The Virtues of Impossibility

Jane McGonigal — the noted game designer and futurist — has a TEDx video about imagining the future.  I’ll be speaking at this topic at CNU on May 4, so having loved her book on gaming, Reality is Broken, I was curious about her take.

You can watch it here, or below, but I also summarize it below.

McGonigal’s project here is to undermine that pervasive feeling that we are unable to shape the future.  “The future is dark,” she says, not in the sense of bad but in the sense of hard to see into.  It is precisely our ignorance of the future that gives us the space to imagine it, and thus to change it.

She tells the story of a futurism class she teaches at Stanford, where she asks students to name things that they are absolutely sure will be true in ten years.  Then, she takes this list to researchers and asks how sure they are of those things.  Sure enough, there are cases where something that people thought was permanent is in fact changing.

I am totally with McGonigal on the urgency of helping people imagine a more diverse range of futures, but we must also notice how easily this message gets out of control, and how much of the craziness — and destructiveness — of the tech industry arises from this belief in the radically open future.

If I were in her futurism class I could easily name things that will absolutely be true 10 years from now.

  • The circumference of a circle divided by the diameter of the circle will still be a little over three.
  • Adult elephants still will not fit into wine-glasses.
  • There will still be a mathematical limit to how many vehicles — each filling the width of a traffic lane — can fit down a single lane in a state of free flow.
  • In the absence of barriers to walking, the density of population will still determine the number of people within a fixed, short walking distance of any point.  That in turn determine the potential market size for fixed transit, which will govern the level of transit intensity and infrastructure that will pencil out in ridership terms.
  • Carrying fewer people in more vehicles will consume more urban space, compared to carrying more people in fewer vehicles.  This mismatch between demand for road space and its supply will manifest as either traffic congestion or a need to widen roads, removing scarce urban land from other uses.

Those are geometric facts.  Within the scale of space and time in which humans live, no invention can change these things.

McGonigal’s rhetoric echoes that of the corporate motivation industry, with its endless implication that nothing is impossible.  But as with heated political rhetoric, the danger is always that some people take this stuff literally, and literally, it’s nonsense.  The facts about the future I listed above are extremely useful, despite not being ridiculous.  They are useful because they keep us from imagining, and spending billions on, truly impossible things.

If we know one thing from watching the tech industry, it’s that the wide-open “anything can happen” futurism that McGonigal praises can be its own bubble, with its own blinders.  Much good has happened from Silicon Valley’s willingness to throw venture capital at the occasional wild idea — “moonshots,” they’re called in the biz — but a lot of VC also goes to investing in mathematically impossible things — like moving people from big transit vehicles into small ones without taking up more space — and into attacking folks like me who try to point that out.

In this new era of “disruption,’ urban transportation suffers as much from delusional imagination as from lack of imagination.  Only when we allow boring math to constrain our imaginings — and when venture capital gets smarter at recognizing those constraints — can we start to have reality-based conversations about how to liberate people to move in our cities.

Not everything is possible, and it’s a good thing, too.

 

21 Responses to The Virtues of Impossibility

  1. Ben April 26, 2017 at 12:12 pm #

    Transit related things that will NOT be true in 10 years:

    — NO flying cars. Probably not even in the next 50 years. Even if the technology makes it possible, society at large will not approve it. All we need is people crashing into one another in the sky and flying over our houses. Not gonna happen. Helicopters are already annoying as it is.

    — Over 50% of cars on the road will still be regular human-driven cars. Thus many of the supposed efficiency benefits of autonomous vehicles will not come to pass. You’ll still be stuck in traffic with everyone else, even if the car is driving itself.

    And like my German friends told me recently:

    “Yeah, there’s a big development here of battery-powered self-driving semi trucks. They are currently being tested on the autobahn. And the companies promoting the trucks, in conjunction with the government, are thinking about creating special lanes just for these vehicles because it will make them safer and more economical to be separated from the other traffic. And they are also considering putting in overhead electrical lines that the trucks can tap into so they don’t have to stop to recharge the batteries. It’s stupid. Because everyone looks at this and says, hey idiots… IT ALREADY EXISTS. IT’S CALLED A FUCKING TRAIN!!!”

