[mdrejhon]

I'd like to see insurance move from a per vehicle to per driver coverage. Otherwise if I loan out my car and the borrower hits someone, I'm liable.

It's definitely something needing consideration.

I expect a special class of driverless rideshare insurance to become available, to eventually satisfy the lawyers. Who provides the insurance is an open question, but probably a combination of the manufacturers & reinsurerers, in conjunction with the app software vendor. Some kind of 21st century Rube Goldberg paperwork trail that finally satisfies all the legal beagles.

TaaS will be a bit slow in coming. It'll be easy to deploy TaaS in Arizona this decade, but not in Canada. Well-maintained roads and climate-stable areas such as Arizona -- I've driven through parts of the area as side trips from my CES convention trips -- the roads are absoltely beautiful & impeccable in many parts.

I see some cities 100x cheaper to insure driverless cars in. 5 years could be realistic for some of them. Even other areas, like Las Vegas, their roads are absolutely stunning in maintenance, smoothness, well marked, rarely faded paint to confuse AI, great driver-view efficiency (but crappy for pedestrians and cyclists mind, you -- few pedestrians dare venture onto road surface, reducing liability risks). Pedestrians never dash out onto the roads in most downtown cores there -- jaywalking is much lower because roads just are that much more dangerous to pedestrians. The infrastructure lasts a really long time with less wear and tear. So there's far less infrastructure deficit in some parts of the states thanks to the area-climate advantage.

As you can see, it would be much easier to cheaply insure driverless vehicles for unattended duty. The auto status quo is quite evident in these parts of the United States and it's no wonder that driverless vehicles would be quite much more trusted in some of these durable, good-weather infrastructures. Drivereless vehicles are going to invade these infrastructures much sooner. What you see in 5 years may be 20 years in Canada.

Now, this bike-sharing user, car-sharing member (who still owns his first and last car), transit-advocate -- wants to see a lightening of a car status quo -- is chagrined to see that it's going to take more than a century (or two) to fix the environmental ovrershoot of automobile status quo -- but clean BEVs and drivereless vehicles will be a necessary mix going forward in slowing backing away from the environmental waste of two billion mostly-empty parking spots in the United States. So I welcome the idea of driverless vehicles and per-capita reduction in car ownership, creating an amplified reduction in parking needs.

That said, there is really a very firm entrenchment (more than many think) -- there are cities I visited that are 100x more car-dependant than car-happy Hamilton, Ontario. Really impeccable well-maintained automobile infrastructure that is easy for AI self-driving cars to key on with low risks of deaths to pedestrians.

 

[kEiThZ]

It'll be easy to deploy TaaS in Arizona this decade, but not in Canada.

Eh. No. Anybody who says this doesn't understand how far ahead sensor tech is and how fast Machine Vision and AI are progressing. Operating in inclement weather and climate is all about the right sensor mix. This kind of automation has been slow because some have picked tech that might not be as effective (LiDAR for example) in poor weather. As for pedestrians darting out? The computer will always react faster than you can. Not only will it react faster than you. But it can determine the risk on several different courses of action and pick the best one. And it can manoeuvre the vehicle with such precision that it can execute escape and emergency actions that you never could.

And if we're talking a decade down the road? The sensors, the AI, etc. will all be so far ahead that there will be no difference between driving in Phoenix or Hamilton. We're talking a decade. Don't forget. That was about the time period that a lot of the world went up dial up or barely having cable to having a smartphone.

 

[TheTigerMaster]

Eh. No. Anybody who says this doesn't understand how far ahead sensor tech is and how fast Machine Vision and AI are progressing. Operating in inclement weather and climate is all about the right sensor mix. This kind of automation has been slow because some have picked tech that might not be as effective (LiDAR for example) in poor weather. As for pedestrians darting out? The computer will always react faster than you can. Not only will it react faster than you. But it can determine the risk on several different courses of action and pick the best one. And it can manoeuvre the vehicle with such precision that it can execute escape and emergency actions that you never could.

And if we're talking a decade down the road? The sensors, the AI, etc. will all be so far ahead that there will be no difference between driving in Phoenix or Hamilton. We're talking a decade. Don't forget. That was about the time period that a lot of the world went up dial up or barely having cable to having a smartphone.

