Friday Reads – 23 March 2018

Welcome to Reconnections’ Friday Reads. This week’s lineup:

Check out our most popular articles:

And some of our other sections:

If you have something you feel we should read or include in a future list, please email [email protected].

41 comments

  1. Interesting to see the Toronto streetcar tunnel debacle here. It’s not clear from the photo, but for a car to reach the passenger platforms, it has to drive about 200 yards along the streetcar tracks from the tunnel entrance. In the most recent case (not the one in the picture) the police found the driver still at the wheel and with his foot still hard down on the gas pedal, even though the wheels had no traction!

  2. I have lost count of the number of people that are saying that, as a result of the fatality involving an Uber driverless car in Arizona, the whole concept should be scrapped. Not only have these comments come without any analysis of the fatality from the relevant authorities, the argument that one fatality should lead to a whole technology being blocked is quite absurd. If applied to conventional cars they would also be banned, as, of course, would be guns (the NRA would go ballistic).

    Of course the fatality is very unfortunate and thoughts are with the family of the unfortunate victim. But there does need to be a proper analysis of the cause of the fatality and an assessment as to what measures can be put in place to prevent a repetition. Only if it is impossible to make any improvements to what could turn out to be a system failure should we really think this is fatal to the technology.

    It’s also worth remember ing that the first pedestrian fatalities from conventional cars took place in the 19th century – and still take place with alarming frequency today.

  3. The self driving car article: having seen the video footage, the pedestrian only walks into the beam of the car’s headlights at the very last moment, so one wonders whether she would have been seen in time even by a human driver.

  4. Re Timbeau and Quinlet,

    The CEO of the firm that supplies the LIDAR equipment to Uber and others has said that its equipment would have seen the pedestrian well before they were visible in the headlights and have made assertions about the quality of Ubers software that interprets the data from the LIDAR and makes the decisions. The CEO also noted that the NTSB hadn’t come to ask them about the LIDAR…

  5. @quinlet

    I think the key point being made in this linked article is that autonomous vehicles are not yet ready to be tested in public – the necessary safeguards for others’ safety are not ready yet, nor are the requirements for the ride-along driver (ie no electronic devices whilst riding in the AV, attention focussed solely on the road etc). Surely multiple sensors should be mandatory on AVs – visual camera, infrared camera and/or radar, for redundant collision detection.

  6. @quinlet
    “…the argument that one fatality should lead to a whole technology being blocked is quite absurd.”
    On reading this, my first thought was the incident that happed on day 1 of the Liverpool & Manchester Railway…

  7. Of course, I meant to say “happened”. That’s what I get for commenting after midnight…

  8. @LBM
    I don’t think it is possible to draw that conclusion without understanding exactly what happened and what went wrong. To draw an analogy, on the same logic you should also argue that the Croydon tram accident meant that trams were not safe enough to be tested in public service.

  9. There’s a lot of internal documentation coming out that shows Uber’s self driving system is far less resilient than competitors like Waymo. Uber was struggling to achieve 13 miles driven between human interventions whereas Waymo averaged 5600 miles between. That’s demonstrates a massive gap in platform capabilities and it would not be surprising to find the car’s LIDAR provided the data correctly showing a pedestrian obstacle crossing perpendicularly to the vehicle’s path but Uber’s software failed to respond appropriately. That’s not a problem with AVs as a concept but a very specific at-fault vendor implementation.

  10. The pedestrian in the Arizona fataoity was WHEELING A BICYCLE – a large metallic object that should show up in LIDAR/RADAR much better than visually – so there is (IMHO) absolutly no excuse for the software-powered detection sytems not to “see” it.

    Incidentally, there is a very amusing & fairly advanced discussion of this precise problem, over on Charlie’s Blog ( C. Stross, SF writer & commentator )
    Link HERE … Please take the time to read the original author’s high;ly amusing but pertinet comments & the (long) discussion.

  11. Mike Watt
    Your comment seems to imply that Uber are “doing it on the cheap” without following the rules & procedures that others are following.
    Oh, shock, horror, what a suprise …….

  12. Greg Tingey
    I’m genuinely shocked. There’s been nothing prior that could possibly have pointed to this being their modus operandi.