    I think the same logic can be applied to the self-driving cars that Google and Uber are developing. Complete rubbish. Why we aren’t investing in more trains, I don’t know.

    • Alan April 26, 2017 at 12:41 pm #

      As we’ve discussed from time to time, most U.S. cities don’t have much in the way of transit usage *today*. Sure, there’s about a dozen metros that really rely on transit to move a large proportion of their commuters – the rest do not. And I suspect that the overwhelming majority of existing *cars* are in areas where transit usage is currently very light.

      So those VC dollars aren’t being invested in mathematically impossible things – they’re being invested into adding a *lot* more functionality into existing private automobiles and trucks that *already aren’t using transit*. If they’re successful at bringing that functionality in at an economically feasible cost, it could *decimate* transit systems outside of those dozen or so metros – and probably destroy the “coverage” routes in those networks as well.

      They don’t need to address the mathematically impossible task of replacing the Lexington Avenue subway line in Manhattan. Their market is the 80-90% of cars that aren’t anywhere near a line of that capacity. And if they end up replacing transit systems in *most* U.S. cities, that will just be a little extra total market.

      P.S. – last I checked, train cars can’t decouple from the larger train and drive individually to their end destination on available surface roads – which is why there’s a lot of interest in autonomous semis, and not investing more in trains.

      • George Lane April 26, 2017 at 6:38 pm #

        “P.S. – last I checked, train cars can’t decouple from the larger train and drive individually to their end destination on available surface roads – which is why there’s a lot of interest in autonomous semis, and not investing more in trains.”

        Boats can’t decouple from each other and continue to a final destination not on the sea/a river/a canal either, yet the enormous majority of freight kilometres are done on sea. Automation will apply equally to logistics centres. Why not have a train pull up, unload automatically, reload automatically, and leave while trucks are automatically loaded with the stuff that was unloaded?

        “If they’re successful at bringing that functionality in at an economically feasible cost, it could *decimate* transit systems outside of those dozen or so metros – and probably destroy the “coverage” routes in those networks as well”

        There is one big assumption here: economically feasible. It will always be cheaper per person to drive a bus with 100 people from A to B than to drive 100 taxis from A to B. If you can pay $20 to get to work in a taxi, or $5 to the station, and $5 on a bus (in the same total time) then most people would take the bus. Then you need to add that all of the autonomous taxis will still be sat in traffic so the bus will be faster.

        Your argument is common, but it always conveniently forgets the geometric cost of road space.

        • Alan April 27, 2017 at 10:56 am #

          WRT trains: yes, ocean/river freight also exists (in large capacity). And we still have trains – because ocean/river freight does things that trains don’t (and that truck freight doesn’t, either). But that’s exactly the “It’s called a train” retort doesn’t hold water – because even modest differences in functionality can support a different transportation system. The difference with autonomous trucks using a dedicated lane, even a dedicated lane with dedicated power, is that the cars: i) don’t have to all start at the same time or the same destination; and ii) don’t have to all end up at the same place.

          That’s why your second argument doesn’t really address the point I was raising above. Outside of the very densest parts of the densest metros, you don’t usually *have* 100 people going from point A to point B. Instead, you have 100 people going from 100 different places *somewhat near* point A, and going to 100 different places *somewhat near* point B (often times with a transfer or mode swtich along the way). Which is why 100 taxis can provide a service (point to point transportation) that the bus cannot – and why the elimination of the *driver* from taxis *might* end up changing the economics of that situation. It might not, of course – that’s a big “if” – but apart from the few metros where you truly have 100 people at a single “Point A,” a cost-effective autonomous car will end up wreaking havoc on the coverage routes and more modestly-ridden elements of the transit network.

          • Alon Levy April 27, 2017 at 11:58 am #

            I don’t know what “few metros” means. There exist metro areas with 2 million people with very high transit ridership, like Vienna, Stockholm, Zurich, and Prague, and at a slightly lower tier, Hamburg, Sapporo, Copenhagen, and Amsterdam. There even exist a few in the 1 million area or even a little less: Geneva, Basel, Karlsruhe, Heidelberg, Strasbourg, Bratislava, Brno, Helsinki.