Also should be noted that EVs have far better traction control over snow and ice relative to ICE vehicles. The cars can very quickly detect slippage, and finely tune each of the four wheels to maintain control. You’ll never see that level of traction control in ICE vehicle with a traditional engine

Modern wireless technology can already locate vehicles to within centimetres on the road. And onboard sensors can trivially see through snow and ice to avoid obstacles obscured by snow.

I’m not convinced that Level 4 or 5 autonomy will be a thing in the near term (within 10 years), but when we do figure it out, expect these autonomous cars to be driving reliably in inclement weather shortly thereafter. I don’t think snow and ice would be anywhere near as big as an impediment as people imagine it to be,

 

[mdrejhon]

Eh. No. Anybody who says this doesn't understand how far ahead sensor tech is and how fast Machine Vision and AI are progressing.

I know, but hold on, that's not why.
I'm being smarter reading between the lines.

I'm aware how fast it is processing. But legislative won't catch up. Legal liability is easier on some American roads that have very strong anti-jaywalking laws that favour the vehicle. Jaywalking is kind of a shaming-invention by the auto industry.
The forgotten history of how automakers invented the crime of "jaywalking"
Pedestrian killed in the road? They'll just make the law favour the car, to solve the initial liability problem in some cities. There are some really STRONG anti-jaywalking laws in some cities down there... repeat jaywalking is a jailable offense in some jurisdictions, did you know? It's sad the way that is.

This is only 1 of about dozens of complications, ranging from mudane (Canadian winter potholes) to winter-faded paint lines, and blizzarded roads. We've got very unique Canadian complications. 90% of this will be easily trainable to car AI, but the difficult 10% will take longer (hotly debated how longer, but I'd say 10+ years for public trust). Possibly more. Even when we successfully do, we might need more line-items of redundancy such as camera redundancy (and self cleaning camera lens etc) than a lot of American cars, in the continuous dirty-salt-spray regime of a wet desalted road that easily obscures 20 lens of cameras simultaneously in mere seconds -- tech that likely needs to be added to Canadian cars that are permitted to selfdrive. We've got some particular customizations we will also need too. The lidar and radar stuff will obviously help a lot too, but doesn't eliminate the danger we have almost no braking safety margin, regardless of human ro AI driver -- an AI will be unable to completely avoid crashes.

Also, in respect to a car's decision whether to operate or not -- is going to be something that Canada faces. Humans decide not to drive in a blzzard, and there will be thresholds for driverless cars. Remember, driverless cars will have the same problems in blizzards as humans -- and they will have to legally & infrastructurally self-care for themselves somehow -- self-pulling over, or self-calling a tow truck -- and they will still undergo pileups if they drive as risky as human drivers and one start dominoing on a slippery road, etc. Unless we want to live with huge increases in cars refusing to selfdrive themselves in minor blizzards.

Those cars will be smarter than we are, but Ministry of Transportation may not tolerate that New Equilibrium, etc. Those 100-car-strandings will still occur with drivereless vehicles that unexpectedly get caught in a major-icing or major-blizzard event that pushes them past their safety tolerances.

You know, those wimpy weather forecasts but suddenly ice rain comes and suddenly ices up the road, creating 100 cars sliding off the road (it has happened to me before -- a 20 km stretch of freeway suddenly got iced, with over 75 slid vehicles, some spun, some rollovers, and some gored into snowbanks, at regular intervals). Self driving vehicles won't be immune to those events. Best case scenario, smarter ones will safely pull over and just strand themselves willingly until dug out by humans, causing the owners to wonder where the heck their car was if it was on ridehail duty. But legislative stuff won't catch up quickly.

New laws will need to mandate instructions to handle those situations with thresholds acceptable to humans. Even arguments about "No drivereless cars shall attempt a journey when forecast is X" (contentious!) to things like "In the event a road suddenly exceeds safety thresholds, driverless cars must do X and X to safe themselves, and if stuck, must call X to get emergency servicing" (contentious too!). Nobody will agree in these new legal rules. It's probably going to take 20 years to ferret the new legal equilibrium out for Canada. Consequently, we will probably have limited empty Ubers in 5 years in good weather, but we're not expecting empty Ubers in a blizzard quite just yet, because the legal framework will use wimpy thresholds for Canada at first, requiring lots of safety drivers for drivereless cars still training itself to Canadian winters, etc. Regulatory will lag very severely here in Canada, unlike empty-car-happy Arizona cities that has already permitted them.

