it says what the ABC news station article claims.
No, the person causing an accident is liable regardless of the level of liability insurance they have. If you have liability insurance on your car for $1M dollars and you cause an accident that kills a busload of people, you are liable for the other $99M of the court's ruling. If you don't have the money of course they can't get blood out of a stone but if you have assets, most of them (exceptions vary state by state) will be seized to satisfy the order -- as will most future income and assets obtained later.
This is why it's pretty stupid not to have coverage (often via an umbrella policy) that covers up to the fairly extreme cases of possible liability unless you're judgement proof (such as you're chronically homeless and unemployed). A good rule of thumb is to have at least enough insurance that the insurance company will fight to keep their money and there's enough in the pot that the injured party will grudgingly settle for the insurance alone in order to get a quick settlement and not have the case drag on for years and, possibly, result in getting even less if a jury is feeling sympathetic to the person causing the injury.
If you're stinking rich - $10B of fairly liquid assets, you can probably self insure for all personal liability though (after all, it's pretty hard for an individual to inadvertently cause via their fault, in their personal capacity vs. business capacity, more than $1B dollars worth of damages).
Actually, it was carelessness or negligence by two people (the safety driver and the pedestrian).
It was malfeasance by others (the persons who ordered the safety feature to be shut off/tuned down and the programmers/techs who turned it off /tuned it down and should have known that would be dangerous -- the "{Hitler, TheBoss} told me to do it" doesn't work. That's not to say that a low level tech who may have been told to "Change the setting for Sensitivity to 5" is liable if they had no reasonable way to anticipate that "5" was an unsafe setting.
There are no relevant statistics on self-driving cars in the real world. This is because there are NO self-driving cars available in the hands of consumers which self-drive in complete control in all the same conditions (snow, ice, rain, darkness, etc) on all sorts of roads (rural, dirt, highways, freeways, under construction, traffic control devices out or missing, road routing changed without notice, detours, worn out lane markings, areas highly congested with pedestrians and children etc) that humans drive in regularly.
Comparing the miles driven on carefully curated roads with safety drivers taking over at planned times (let alone unplanned times) to "real world" miles is absurd. One also has to compare the efficiency of getting from place to place -- if a self driving car decides to stop-and-go or just freeze on an on-ramp during rush hour because there's no "lawyer safe" thing to do where a human would have efficiently and quickly dealt with the same situation, the human piloted car can't reasonably be compared to the self-driving car as the latter is not performing the same function (at an extreme, a self driving car that never went over 3 MPH would probably result in very few fatalities - except due to passenger old age - but would be useless).
Thus, the claim is impossible to verify at this time.
I think it's likely that some categories of accidents will become much less common with self-driving cars (esp. those involving drunk drivers piloting the car in question and those that existing collision avoidance, lane keeping, etc would also reduce). On the other hand, I expect some categories of accidents will be more common with self-driving cars - mostly those where a human driver has to make a logical decision based an somewhat unusual circumstances (such as a downed tree or noticing that there is a stop sign but it's been run over or a new traffic control sign appearing warning of a difficult to anticipate transient situation).
(Hopefully, true self driving cars won't have a penchant for smashing into big red fire trucks, sometimes with flashing red lights, like Tesla autopilot does. I assume/hope the adults are implementing the real self-driving cars and Tesla is the only one that hired fresh outs from community college to do design and write their software and follow pedo-Musk's every whim.)
I always assumed that the CSRs could see every keystroke I type in the chatbox.
If I have a long question (often the first question is long for example), I usually type the text into emacs (and, if I'm being picky, spell check it as well) and then cut and paste into the chatbox just to avoid confusion (such as me leaving out a "not" and later correcting it and the CSR never noticing my correction).
When someone claims person Y's life was worth $Z, I inquire as to if they had >=$Z life insurance on Y. If not, they are lying about the value of Y's life to them.
In this case, the question is "Were these files worth backing up in case a house fire destroyed them?". If the answer is "No", then I know how little they are worth (given the low cost of offsite backups either via sneakernet or via the internet or other communication medium). If the answer is "Yes", then the damages (if any) would just be the cost of retrieving the backup and restoring the files.
Why should a third party care more about your files than you do -- esp. when you explicitly decided NOT to back up your files and the third party's destruction of them was the result of an unintended bug?
So, tell your boss that you expect to be paid an hourly rate (or at least comp time) for anything over eight hours on a "travel day".