  13. Someone said: “Move fast & break things” as the underlying ethos of (parts of) “Silcon Valley”
    Certainly it used to appear that Kalanick represented that ethos, I’m afraid.
    Mayb OK, mayb not, but, when the things you break are other people’s lives ….
    { And, yes, I did notice the /Fe* content of your post ]

  14. @Greg – ah, but that the bicycle made her an odd shape and thus the Uber software got confused as is obviously is’t clear what to do with ‘big object in the road that you might crash into’.

    Plus she was jaywalking, so clearly that adds to the fact that Uber couldn’t possibly have predicted it as how can it compute people breaking the law?

    Or at least this is what those saying there was nothing wrong with the tech are saying.

    It makes it sound worse though – that rather than a failure of the tech, that it’s a failure in the specifications of the tech: that the unexpected/unusual/unlawful wasn’t to be processed, that collisions were only to be avoided if the software could work out what was going to be hit – and perhaps even (the subtext of some more extreme comments I’ve seen) pedestrians jaywalking are fair game to run over.

  15. It might be worth pointing out that none of the “extreme comments” derided by Si in a recent message were made on this forum – and none such would be allowed to remain if they had been made here.

    Comment on the event, or perhaps on what may (plausibly) have gone wrong, or what should be done about it. But second-level comments on other people’s daft comments elsewhere are less welcome.

  16. A perhaps interesting comparison to look at in order to consider the path that automation of driving will take is the aviation industry.

    It is now possible for an aircraft to do almost every part of a flight autonomously. This has no doubt saved thousands, if not millions of lives due to reduction in accidents due to pilot error and/or by relieving pilots of certain duties to allow them to look out for and correct issues earlier than might otherwise have been possible.

    The flipside however, is that there have been several incidents of autopilot systems causing or at least contributing to accidents – for example AF447

    Equally (aside from AF447) there are several examples of the problems encountered by having an autonomous system working together with humans (Überlingen mid-air colision ).

    As is the case with aviation, this accident perhaps serves as an indicator of what is to come in the development of autonomous cars – that rules and regulations are introduced as a result of (deadly) accidents to prevent them happening again, with the end result being far safer than they system we have now, but at a cost of numerous lives.

    Whether effective regulations can be introduced preemptively to avoid this human cost is very much up for debate, but history suggests that it is unlikely, and that those lives are a price we as a society will have to pay for the development of this technology.

  17. It seems to me that, at the very least, we would like Uber and other advocates of driverless cars to perfect their software somewhere else from anywhere I am likely to be.

  18. About where and how driverless car tests should take place, the standard approach for drug testing might be a good guide. Test as far as you can go without risking humans at all, when that hurdle is past, test using only consenting volunteer humans, and only when such tests are complete, let the product loose on the general public.

  19. @Malcolm

    That could well be the best approach. However the drug test approach is also notorious for being extremely expensive and a bureaucratic nightmare. Tech companies will fight tooth and nail to avoid such a regime being imposed on them. This is partially what I meant when I said ‘history suggests that is unlikely’.

  20. More of an issue than the crash is the worrying lack of regulatory response.

    The autonomous vehicle was reported to be speeding. It was also reported not too have attempted to brake at any point.

    Both suggest that Uber has encoded the behaviour of current drivers, rather than aspiring to making the system drive “perfectly”. That means we will have to put up with a lot more bad behaviour on the roads.

    And – rather than investigating this – the local PD has been complicit in excusing the behaviour by putting the blame onto the victim.

  21. @ Malcolm 17:17

    You are quite right re:extreme comments. It just really irked me having seen comments elsewhere along the line of ‘that she was jaywalking needs to be in the headline’, like it was the most important to know that this person was crossing the road not at an official crosswalk and thus it was her fault (and the local PD is pushing this line too, according to Bob – yuck!).

  22. The “The Perfect Selfishness of mapping apps” article is rubbish. It just goes on and on with a lenghty text that just states the obvious that there has been an increas of traffic on local streets and also that we don’t really know if map app usage is bad or good or even neutral.

  23. The jaywalking issue is surely a red herring. The cause of the accident was the car’s failure to detect an obstruction in the road.