There's a lot of interplay factors that you have to think through.

 

[kEiThZ]

I'm aware how fast it is processing. But legislative won't catch up. Legal liability is easier on some American roads that have very strong anti-jaywalking laws that favour the vehicle.

Your previous argument was that this would all be easier in Arizona than Canada. Why would legislative trends be very different in Canada than the US? If anything, I think the Americans may have a tougher time given their more litigious society.

 

[kEiThZ]

Remember, driverless cars will have the same problems in blizzards as humans -- and they will have to legally & infrastructurally self-care for themselves somehow -- self-pulling over, or self-calling a tow truck -- and they will still undergo pileups if they drive as risky as human drivers and one start dominoing on a slippery road, etc. Unless we want to live with huge increases in cars refusing to selfdrive themselves in minor blizzards.

You know, those wimpy weather forecasts but suddenly ice rain comes and suddenly ices up the road, creating 100 cars sliding off the road (it has happened to me before -- a 20 km stretch of freeway suddenly got iced, with over 75 slid vehicles, some spun, some rollovers, and some gored into snowbanks, at regular intervals). Self driving vehicles won't be immune to those events. Best case scenario, smarter ones will safely pull over and just strand themselves willingly until dug out by humans, causing the owners to wonder where the heck their car was if it was on ridehail duty.

Sorry. But you just don't understand the tech if this what you think. I don't mean to be harsh. Right now, we monitor every modern airliner with hundreds of points of data in real-time. We make conditional maintenance decisions before they even land at their destination. This is without serious AI. Do you think we're just going to let loose self-driving vehicles randomly on the road?

It's an entirely different construct. Those AEVs will be closer to some combination of how we operate aircraft today, along with a neural net and continuous monitoring behind them. They will be fed weather, traffic and road friction information continuously. Each vehicle will be analyzing sensor visibility and wheel slip and feeding it back to the neural net. Large AI systems will be monitoring all this and making decisions on where to deploy vehicles, reduce service, slow down service, etc. The risk of those 100 car pileups will be so much lower. And if that happens the response time will be sufficiently faster.

There will be no real ramp to the capability. I don't think networked Level 5 autonomy will even be allowed before vehicles are capable of the above.

 

[mdrejhon]

Sorry. But you just don't understand the tech if this what you think. I don't mean to be harsh. Right now, we monitor every modern airliner with hundreds of points of data in real-time. We make conditional maintenance decisions before they even land at their destination. This is without serious AI. Do you think we're just going to let loose self-driving vehicles randomly on the road?

I'm not fixating only on tech or only on legal. (I did so only as a counterbalance)

By now, it should be obvious I'm seeing the entire apparatus -- the entire chain.
What I've written is merely simply cherrypicked examples, but it seems you've fixated on the cherrypicks.

Backing a bit to see the whole picture:
-- We can see how slow and polarized government is. It's going to take a while.
-- I'm a high tech worker, software developer (tech familiarity), read my LinkedIn, I'm also an ex-bank-employee (fiscal system familiarty). I've seen quite a lot of tech and Dilbert moments. And the different company cultures too. (To credit, even Zuckerberg and Musk disagrees fiercly over how dangerous AI is -- and that also rears itself in how AI is developed differently by different companies. We all need to trust the AI, as not all AI is created equal. It's amazing stuff).
-- And my spouse ran for city councillor. (govt efficiency familiarity), and he also took paralegal training (legal efficiency familiarity) in a fiercely-competed ward in Hamilton, Ontario. He's delegated a few times at the provincial level too. I've seen first-hand the inefficiencies/dirty stuff that goes on at multiple government levels, agencies. I'm not going to comment further, but if you're familiar with politics, your gray hair grows faster no matter what color you affiliate with.

So repeating: I know enough to se the entire apparatus. I don't claim to know everything. But I know enough that it's not going to be as easy walk here.

Anyway, I've witnessed first-hand a lot of inefficiencies/delays in multiple spheres, from legislative to managerial to other complications.
Plus, I know AI is going to be big. I'm 33% AI stock in my RRSP - I am long NVDA stock in my RRSP (an AI play) -- it's currently overweighted at entirely 33% of my RRSP, and that stock has done well over the last few years. So, your assertion I don't know tech, is total nonsense.