            Taxis are an efficient replacement for coverage routes. But cutting coverage routes entirely is also an efficient replacement for coverage routes. By definition, the vast majority of ridership is not on coverage routes.

          • Alan April 27, 2017 at 12:50 pm #

            There’s maybe ten or twelve US metros that have any significant transit usage, as a proportion of overall commuter or transportation use: NY, LA, Chicago, DC, Philadelphia, Boston, SF, Atlanta, Seattle, Honolulu….maybe Minneapolis or Portland. Once you get beyond those metros (and maybe even just the first six), transit us is already a very small proportion of commuting or overall travel. Almost everywhere else, you have metros where you just don’t have the scarcity of urban space (or geometric cost of roadways) over a large enough area that transit *has* to be part of the equation. These are metros where *today* so few folks are using transit that even the smallest increase to the efficiency of the roadway network (or even none at all).

            It is possible to design metros of one or two million people that have high rates of transit usage, but for the most part our metros (outside of the handful above) aren’t designed that way.

            It is true that the vast majority of ridership is not on coverage routes. The vast majority of ridership is also limited to a handful of metros (NYC and a few others). Outside of those six or seven cities, cutting into coverage routes will take out a non-trivial amount of the systems – and I suspect that while AV’s aren’t going to replace super-high-density lines (the Lex Ave subway above, for example), they can replace a number of routes that transportation planners might not label as “coverage,” but aren’t space constrained enough to be unworkable for personal autos.

          • George Lane April 27, 2017 at 1:02 pm #

            “The difference with autonomous trucks using a dedicated lane, even a dedicated lane with dedicated power, is that the cars: i) don’t have to all start at the same time or the same destination; and ii) don’t have to all end up at the same place.”

            Yes, that is correct, I recognised that and demonstrated how automation will make trains even more efficient than they already are.

            “…you don’t usually *have* 100 people going from point A to point B. Instead, you have 100 people going from 100 different places *somewhat near* point A, and going to 100 different places *somewhat near* point B (often times with a transfer or mode swtich along the way). Which is why 100 taxis can provide a service (point to point transportation) that the bus cannot – and why the elimination of the *driver* from taxis *might* end up changing the economics of that situation. It might not, of course – that’s a big “if” – but apart from the few metros where you truly have 100 people at a single “Point A,” a cost-effective autonomous car will end up wreaking havoc on the coverage routes and more modestly-ridden elements of the transit network.”

            I feel as though you are agreeing with me here: cheaper taxis would reduce the need for coverage routes while increasing demand on ridership routes. We already see this with Uber having so many rides to metro stations.

            It’s important to remember that 1 bus is always going to be cheaper than 100 taxis, so even the taxi companies might set up their own bus routes to get punters to transfer.

          • Mike April 27, 2017 at 1:06 pm #

            Alon: he said US cities. Most of the US has a political allergy to density and transit. What will the Columbuses and Boises and Tacomas do: can they find 100 people going from point A to point B at the same time? Some cities are making their downtowns dense and walkable, but that’s only enough can fit only 5% of the metro’s population and only the top 5% can afford to live there. Fixing these metros’ land use so that there will be more 100-person groups for a bus or train, and getting them to agree to spend tax money on major transit improvements, is a harder nut to crack than accommodating autonomous cars.

          • Alon Levy April 27, 2017 at 6:10 pm #

            Okay, so we’re talking about one country with uniquely urban policy: no fuel tax*, low-rise zoning in inner-urban residential areas and single-family zoning elsewhere**, parking minimums, education policy that encourages white flight, suburbs with too much autonomy, lolzy-high infrastructure construction costs at all levels, buses with neither dedicated lanes nor signal priority, FRA regulations that discourage mainline passenger rail, poor intermodal integration.

