Getting it wrong on self-driving vehicles (crosspost from Crooked Timber)

A few years ago, I got enthusiastic about the prospects for self-driving vehicles, and wrote a couple of posts on the topic. It’s now clear that this was massively premature, as many of the commenters on my post argued. So, I thought it would be good to consider where and why I went wrong on this relatively unimportant issue, in the hopes of improving my thinking more generally.

The first thing I got wrong was overcorrecting on an argument I’d made for a long time, about the difference between radical progress in information and communications technology and stagnation in transport technology. The initial successful trials of self-driving vehicles in desert locations led me to think that ICT had finally come to transport, when in fact only the easiest part of the problem was solved.

There was also an element of wishful thinking. As commenter Hidari observed, the most obvious use of self-driving vehicles is to provide mobility for 75+ Baby Boomers. As someone approaching that category, and having never liked driving much, this is an appealing prospect for me. And I liked the idea of taking other bad drivers’ hands off the steering wheel.

That framing of the issue is very different from the way a lot of commenters saw it. Should self-driving cars be seen as automated taxis, and if so is automation desirable or not? Is any improvement in car technology a distraction from the need to shift away from cars altogether? I don’t have good answers to these questions, but they indicate that resistance to self-driving cars won’t be purely a matter of technological judgement.

Finally, having put forward a position, I am usually tenacious in defending it. Within limits, that’s a good thing, particularly in the context of a blog where the discussion doesn’t have any direct implications for what happens in the world. It’s good to put up the strongest case, and test it against all counter-arguments. But that approach carries the risk of being obstinately wrong.

I’m hoping discussion here will help me deal with more consequential errors of judgement I’ve made. So feel free either to discuss the original question of self-driving vehicles or the broader issue of how to think about mistakes, and particularly mistakes I’ve made.

17 thoughts on “Getting it wrong on self-driving vehicles (crosspost from Crooked Timber)

  1. if you haven’t done much programming (ie years on a paid product) you can’t really appreciate computer programming’s limitations. that said, nor can most programmers.

    wilkes at Cambridge computing lab had an epiphany when he realised he was going to be spending most of his time correcting errors in his programs. he thought having made edsac his problems were over. and edsac wasn’t doing an awful lot more than printing out trigonometrical tables.

    keep up the good work!

  2. Is the mistake here of any consequence? Self-driving vehicles are taking longer than many expected, but they are surely coming. There is SFIK no computability barrier, it’s just more data needed and more processing. Moore’s Law no longer holds, but computer processors and memories still get better every year, as do cameras and LIDAR sensors. Self-driving fans did underestimate the non-technical issues of public and regulatory acceptance, but several metros have run driverless trains for decades without any signs of social rejection. Private cars will be last, after buses, vans and trucks on short fixed routes.

  3. As well as the computing aspects, there are moral/ethical/legal issues to be sorted out before self-driving cars become mainstream. Does/should a self-driving car put the passenger’s life at risk by swerving when a child runs out onto the road? What about a cat?

    A quote attributed to Bill Gates is “We always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten.” Do you think this may apply to your views on self-driving cars?

    But don’t be too hard on yourself. You are correct far more often than most pundits.

  4. Does/should a self-driving car put the passenger’s life at risk by swerving when a child runs out onto the road? Yes, it should do that but according to calculations and heuristics (which is what humans do). It would not be programmed to swerve a car with 2 adults and 2 children in it, doing 100 kph, into the path of an oncoming truck doing 100 kph also, just to avoid one child. It would be programmed to seek the verge or other clear path, thus braking and swerving (not simultaneously) at optimal rates (and probably better than all but racing car drivers could achieve).

    What about a cat? No, it should run over the cat.

  5. What would I read to find out where we are now with self driving cars?

    I feel the momentum is towards a new local minimum with electric / self driving cars, while missing the opportunity to design a better transport system.

    I would like a 200km/hr urban roller coaster network.

  6. Self driving cars will become a thing when the insurance wonks work out who has the least accidents. Then the cost of insurance will diverge —and pretty widely too— and then the market will reply.

  7. “if there is a group in the United States that stands to benefit most from the life-saving potential of self-driving technology, it’s those who live in the greatest poverty, but only if they can afford the technology. Driverless-car technology might have the potential to improve public health and save lives, but if those who most need it don’t have access, whose lives would we actually be saving?”

