The Coming Disruption of Transport

[kEiThZ]

^I don’t understand why we always make this debate about Elon

Because the origin of virtually all of these fanciful claims and the fount of faith in them is Elon. I don't think most of Tesla's own line engineers (and I actually know some) would be anywhere near as bullish.

But the stock price demands that he keeps telling the faithful that the next great thing is just around the corner, and of course his next product is going to obliterate the need for public transport immediately.

In case anybody wants to argue that I am just a Tesla hater, I'll just say I invested in the company in 2012. Tesla stock paid for my big fat Indian wedding. I saw potential there before most people had heard of the company. But we're entering silly session now that Musk has got a taste of celebrity and is more focused on headlines than tech. Want that same 2012 spirit In an EV startup? Look at Rivian.

 

[kEiThZ]

The question is not, are the engineers happy - the question is, are the risk managers and regulators and insurers and litigators happy. And the politicians who have to answer for what is permitted on the roads.

I'd agree with you with the caveat that guys like Musk are extremely effective at arm twisting politicians and dazzling them with bullshit. Look at all the, "We're not going to build HSR because we're going to build Hyperloop!" nonsense that gets put out in North America. Musk is great at providing politicians with excuses on why they shouldn't invest in public transport. And imminency is very much part of that schtick.

In the short term, fuel prices are escalating and labour costs much less so. That gives some idea what the short term leverage may be. Rail will prevail a little longer based on fuel efficiency. Drivers only produce small amounts of carbon, too.

All commodity prices are going up. Charts for copper, lithium, nickel and cobalt over the last two years make recent oil spike look timid. It's actually killed the routine decline in the price of renewables and batteries. It'll get back on track. But it'll take a while.

Weather may actually create an economic differentiator that tilts our economy….. if AV’s work well enough in the sun belt, it could create an economic advantage for commerce in that region that attracts industry and distribution investment that the snowbelt will have to compete with somehow.

I'm not sure I buy the idea that AEVs will necessarily improve economic competitiveness in the long run. Short term maybe. In the long run, if it results in perpetuation of the suburban growth ponzi scheme, that will be a disaster for society and massively increase infrastructure costs. There's no fundamental way to get past the geometry problem. No matter how many pedestrians get run over.

 

[afransen]

We are easily decades away from the robotaxis vision that Musk sells on stage.

Care to explain Waymo, that is operating commercially today, sans safety driver?

 

[afransen]

Local bus routes in cities like Toronto and Montreal are here to stay. There's enough demand that AVs can't keep up. Given our regulatory environment, larger vehicles like buses and trucks will probably face difficulty being allowed (current regulations, as I understand them, require a driver in the car anyways. I don't see that going away soon).

If it is just regulation not driven by practical safety considerations, those regulations won't stand for long. And there are autonomous ridehail and trucking operations in the US that are performing trips without a safety driver.

Given inertia and capacity demands, my personal prediction is that AVs will be widely used in the 905 to feed GO stations and act as community shuttles, while bus service on major arterials will continue to be served by larger vehicles. The advantage of trains and buses are not that they're cheap, their high capacity.

I think you're confusing what you wish would happen with what will happen. There is not government agency that will be mandating the AEVs will bring passengers to GO stations only. Many will take them to their destination. The only curbs on that will be cost (requiring road tolls) and speed (oppressive congestion induced by AEVs).

The most common rebuttal to the above points I've seen is that AV efficiency will offset the new traffic. But that doesn't solve the inherently inefficient geometry of cars, nor does it make it safer for cars to travel closer together on the roads. When you're operating a giant moving metal box, you need some space around you for unexpected events like road conditions or undetected obstacles. You can't just pack everything close together, because even if driver efficiency is totally eliminated (computers have inefficiencies too), unexpected conditions can still occur.

Yes, we will have a tragedy of the commons. This is why we need to plan for this eventuality.

 

[afransen]

But trains of AV electric trucks trundling down the TCH hauling coal and grain? I’m not buying that, yet.

This is strawmanning. The claim was about intermodal. Door to door trucking will likely still be more expensive than rail, but it may be so much faster and the cost premium much reduced, shifting a lot of higher value consumer goods (ie, not commodities like coal and grain) to door to door trucking.

 

[afransen]

In the long run, if it results in perpetuation of the suburban growth ponzi scheme, that will be a disaster for society and massively increase infrastructure costs.

This is why I have been expressing concern that there is no plan to manage the technology as it gains adoption.

 

[MrGoose]

This is strawmanning. The claim was about intermodal. Door to door trucking will likely still be more expensive than rail, but it may be so much faster and the cost premium much reduced, shifting a lot of higher value consumer goods (ie, not commodities like coal and grain) to door to door trucking.

Why? Thats not efficient, because you're trying to stuff a long-distance truck/ train i.e. high volume, optimized for highway speed, probably very long and large, into small low volume streets which are better suited for a deliver van or cargo bike.

Care to explain Waymo, that is operating commercially today, sans safety driver?