            I don’t know what I’d recommend in Boise. In Columbus, there are reasonable corridors for on-street light rail (or buses with dedicated lanes and signal priority if you think that’s equivalent, which it isn’t). Lots of capacity to redevelop downtown parking lots and replace them with office towers, and to build townhouses and small apartment buildings in areas that are single-family today. Columbus has healthy population growth, unlike Detroit or Cleveland or St. Louis, so the city could expect to fill this zoning capacity.

            I know it’s possible, because Calgary and Vancouver have done just that. They’ve combined aggressive upzoning (just downtown in Calgary, or all over the Expo Line in Vancouver) with new rail construction and good bus-rail integration. Calgary is not Karlsruhe and Vancouver is not Vienna, but both have higher transit mode share than any US metro area except New York, and are growing their mode shares as they add new lines. Cost per rider used to be low, especially in Calgary, and is still affordable in ways almost nothing in the US.

            *Even when US governments have a fuel tax, they dedicate it to road construction and a little bit to transit. Good thing cigarette tax revenues aren’t deeded to tobacco advertising…

            **Most of Europe has low-rise zoning throughout the city, and English Canada has single-family zoning outside the high-rise inner-urban neighborhoods. US cities tend to have the lower of both worlds.

          • Alan April 28, 2017 at 5:43 am #

            Alon: Yes, we are talking about one country. Jarrett has been pushing back against claims that AV’s will have a transformative effect on transit…but as far as I know, those claims have mostly been made about the U.S. market. There are clearly some metros in the U.S. that exhibit the space constraints that Jarrett discussed – but most do not.

            George: One bus may be cheaper than 100 taxis, but it’s not clear that you need 100 autonomous cars to replace that one bus. Again, in the super-high density metros you might regularly have 170 riders on a bus – but outside those metros, those buses aren’t going to be nearly as full for nearly as often. You don’t need 100 cars to accommodate 10-20 boardings per hour. A typical diesel city bus costs upwards of $300,000 – about 10-15x the cost of a basic economy car. So depending on the cost delta for the autonomy features, it’s entirely possible that you might be able to get a dozen AV’s for the cost of an bus – which means that any bus route that’s hitting fewer than 20-30 boardings per hour is ripe to be taken out.

          • George Lane April 29, 2017 at 5:45 pm #

            I’d love to know which freeways would only have 10-20 boardings per hour on them. Your argument is that AVs will be so popular that no one catches the bus, and so unpopular that only 20 people per hour will use any one road. For your argument to be valid both of those things need to be true at the same time, however, those two things cannot both be true at the same time.

          • Alan May 1, 2017 at 5:44 am #

            George,

            I’m not saying that *freeways* have 10-20 boardings per hour – I’m saying that *bus routes* that are coverage routes outside of the higher-density areas are likely to have about 10-20 boardings per hour. It doesn’t take a whole lot of cars to replace those buses. There are certainly many, many more vehicles on the *roads* (freeways or otherwise) those buses travel on – again, outside of the top dozen or so cities, fewer than 10% of commuters use transit, so the overwhelming majority of trips are by private passenger car in those cities.

            So AV’s could, in theory, replace huge swatches of transit systems outside of those dozen or so of the densest metros – and a fair number of routes *within* those metros, but outside of the denser core. In other words, AV’s will be so popular that no one catches the bus – and *buses* are currently so underutilized in all but the densest metros that there no real spatial limitation to replacing those buses with private cars.

          • George Lane May 1, 2017 at 3:11 pm #

            “So AV’s could, in theory, replace huge swatches of transit systems outside of those dozen or so of the densest metros – and a fair number of routes *within* those metros, but outside of the denser core. In other words, AV’s will be so popular that no one catches the bus – and *buses* are currently so underutilized in all but the densest metros that there no real spatial limitation to replacing those buses with private cars.”

            Exactly. AVs will reduce ridership on coverage routes in low density metros while increasing it on their ridership services.

            You also cannot seriously believe that there aren’t at least 50 metros in the US with severe peak hour congestion in which adding more vehicles simply won’t work?