  8. I fear JQ has been drawn into the gravitational hype field surrounding everything Tesla does. Musk is drunk on his own celebrity. The outrageously dishonest labelling of Tesla’s current software as “FSD” = Full Self Driving is just a striking example. Tesla makes real progress, sure, bu it’s hard to get a dispassionate view. I suggest looking instead to other major players that behave like normal businesses.

    Waymo is Google’s self-driving subsidiary, with bottomless funding. Its approach is cautious, conservative and safety-conscious. Waymo has been operating fully autonomous robotaxis in Phoenix since November 2019, and has started in San Francisco (Wikipedia). As I understand it, their system relies on extremely detailed maps, down to individual lampposts, so rollout is inevitably slow. But it works, expensively and safely.

    NVidia is the leader in electronics for autonomous vehicles. I’m against their proposed takeover of ARM and hope regulators kill it, but NVIdia is undoubtedly a very capable outfit. Car computers have to be very powerful as well as reliable and secure – it’s too risky to rely on the cloud for second-by-second operation, though that´s where the AI algorithms and map databases are updated. NVidia expect to sell processor SoCs capable of Level 5 self-driving in another couple of design iterations, roughly annual. NVidia is a pure electronics company and will sell to anybody. Tesla designs its own, like Apple.

    I see no reason to think Waymo, NVidia and their competitors will not get there by 2925.

  9. “Now it’s been ten thousand years,
    Man has cried a billion tears
    For what, he never knew, now man’s reign is through.
    But through eternal night, the twinkling of starlight
    So very far away, maybe it’s only yesterday.” – Richard Lee Evans.

  10. “Within limits, that’s a good thing, particularly in the context of a blog where the discussion doesn’t have any direct implications for what happens in the world. It’s good to put up the strongest case, and test it against all counter-arguments.”

    I am unsure I agree with that as a general claim. If you advance any proposition shouldn’t part of the consideration be “what could possible happen that is very harmful if I get it wrong. Minimizing regret and avoiding hubris seems to me a wise choice. Applied in the current context I don’t think there are drastic consequences of getting it wrong by overstating the case for electric vehicles so a tick. If the 1% chance that climate change arguments are wrong eventuates then no massive damage in getting rid of polluting fuels. On the other hand suggesting “cold shower” privatizations in the former Soviet Union had terrible consequences for welfare when that turned out to be the wrong argument so no tick.

    Considering the regrets that might arise if your argument is wrong seems an important ethical principle in economics and other public policy discussions.

    On a blog if you are arguing a point I think the same caution should apply as in giving testimony to a Senate committee. Indeed the former might be a rehearsal for the latter.

  11. I agree that your mistake was not very consequential, but I really disagree that the issue of the arrival date of self-driving cars is unimportant. I think they will utterly, utterly transform our lives and our cities. Once they are entrenched and widespread, every street in the city will be one way (except the cul de sacs [culs de sac?]); the huge amount of space given over to parking will suddenly be freed for other uses; children, older people, the disabled and the inebriated will be more mobile than drivers currently are; property prices 50-200km from the city will soar as people effortlessly commute.

    They will be able to form trains on long-haul routes, speeding along at 200km/hr plus with the lead vehicle bearing the brunt of the wind resistance and cars peeling off and glomming on as required. They’ll be able reliably communicate with each other, so two cars can speed through an intersection missing collision by mere centimetres, or follow the car in front almost touching. No reaction times, no mystery about how others will behave, no aggression or selfishness.

    I think your mistake might have been the failure to clearly map out a pathway to widespread adoption. My guess is that a self-driving-only expressway between two cities will be the clincher. Imagine if the koala-killing highway from Brisbane to the Gold Coast had a speed limit of 200km/h and a policy of self-driving cars only. It would reduce the trip to the beach by half. I’m sure there are larger cities with a similar possibility.

  12. seqaugur, reminding us of Snow Crash, “glomming” in my lozenge shaped ‘transport’.

    And if we “speed through an intersection missing collision by mere centimetres” I will definitely need a pair of;
    “Peril Sensitive Sunglasses – Joo Janta 200 Super-Chromatic Peril Sensitive Sunglasses have been specially designed to help people develop a relaxed attitude to danger. At the first hint of trouble, they turn totally black and thus prevent you from seeing anything that might alarm you.