I agree with a lot of your points. We'll definitely solve the problems. But we're not going to solve it in 5 years. I was writing cherrypicked examples of why it's not going to take as little as 5 years in Canada to trust the entirety of the Ontario road networks 24/7/365 to driverless cars, but I think you latched on them selectively. The first-pass imperfect rules we develop years for selfdriving cars in a blizzard, will not be the final (better) rules we write in future years, it will all be refined. Part of the point is that at the beginning it will be woefully imperfect and contentious, you see.

Move on, next debate.

There will be no real ramp to the capability. I don't think networked Level 5 autonomy will even be allowed before vehicles are capable of the above.

Mutual agreement there.

Level 5 is coming, just not as quickly in Canada as in some parts of USA. Choose your number. 2 yr vs 5 yr. 5 yr vs 10 yr. 10 yr vs 20 yr. It doesn't matter. The number fixation becomes semantics at this stage; a major point is it's going to take longer here than there. That's part of the broad, aggregate point -- statistically speaking, the odds of meeting the video mentioned targets in 5 years is far less likely in, say, Ontario than, say, Arizonia.

As you can imagine, it is not going to be sudden Level 5 simultaneously in all cities in all states/provinces. It'll sort of phase-in here-n-there, some sooner than others. I'll easily bet my mortgage on that. And, much of us are in mutual agreement it's not going to be a sudden 80% North America Wide reduction in cars in just a mere 5 years.

 

[crs1026]

Your previous argument was that this would all be easier in Arizona than Canada. Why would legislative trends be very different in Canada than the US? If anything, I think the Americans may have a tougher time given their more litigious society.

The thing to remember is that while AI will lead to fewer accidents than at present, there will still be some. And there will have to be insurance for that situation.

I would assume that in an AI environment, with people just being passive loads, there will be no driver element to liability. The only possible places to point for liability are the road maintainers, the other vehicle, or the vehicle design and software. In theory, that should lead to no-fault payouts to victims.

Insurers work from actuarial assumptions. So, to start, if Brand A is predicted to have 1 accident every 10 giga-kms, and Brand B has an accident every 9 giga-kms, and the average accident creates $XM in benefits paid, it's pretty easy math to set the premiums for a vehicle. But.... once experience kicks in, and insurers note that Brand B's claims rate isn't decreasing at the same level as Brand A, or if it increases because some latent flaw hasn't been fixed, they will go looking for reasons why. And that may cause them to ask for fine detail on what is causing the accidents.

It will be impossible for the average survivors of an accident to mount a legal challenge when an AI vehicle screws up - the vendors will have huge legal departments, the average accident victim doesn't, and it will take enormous resources just to understand what went wrong.... let alone prove it in court. But insurers have big legal departments also, and they have financial interest in not absorbing costs if they can pass the liability to someone else. So eventually it's insurers who will be chasing manufacturers for why their vehicles seem to fail under situation X.

- Paul

 

[lenaitch]

I’m not a Uber user, so I will ask.....what do the current ridesharing apps do about tax records? I would guess it’s an auditor’s windfall, since it’s cash-free and every transaction is recorded on line. Drivers must be accustomed to declaring the income. Unlike old school taxis, where “off the meter” rates were practised by many in cash fare days, and tips always were unrecorded.

Neither am I so I cannot definitively answer the question, but have wondered it myself. I have to believe there is a reporting requirement by the 'collector of money' for want of a better term, regardless of whether the providers are employees or independent contractors. Ultimate income reporting is the responsibility of the individual, but all reporting forms I get to report the CRA are copied to them as well by the issuer.
The question would apply to AirBNB, food delivery and all of the so-called gig economy employment streams as well.

 

[lenaitch]

Sorry. But you just don't understand the tech if this what you think. I don't mean to be harsh. Right now, we monitor every modern airliner with hundreds of points of data in real-time. We make conditional maintenance decisions before they even land at their destination. This is without serious AI. Do you think we're just going to let loose self-driving vehicles randomly on the road?

It's an entirely different construct. Those AEVs will be closer to some combination of how we operate aircraft today, along with a neural net and continuous monitoring behind them. They will be fed weather, traffic and road friction information continuously. Each vehicle will be analyzing sensor visibility and wheel slip and feeding it back to the neural net. Large AI systems will be monitoring all this and making decisions on where to deploy vehicles, reduce service, slow down service, etc. The risk of those 100 car pileups will be so much lower. And if that happens the response time will be sufficiently faster.