Because putting a LiDAR bulb on the top of each AEV and then asking it to perceive the environment around it is asking it to fail. Eventually the computer will encounter a corner case and it will act unpredictable if not just shut down.

The built environment of cities is so much more varied and chaotic than a computer could possibly handle without being handheld.

Think about all the road markings and the design that goes into roads to keep human drivers from crashing into each other. For best AEV results, you'd have to duplicate those road markings but in preferably a computer-friendly format. That means RF lines on the roads, maybe QR-code style signs etc.

 

[MrGoose]

This is why I have been expressing concern that there is no plan to manage the technology as it gains adoption.

The trick to managing AEVs is redesigning our cities around humans and then banning AEVs/ all personal motor vehicles from cities.

 

[kEiThZ]

Care to explain Waymo, that is operating commercially today, sans safety driver?

In a substantially sandboxed mode with remote human backup and monitoring. That is good for development. But not substantially scalable. Getting to what Musk or Seba envisions from here is orders of magnitude more complicated. Now, AI also improves exponentially. So we'll get there eventually. But I think we're closer to decades out from removing the steering wheel, than years away.

Also, these estimates ignore market lag. How many people have the money to buy a brand new car as soon as autonomy is announced? How many cars can a TaaS service bring online as soon as they achieve autonomy? What happens to the cars already on the road? Given that cars stay on the road for about 12 years, I think that is about the minimum time from when TaaS is launched in a given market for it to replace cars in that market. By this metric, we're a very long time away from replacing the TTC and GO with TaaS for all.

This is why I have been expressing concern that there is no plan to manage the technology as it gains adoption.

We're not at the point of genuine mass adoption yet. We don't see people buying cars based on the level of automation yet, or basing housing deciding based on owning a substantially automated vehicle. When that happens we'll probably start having these conversations.

I do agree that they need to happen. Folks like Musk drive the conversation entirely through an anti-urbanist and elitist lens. In his future vision, the poor and condos don't exist, and you'll spend two hours in your cars auto-commuting from your distant exurb. Think of how dystopian this vision actually is. Seba is better and presents a pro-urban where TaaS becomes a public good, effectively. But this ignores the fact that a lot of people don't actually want to share vehicles. They'll happily clog up the roads for their personal comfort. So at some point we will need to have real conversations about what we want our cities to look like and how serious we are about sustainability.

 

[DirectionNorth]

If it is just regulation not driven by practical safety considerations, those regulations won't stand for long. And there are autonomous ridehail and trucking operations in the US that are performing trips without a safety driver.

LOL. What do our regulations for light trains and their separation from freight look like?

I think you're confusing what you wish would happen with what will happen. There is not government agency that will be mandating the AEVs will bring passengers to GO stations only. Many will take them to their destination. The only curbs on that will be cost (requiring road tolls) and speed (oppressive congestion induced by AEVs).

I am not saying that it is going to be the only use of AVs, I am saying that they will not replace arterial/high demand transit routes.

Yes, we will have a tragedy of the commons. This is why we need to plan for this eventuality.

But I don't think it will happen as fast as you think, because cities are complex environments that computers can't simulate. Nevertheless, you are right that we should be making plans for when it happens.

 

[crs1026]

Why? Thats not efficient, because you're trying to stuff a long-distance truck/ train i.e. high volume, optimized for highway speed, probably very long and large, into small low volume streets which are better suited for a deliver van or cargo bike.

Because putting a LiDAR bulb on the top of each AEV and then asking it to perceive the environment around it is asking it to fail. Eventually the computer will encounter a corner case and it will act unpredictable if not just shut down.

The built environment of cities is so much more varied and chaotic than a computer could possibly handle without being handheld.

Think about all the road markings and the design that goes into roads to keep human drivers from crashing into each other. For best AEV results, you'd have to duplicate those road markings but in preferably a computer-friendly format. That means RF lines on the roads, maybe QR-code style signs etc.

I am surprised that a global standard for AI-friendly road marking and wayfinding isn't already out there and being implemented, at least where new roads are being built or for the core highway network.

I'm also surprised that the various developers have set the threshold for initial rollout such that one needs an all-knowing vehicle capable of handling every last scenario in North America. That makes the design process so much more complex.

I would have thought that the first step would be to exploit the "simple" applications - such as fixed routes for things like slow speed shuttle bus applications. A much more finite number of navigation challenges and road conditions, and potential to strategically target road anomalies that are the most challenging. Things like the shuttle between an airport terminal and the rental car operation, for instance.

As for intermodal, it must be much simpler to figure out the AI to allow drayage between specific key customers and the intermodal yard, than to figure out the AI for the entire TCH between container terminals on the Pacific Coast and the GTA, for instance. That's a very immediate savings without boiling the ocean. In the GTA, for instance, I bet one could pick the top 20 or 30 logistics centers and automate the drayage between these and the two (soon to be three) rail intermodal terminals. A very small set of roads to figure out and equip for smart vehicles.