          • Alan May 2, 2017 at 6:02 am #

            The key is that nearly all metros in the U.S. are “low density” metros on this issue – again, there’s barely a dozen where commuter ridership gets over 10% of peak hour trips. In those metros, even many the “ridership” routes aren’t necessarily going to have enough riders to impose spatial limitations on switching over to AV’s.

            And yes, there’s very few metros in the U.S. with severe peak hour congestion that cannot add more vehicles. Again, in most metros the proportion of people riding transit is in the single digits. If most of those people switched from transit to passenger cars, that delta – the number of new cars – would be just a few percentage points on existing volumes. That’s probably within the day-to-day fluctuation of roadway usage. To say nothing of the fact that there’s probably not many of those 50 metros that don’t have 3-5% more people in the metro area than they did in 2010 – and 3-5% more private passenger cars – and still manage to function.

            Heck, even in *dense* metros you might be able to lose a sizable number of transit users and still “work” – the obvious example being DC. Transit ridership has fallen more than 14% in the last year, and is down more than 20% since the peak ridership (during a time of rising population) – and while peak hour is still horribly congested, you haven’t seen “carmageddon” in DC.

      • Alon Levy April 27, 2017 at 7:01 am #

        Actually… trains decouple all the time in Japan. There’s something called the Mini-Shinkansen: two trains are coupled on the Shinkansen line, and then they decouple, with the bigger train staying on the Shinkansen and the smaller one diverting to a legacy line. Decoupling also exists on some Tokyo commuter lines, where they run 15-car trains in the core but shorter trains (I think 10 cars?) on some outer branches.

        More extensive decoupling is under proposal. Reinhard Clever proposed it for California HSR in urban areas, where speeds are low and it’s fine to use a train segment without a streamlined nose. Driverless trains make this easier. So far it’s not being used anywhere with driverless trains, but it could be. For example, Vancouver should look into doing it if it extends SkyTrain deeper into Surrey and Langley, since there are 3 useful branches, and decoupling would allow trains to run to all three branches frequently.

        • BDawe April 27, 2017 at 9:15 am #

          That would seem to clash with the capacity imperatives of the Expo Line, which is moving towards open gangway sets to maximize the capacity

          • Alon Levy April 27, 2017 at 9:27 am #

            Does the Expo Line even run fixed-length trains? My recollection is that it couples trains of different lengths already, unlike the Canada Line. Shouldn’t be a problem ordering new trains exactly one half as long as the platform, with internal open gangways, but no gangways at all between the two coupled trains. One train goes to Lougheed, the other goes to Surrey and branches.

          • BDawe April 27, 2017 at 3:03 pm #

            The Expo Line today runs six-car (three-married pair) Mark I trains, four car(two married pairs) Mark II trains, and four car (one open gangway unit) Mark III trains

            The Mark IIs sorta meet your description, but they don’t take up the whole platform quite

        • el_slapper May 2, 2017 at 5:04 am #

          It’s also marginally existant in France. I’ve seen a few suburb trains do that in Paris-Nord, and my night train from Paris to Annecy did this routinely in the morning(I think it was in Chambéry) in 1998.

          It’s not an easy manoeuver. Especially when in Paris the speaker has to warn everybody that “the first 4 cars go to Chantilly-Gouvieux, the last 4 go to Persan-Baaumont”. Plenty of people ended up in the wrong cars.

  2. Dexter Wong May 2, 2017 at 12:34 am #

    I have seen light rail cars joined as a train reach a junction point, separate and proceed on separate lines (laid in the street) as single cars (and the reverse). So you can’t say in general that train cars cannot separate and proceed on different lines.

  3. lue March 25, 2019 at 11:31 pm #

    More extensive decoupling is under proposal. Reinhard Clever proposed it for California HSR in urban areas, where speeds are low and it’s fine to use a train segment without a streamlined nose. Driverless trains make this easier. So far it’s not being used anywhere with driverless trains, but it could be. For example, Vancouver should look into doing it if it extends SkyTrain deeper into Surrey and Langley, since there are 3 useful branches, and decoupling would allow trains to run to all three branches frequently.