  24. Timbeau
    Sorry, but the issue of “jaywalking” – a peculiar US corrupt perversion is unfortunately, directly relevant, because it specifically allows careless/useless drivers & automated apps ( like Uber ) to get off the hook., by blaming the victims

  25. Timbeau. I agree with your assessment, but you and I have a European mindset. Many US states operate differently, they put the car driver at the top of the tree and consider Jaywalking an offence.
    I struggle to see how they can programme a car to cope with the near anarchy of a street in soho or naples.

  26. @Island Dweller

    Europe, USA or the remotest parts of the Silk Road: if a vehicle can’t deal with obstructions in the road ahead (be they “jaywalking” pedestrians, stray animals, stalled cars, something fallen off the back of a lorry, fallen trees or whatever), it is sooner or later going to have an accident.

  27. To say that anti-jaywalking laws are “blaming the victims” is not always correct. Careless walking can indeed cause accidents, but the victim is quite often a different person – as when a car swerving to avoid a jaywalker hits something or someone else.

    The difference between the situation in the USA and the situation elsewhere is also not quite as great as it sometimes seems. Societies everywhere dislike accidents caused by pedestrians, and seek to minimise them, sometimes by punishing offenders and sometimes in other ways. The word may be largely limited to the USA, but the actions, and also public disapproval of them, are more widespread.

  28. @ Malcolm – it is really true to say “societies everywhere dislike accidents caused by pedestrians”? That rather puts the onus on one half of the parties involved in collisions. Surely there is as much or more “dislike” of the damage caused by motorised vehicles driven by *drivers*? The US situation, as stated above, puts the motorist at the top of the tree and seeks to move responsibility elsewhere. This strikes me as bizarre but we see the same ridiculous mindset here with motorists expecting priority over everything and everybody – no doubt an imported mindset as we followed US thinking about roads and building our way of jams from the 50s onwards.

    The worrying thing about autonomous vehicle development seemingly being led from the US is that their “car on top” thinking removes a whole pile of legislative pressure and control from the companies doing the developing. How soon before we see the inevitable “pedestrians and other road users must be removed from roads and put behind crash barriers to allow autonomous vehicles to work properly” demands? How on earth a company can develop something so poor that it cannot cope with reasonably foreseeable circumanstances is beyond me.

    At some point soon there is going to be an almightly collision (pardon the pun) between those lobbying for car reduction, carbon reduction and “calmer streets” and those in very large businesses promoting autonomous vehicles who will want vehicular priority and “safer streets” which restrict what the other lobby want. I can foresee London being one of those places where that “collision” occurs given the direction set my the new Mayor’s Transport Strategy.

  29. WW: There is a lot in what you say.

    I stand by my claim that all societies dislike accidents caused by pedestrians. I did not make any associated claim as to a comparison between this particular disliking and parallel ones about other causes of accidents. There may well be different tendencies to exonerate particular categories of accident-causer to different extents in different societies. But it is not a hard and fast case of USA -> motorists are all blameless vs elsewhere -> pedestrians are all blameless.

    Where I do observe a contrast is in the attitude to behaviour (on anyone’s part) which looks safe-ish to a common-sense observer, but is against some laid down rule. Things like crossing the road against a red pedestrian-light in broad daylight when there is no traffic around anywhere near. This is widely disapproved of, and sometimes legally punished, in certain places (e.g. Germany, Scandinavia, some parts of the US), and widely tolerated in many other places.

  30. I suppose that a good example of the differing priorities between cars and pedestrians is shown by the recently produced Chrysler Commarro which is on sale in the US but forbidden in Europe because the design of its front means it is likely to cut the legs off any pedestrian it hits at speed.

    Having said that, it is not just a case of the motor industry seeking always to put the car in front of the pedestrian. Remember the Ford Pinto where the designers accepted that the design of the fuel tank meant a limited number would explode, but that the costs of damages from those explosions would be less than the cost of a redesign.

    Those examples of what the auto industry will do, even now, to disregard safety must be seen in the context of the somewhat hysterical calls, in some quarters, for autonomous vehicles not to be allowed at all as a result of one – as yet not fully explained – fatality in Arizona.

    Having said that, I think WW is right to say that there will be some calls for safe corridors for autonomous vehicles in cities, unless we work much harder now to decide how autonomous cars should fit into the city (rather than the other way round). This would be a much more useful line of consideration than pretending that autonomous cars won’t happen.