There will be no real ramp to the capability. I don't think networked Level 5 autonomy will even be allowed before vehicles are capable of the above.

This assumes a highly regulated road travel 'industry'.

Regrdless, alternative, non-embracing viewpoints are being dismissed, so I will bow out.

 

[crs1026]

Sorry. But you just don't understand the tech if this what you think. I don't mean to be harsh. Right now, we monitor every modern airliner with hundreds of points of data in real-time. We make conditional maintenance decisions before they even land at their destination. This is without serious AI. Do you think we're just going to let loose self-driving vehicles randomly on the road?

The issue is the variability of the roadway. Runways don’t develop diversions with orange pylons. I have no doubt that an AI vehicle could autopilot its way down a roadway in fog or snowstorms. I have a hunch that there will be embedded wayfinding markers too, such as RF tags in the pavement or visual bar-code markers along the roads to recalibrate position and allow dead-reckoning adjustments every so many metres.

The interesting thing will be the average road crew. Right now we have utility work in our area - as the crews work, they juggle their pylons every couple hours. Roads are not constants. I can imagine moving to a marine-like “”red and green bouy” approach to give clues to AI cars as to where the road has gone. But it will take a huge “Disruption” to get seat-of-the-pants contractors to use them to mark roadways consistently. A fixed gps database refreshed annually may suffice for Garmin devices (most of the time), but AI cars will need dynamic data on the latest changes to roadways. All it takes is one crew doing utility cuts....sinkholes? flash floods?

On the positive side, I can see automated real time pothole reporting. And, perhaps road work will no longer have money-for-nothing paid duty cops assigned.

- Paul

 

[MisterF]

The issue is the variability of the roadway. Runways don’t develop diversions with orange pylons. I have no doubt that an AI vehicle could autopilot its way down a roadway in fog or snowstorms. I have a hunch that there will be embedded markers too, such as RF tags in the pavement or visual bar-code markers along the roads to recalibrate and allow dead-reckoning adjustments every so many metres.

The interesting thing will be the average road crew. Right now we have utility work in our area - as the crews work, they juggle their pylons every couple hours. Roads are not constants. I can imagine moving to a marine-like “”red and green bouy” approach to give clues to AI cars as to where the road has gone. But it will take a huge “Disruption” to get seat-of-the-pants contractors to mark roadways consistently. A fixed gps database may work for Garmin devices (most of the time), but AI cars will need dynamic data on the latest changes to roadways. All it takes is one crew doing utility cuts....

On the positive side, I can see automated real time pothole reporting. And, perhaps road work will no longer have money-for-nothing paid duty cops assigned.

- Paul

Autonomous cars don't rely on mapping to navigate. They constantly sense everything going on around them and can see way more than a human can. There's nothing about a construction site or changing road markings or potholes that would confuse an autonomous car any more than it would confuse a human driver.

 

[kEiThZ]

Autonomous cars don't rely on mapping to navigate. They constantly sense everything going on around them and can see way more than a human can. There's nothing about a construction site or changing road markings or potholes that would confuse an autonomous car any more than it would confuse a human driver.

This. I don't know if it's a communication problem or people just don't understand how AI and sensor fusion works in a self-driving environment.

 

[kEiThZ]

What I've written is merely simply cherrypicked examples, but it seems you've fixated on the cherrypicks

Not so. I'm struggling to understand your viewpoint. Your earlier examples seems to suggest that L5 autonomy is easier to implement in other jurisdictions because of their operating environment. I don't agree that those differences are all that substantial.

Then you suggest that it's legislative. And I'm still not understanding what specifically you think are the legislative differences that would enable faster implementation in the US. I see a place which lags on technology. I spent the last few years just outside the Bay Area and still had so many places without tap or chip and PIN for credit. On the other hand, Canadians adopted electronic payments, NFC payments, etc at faster rates. I also mentioned the issue with liability in the US. They are much more litigious. And the payouts are larger. Doesn't sound like a place I'd want to launch my autonomous vehicle service.

 

[Admiral Beez]

I am always struck by how inflexible personal vehicle ownership is.

Try motorcycles. Some guys are so fanatical about their machines, they’d sooner let you date their girlfriend than ride their bike.