Suppose a nurse looks at a patient's "chart" at 01:45 after the time switch and sees a medication was given at 01:30. Suppose the patient is allowed to, on request, have the medication not more often than once an hour (and, perhaps, there are additional limitations on the number of doses in a 24 hour period). If the patient asks for another dose at this time, does the nurse say "Sure" or "I'm sorry Mr Senile, you can't have another dose for another 45 minutes. You have probably forgotten, but you had a dose just 15 minutes ago by Nurse Jones who got called to a code blue a few minutes ago."?
We could of course, for example, on the west coast require "PDT" or "PST" to be included in all user input of times in ambiguity and displayed as well and train everyone for this once a year event (the change to "spring ahead" is less confusing and is not ambiguous even without the time zone designation differentiating from daylight and standard time).
I've read multiple news reports in just the past few years of people intoxicated on drugs or alcohol stepping out into busy traffic on freeways where pedestrians are not allowed at all, let alone in the roadway (note, the results are almost always the same but the good news is the addiction problem is fixed forever). I've never read (except perhaps on the National Enquirer cover or in The Onion) of your scenario happening.
Hmm... I wonder if you would like to be sentenced to prison. You would be surrounded by inmates who think outside the rules and aren't knee-jerk authoritarians which would thrill you. On the other hand, you would also be subject to a lot of knee-jerk authoritarians with badges which you might not like -- but that would give you even more opportunities than usual to show how much you think outside the rules. Tough call :)
In all of the cases in the survey I took at this site, SOME form of loss of life was unavoidable (although some were between humans and non-humans).
However, in many real world cases, there are likely to be differing probability of human death and, certainly, self-driving software should take that into account. A car hitting an elderly pedestrian squarely at 35 MPH is very likely to kill the pedestrian (maybe >80%?) but in a modern car with the highest crash protections, hitting a concrete barrier at the same speed is much less likely to kill the passenger (maybe 5%?).
In collisions with inanimate objects there are also differing risks of death or injury. If a semi-truck runs a red light as you approach the intersection and the car you're driving is about to be stripped of everything above the beltline - including the heads of you and all of your passengers - if you do nothing, I'm betting you would correctly swerve into a "soft target" like a trash can to avoid the truck. Both are "collisions" (a.k.a. "accidents"), but humans do make distinctions between collisions based on the imprecise odds of death and injury among all options. I expect self-driving cars to do the same.
These maneuvers are not necessarily "insane" - but there may be something in the "escape path" with some probability that it is each of (a) human, (b) domestic pet, (c) small wild rodent, (d) a patch of flowers in front of a business or home.
It sounds like you are proposing that if p(a) = 0.0000001 and p(d) = 99.999 you would still elect to kill the pedestrian? Most humans would not make that decision -- including humans on the jury for the civil case the pedestrian's widow and children file.
[Yes I know that it's quite possible that there is a small wild rodent or a small child hidden in the flowerbed so it's a little more complicated than I make it out to be.]
Should it swerve if that would result in no damage or injury to anyone if it detects that it will impact the pedestrian at a lethal speed if it just slows down?
Assuming that your answer is "yes" - what if it detects a domestic dog in the path of swerving, should it kill the pedestrian instead? What if it detects a squirrel in the path of swerving? What about a mouse? What about a mailbox? Remember, you're writing the code and/or developing the training sets -- your call, the only way out of making the call is to shut down the "self driving car" company you founded and invested all of your savings in.
This decision, of course, is no different than one you would be forced to make as a human driver in a similar situation -- but at least the jury in the human driver case can sympathize with the "split second" nature of it -- they won't be so sympathetic to the five year old block of code with the comment in front of it that says:
However, there are laws against some sorts of crossing between intersections in some areas. Where I live, there is one law that says that if you are within X feet (I don't recall what X is) of a controlled intersection, it is illegal to cross except at such an intersection.
Also, there are places where "no pedestrian" signs prohibit pedestrians so crossing a road from/to one of those areas is also illegal.
But, yes, where I live, you're certainly free to cross an undivided road on foot if there's no intersection within miles and no sign prohibiting it!
So, a self driving car in a case like this (or any of the millions of other similar cases) should never take evasive action - just smash into the pedestrians or barrier in front of it? Or, do you suggest that self-driving cars should ignore bicyclists and pedestrians (as they are both light and probably won't kill any passengers in the car) when evaluating evasive maneuvers necessary to "save the car and the passengers"?