    “A double-pair is frequently worn by Zaphod Beeblebrox.”, and me at 200kmh thru an intersection!

    SF becomes reality.

    “7 science fiction inventions that became reality
    “Sci-fi authors wanted to prevent dystopian futures. Instead they predicted them.”

    And the robot guard dogs in Snow Crash too.

  13. The first date of interest (n) is when fully autonomous vehicles are approved for a whole country – Waymo are already there for a single city. I already guessed 2025 for the technology, add 3 years for the red tape and mass production, so 2028. The second date is the ban on human drivers on public roads, since they can’t be trusted. I’d guess n +15 years, a reasonable life for a car. This takes us (well. some of you, I won’t be around) to 2043.

    Towards the end of the second period, there will be a lot of social and political pressure to bring the ban forward, as nearly every road death will be avoidable. “Toddler, 3, killed by manual truck. Third road death from killer drivers in Qld in a week!””

  14. Joe: “if there is a group in the United States that stands to benefit most from the life-saving potential of self-driving technology, it’s those who live in the greatest poverty..” The poor are more likely to be pedestrians and cyclists, both of whom will benefit greatly from cars and trucks driven safely by undistracted and sober AI. We can be quite sure that the AI processing kit will get cheaper, and pretty confident that the same will hold for the sensors (LIDAR and webcams). Cf. $60 Chinese smartphones in South Africa. A program may not leave any change from $1bn for development, but the marginal cost of making an additional copy of it is $0.

  15. JQ – ” It’s good to put up the strongest case, and test it against all counter-arguments. But that approach carries the risk of being obstinately wrong.”

    The game is not to have a case win, but to aid discovery – somewhat like the difference between the doubly lob sided voodoo English ‘justice’ system and the Code Napoleon influenced systems.

    The greatest obstacle to discovery is not ignorance – it is the illusion of knowledge

    – Daniel J. Boorstin

    The Knowledge Illusion: The myth of individual thought and the power of collective wisdom.
    By Steven Sloman, Philip Fernbach – 2017

    “We all know less than we think we do, including how much we know about how much we know. There’s no cure for this condition, but there is a treatment: this fascinating book. The Knowledge Illusion is filled with insights on how we should deal with our individual ignorance and collective wisdom.”
    —Steven Pinker, Johnstone Family Professor of Psychology, Harvard University, and author of How the Mind Works and The Stuff of Thought

    The knowledge illusion, Two extracts from a new book by Steven Sloman and Philip Fernbach
    The knowledge illusion is the flip side of what economists call the curse of knowledge… The curse of knowledge is that we tend to think what is in our heads is in the heads of others. In the knowledge illusion, we tend to think what is in others’ heads is in our heads. In both cases, we fail to discern who knows what.

    The Knowledge Illusion – Philip Fernbach and Steven Sloman on ignorance and irrationality
    … Evolution’s endowment
    The human mind evolved abilities to take actions to better enable our survival. Remembering vast amounts of information was not helpful to our evolution. As our brains developed in complexity (and shrunk), we got better at responding to more abstract environmental cues and new situations. Furthermore, unnecessary details are counterproductive to effective and efficient action if a broad understanding suffices. This is why our attention and memory systems are sometimes limited and fallible. We may have evolved a different set of logic if it was conducive to our evolution and survival as a species.
    First we overestimate, next we ignore
    Humans use causal logic to reach conclusions and project into the future. We share similarities with computers in that we undertake complex tasks by combining simpler sets of skills and knowledge together. However, the illusion of explanatory depth means we often overestimate the extent and quality of our causal reasoning.
    We acquire information slowly and spend only a small fraction of our lives deliberating. Doing otherwise would leave us unable to handle systems that are too complex, chaotic or fractal in nature. The illusion of understanding means we tend manage complexity by ignoring it. Likewise, we ignore alternative explanations when we’re unable to retrofit it into our pre-existing model of understanding.

  16. To the extent that your enthusiasm for self-driving cars is based on a forecast as to when they will become reality: what do you expect ?
    How much forecasting is 100 per cent accurate ? And if it is, how much of that is sheer luck?
    Treasury forecasts things all the time and gets it wrong with monotonous regularity. They never look back at how wrong they were anytime.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s