- Paul

 

[afransen]

Why? Thats not efficient, because you're trying to stuff a long-distance truck/ train i.e. high volume, optimized for highway speed, probably very long and large, into small low volume streets which are better suited for a deliver van or cargo bike.

Huh? In what way is rail freight relevant to getting trucks off tight urban streets?

If I am shipping strawberries from California to Toronto, I go by road even if it is more expensive than rail. Time sensitive or high value goods are vulnerable to shifting away from rail onto trucks. It also opens new possibilities from supply chain perspective for firms if they can reach more of the continent within 1-2 days by truck (if those trucks can roll more or less continuously (pausing to charge). If I am Amazon, I might also be shifting air freight back to road if it can roll 24hrs/day without paying for team drivers.

 

[afransen]

The trick to managing AEVs is redesigning our cities around humans and then banning AEVs/ all personal motor vehicles from cities.

Banning cars is definitely a bold solution. What's our plan to make that politically feasible?

 

[afransen]

In a substantially sandboxed mode with remote human backup and monitoring. That is good for development. But not substantially scalable. Getting to what Musk or Seba envisions from here is orders of magnitude more complicated. Now, AI also improves exponentially. So we'll get there eventually. But I think we're closer to decades out from removing the steering wheel, than years away.

Also, these estimates ignore market lag. How many people have the money to buy a brand new car as soon as autonomy is announced? How many cars can a TaaS service bring online as soon as they achieve autonomy? What happens to the cars already on the road? Given that cars stay on the road for about 12 years, I think that is about the minimum time from when TaaS is launched in a given market for it to replace cars in that market. By this metric, we're a very long time away from replacing the TTC and GO with TaaS for all.

We are only 15 years post-iPhone. Think of all the business models that have been disrupted in that time and how far technology has advanced. Decades is a very long time. My personal estimate is about 10 years to see major disruption from autonomy. That is well within the planning horizon for our major infrastructure investments though. If Tesla is successful, they can deploy a pretty large number of autonomous vehicles pretty quickly. They have a track record of being able to move quickly that more bureaucratic organizations (even including Alphabet) can't match. TaaS will attrite the car fleet far faster than the replacement rate. One TaaS vehicle can do the work of several private vehicles, and if TaaS is coincident with the transition to electric, a lot of low utilization older gas cars will be displaced by a relatively small number of TaaS vehicles. And the fleet size is one thing, vs share of VMT.

I don't think TaaS will replace TTC/GO. I don't think that should be a goal, nor would it be desirable. But those organizations should include in their long range planning how they remain relevant in a transportation landscape with TaaS. They need to get better at moving people longer distances at higher average speeds, and plan for how their short distance services may need to change to remain relevant. Does TTC start operating on-demand minibuses? How can terminals be adapted to support that operating model? Does GO start operating station-to-station minibus service?

We're not at the point of genuine mass adoption yet. We don't see people buying cars based on the level of automation yet, or basing housing deciding based on owning a substantially automated vehicle. When that happens we'll probably start having these conversations.

I do agree that they need to happen. Folks like Musk drive the conversation entirely through an anti-urbanist and elitist lens. In his future vision, the poor and condos don't exist, and you'll spend two hours in your cars auto-commuting from your distant exurb. Think of how dystopian this vision actually is. Seba is better and presents a pro-urban where TaaS becomes a public good, effectively. But this ignores the fact that a lot of people don't actually want to share vehicles. They'll happily clog up the roads for their personal comfort. So at some point we will need to have real conversations about what we want our cities to look like and how serious we are about sustainability.

I think Musk definitely has blind spots when it comes to how this technology will affect cities. It frankly doesn't matter what proponents think, the technology will have the effects it has. I think many people will pay a premium (UberX style) for a private ride. But shared rides have the potential to be a lot cheaper and that will be relevant for people using it for every day transportation and not occasional use. I think we'll see different vehicles for the different use cases. Small 2-4 seaters the size of current sub compacts for private rides, and larger vehicles for shared rides. I think people would tend to be more comfortable sharing with more than 1 other person.

I agree that 2 hours of commuting from distant exurbs is not desirable. We already have that today and when the penalty of this is reduced with autonomy and electrification it will get much worse. We need to become comfortable with the idea that we will have the toll our highways sufficiently to contain this use case and keep traffic from becoming even more catastrophic. Those tolls will also drive people into shared vehicles, even if they have to transfer to a private vehicle when they get closer to their destination. It will be like carpool lots on steroids.

 

[afransen]

LOL. What do our regulations for light trains and their separation from freight look like?

Ah yes, I forgot about the trillion dollar public transit industry and all the lobbying dollars it can bring to bear on policy makers, as well as public outrage about not letting transit vehicles on certain rail corridors.

I am not saying that it is going to be the only use of AVs, I am saying that they will not replace arterial/high demand transit routes.

That is certainly an aspiration. How do we know that is the case? If people ride the bus 7km for $3.50, and a shared AEV can carry 6 passengers and can be operated for less than $0.50/passenger km, why won't people switch? Traffic is bad? Why won't that affect the buses?