Consider how you would write the requirements spec or testing criteria for self driving software. Would you just identify desired actions in such cases as "Undefined"? If not, what would you define.
That's naive -- the software has to do something here, it receives a bunch of inputs and analyzes a bunch of possible outcomes and somehow has to score them to decide to take an action (including "do nothing" which is, in itself, an action).
For example, surely a car should swerve to avoid a car that has run a red light if that will avoid a collision of any sort rather than just run into the red light runner and likely kill or seriously injure individuals in both cars. But, what if swerving would mean impacting a vehicle in the cross street (in another lane) that did follow the law and stop at the red light and this impact, due to being off center, presents a much smaller but still significant likelihood of death -- but to the completely uninvolved driver of the lawfully stopped car?
An alert driver will actually make a decision in such cases -- it may not be the right one and it may not be made quickly enough and most people probably don't know how they would really make such a split second decision. Our brains are not available for code review. Hopefully the self driving software/dataset and design has been reviewed and the reviewer has to determine if the code/dataset meets the requirements in this case - a SEGV probably isn't acceptable for example. If nothing else, fuzzing tests during development would result in situations like this and require a human (perhaps with a lawyer stitched to their side) to decide if the outcome was acceptable.
The survey presented assumes brake FAILURE in the scenarios. I.e., the brakes are applied but they fail to stop the car as expected.
In such a scenario, the software is presented with several inputs including at least:
1. In spite of applying the brakes the car is not slowing down quickly enough (perhaps it might not even be because of a defect in the car -- perhaps a leaking tanker truck full of vegetable oil had passed ahead of the vehicle you are in just 20 seconds earlier).
2. The car is approaching an unanticipated barrier (perhaps related to construction or perhaps as the result of some sort of localized natural or man-made disaster) at a high speed that, if impacted, will likely kill everyone in the car.
3. The software detects what it thinks is a "passenger safe" path around the barrier, but detects that a living entity (a dog or a person perhaps) is occupying that space and that the living entity will likely be killed if that path is taken (of course, if it's the vegetable oil case, the software might not know enough - do the weird road conditions also exist on the swerving path which might make steering ineffective?).
The software has to do SOMETHING predictable in this case - even if it is just to throw up its hands and reboot! Any unpredictable response to fixed inputs in this area should be caught in review before the product alpha. So, what is the requirement in this scenario?
So, if a pedestrian high on drugs illegally steps out in front of your car in the middle of a high speed road where there is no pedestrian crossing, you would choose to drive your car into a barrier to avoid the pedestrian even if that would result in the almost certain death of yourself and three other family members in your car? What if you were driving a carpool of kids and two of the passengers who would die aren't even your kids -- they are neighbor's kids?
The US is a nation composed largely of immigrants and their offspring, many who have arrived comparatively recently. In many cases they came not because it was convenient (getting to the US from Poland or Italy, for example, was not "convenient" before air travel - esp. for poor people) or because it was easy or because it was low risk. They subjected themselves to substantial risk, expense, and inconvenience to make the trip and survive in the US.
These immigrants, of course, left behind those that didn't have the same drive or interest in creating a better situation for themselves and their families. It would not be surprising that those who had the gumption to better themselves rather than sacrifice themselves for the "common good" would be looking out for themselves and their families more strongly than those that lacked such gumption and remained behind.
As well, the US has historically been one of the most diverse populations in the world (due to the source of our population) so the tribal "common good" notion is probably unsurprisingly much stronger than in monocultures like Japan or most of the Nordic countries.
The US seems to have done pretty well - esp. in light of having to deal with its very diverse population.
In the scenarios in the linked survey, there are ones where the pedestrians have a green light and ones where they are walking against a red light. I always prioritized legal pedestrians over passengers and passengers over illegal pedestrians. Note that in every case that which caused the situation was a brake failure - not something that the passengers are directly responsible for at that moment (unless they are knowingly using unlicensed/uninspected/unmaintained vehicles and that's why the brakes failed - but the scenarios don't suggest that's the case).
Legislation might require that the passengers in the car be prioritized below law abiding pedestrians in failure cases. This would encourage people to buy/rent/share the most reliable cars -- i.e., those that don't have as many failure cases that require making such decisions. It also makes the person responsible for the selection (the passengers) accountable for their actions.
Apple users obviously don't object to proprietary walled gardens else they wouldn't be buying Apple products. This is just a few more bricks on top of the garden wall and I would expect it to be celebrated.