  31. Apportioning blame is all very well – would liability have been different if the victim had (presumably legally) been riding the bike? Or if it had been another car?

    But whether an obstruction is legally there or not, it is better (for the safety of the car’s occupants, if no-one else), that it is detected and avoided, and it seems that the automatic car in this case failed to do so.

  32. @MALCOLM

    ” Things like crossing the road against a red pedestrian-light in broad daylight when there is no traffic around anywhere near. ”

    It was always my understanding in the Common Law that lights to pedestrians are always “advice” and not commands. The red lights to a car driver, who has taken and passed a test to show they know what is expected of them when they drive (or ride a motorbike).

    Because a member of the public cannot be expected to have read “The Highway Code”, licensed motor vehicle drivers are required to look out for pedestrians and give them right of way on non-motorway roads. Drivers sign up to, for example, “watch out for pedestrians crossing a road into which you are turning. If they have started to cross they have priority, so give way.” (160).

    The only form of “Jaywalking” in the UK is on Motorways, as it is specifically banned is section 17 of the Road Traffic Regulation Act 1984 and regulation 15 of the Motorways Traffic (England and Wales) Regulations 1982. I was once told that this was Motorways restriction didn’t conflict with a person’s rights are they were not an existing right of way.

  33. An update on the Arizona crash – Intel who supply some of the optical sensors and some software (also see previous comments from the LIDAR supplier) say that their systems detected the pedestrian when shown a degraded copy of the video of the incident on a monitor infront of the sensors but that Uber had turned off that part of the intel software on the car involved in the incident…

    The Arizona Governor has now also said Uber doesn’t have a safety first approach.

  34. @Briantist
    “I was once told that this was Motorways restriction didn’t conflict with a person’s rights are they were not an existing right of way.”

    Which is also why if a road is “upgraded” to motorway, as much of the A1 has, an all-purpose road has to be provided parallel to it.

  35. @timbeau – I can think of a good number of stretches of the A1(M) which don’t have an all purpose road beside it. It’s quite easy to extinguish rights of way actually – happens all the time.

  36. Graham/Timbeau: It remains true that when an all-purpose road is converted to a motorway in exactly the same place provision for access to any adjacent properties is necessary. Exactly how this is done varies from place to place, but in at least some places (e.g. A1(M) southwards from Peterborough), this is achieved by building a new all-purpose road on one or both sides. (These roads also fulfil other local needs). In other cases various other expedients may be used (including no extra provision at all if all adjacent fields already had alternative access). It may be legal to render someone’s house or field utterly inaccessible (by signing the right documents), but it does not normally happen.

    And there are also other parts of the A1(M) which are new-build on a different course, and the old all-purpose road has just been renumbered and retained.

  37. @Malcolm/Graham

    I did say “much of” the A1. The sections from Alconbury to Stilton in Cambridgeshire, and the much longer stretch from Bramham Crossroads to Scotch Corner are the most obvious. This is not the old A1, which was dual carriageway and whose alignment has been taken by the motorway, but a new road built alongside

    See B1043, A168, A6055,
    https://www.google.co.uk/maps/@54.2972586,-1.5607352,3a,75y,344.99h,92.52t/data=!3m6!1e1!3m4!1s_-eViWixE8gBT-9W8Luq5A!2e0!7i13312!8i6656

    or the A702 in Scotland
    https://www.google.co.uk/maps/@55.4776308,-3.695648,3a,75y,137.98h,93.54t/data=!3m6!1e1!3m4!1s1rrdbSHQzkkFOzT3KzpJvw!2e0!7i13312!8i6656

    These must be some of the quietest classified roads in the country!

    There are of course also many parts where a completely new alignment was necessary because it was not practical to widen the roads through the middle of places like Stevenage and Doncaster to motorway standard.

  38. Briantist writes: “The red lights to a car driver, who has taken and passed a test [are commands]”

    Um…if that is correct, then cyclists could not be prosecuted for ignoring a red light since they haven’t taken a test and don’t require a licence (and moreover, can be children).

  39. But the regulations and laws refer to ‘vehicles’ which include cycles. The testing and licensing of drivers is irrelevant .

  40. @Quinlet
    Which leads to the odd situation at “Toucan” crossings that cyclists can’t cross the road on the red light, but it is only advisory for pedestrians.

Comments are closed.