GM's big bet on driverless cars turns sour
theverge.com
theverge.com
So, the regulators must build a more rigorous testing/scoring system before these systems are allowed on the open road without a human minder. The existing ones designed for humans have many assumptions about human behavior built into them that clearly makes it unsuitable for nonhuman drivers.
It never occurred to me that the AIs driving these vehicles should be tested, but that makes absolute sense.
I wonder what tests these systems must pass before allowed on the road. I remind that it's important that the vehicles do better than human drivers and sadly that's a pretty low bar. I hope we aren't evaluating this primarily by counting deaths, injuries and property damage and comparing between the two.
It's funny it was afyer all just Waymo code dump, NVIDIA demo, and everyone else desperately trying and failing to make some departures from it(other than MobilEye?)
Twenty years ago there were no electric cars. Fifteen years ago the idea of a mass-market electric vehicle was considered a joke. Ten years ago people talked about the chicken-and-egg problem of cars and chargers and how you can't have one without the other, and no one was building chargers. Today, Tesla has some of the best selling vehicles worldwide, as an EV only company; other companies are adopting their standard on charging; the idea of them building a charger network, originally seen as completely ridiculous, has ended up as the premier way to charge your vehicle and is being opened to other companies' cars. I'd be tempted to guess that they've had a larger impact on reducing emissions than any other single company on the planet.
Twenty years ago, private spaceflight was considered impossible; this was the domain of governments alone, and the various x-prizes and the like had done nothing to change this matter. Fifteen years ago, reusable rockets was considered impossible (NASA and their contractors can't even do it, how could anyone else?) Ten years ago, SpaceX cut the costs of delivering stuff to space by what, 10x? 20x? Now they're casually demolishing all players in the satellite-ISP space single-handedly, and are looking at reducing delivery costs another 5-10x; the EU's current rockets and space orgs are collapsing under the weight of their lack of competitiveness.
I don't have to agree with the man's recent decisions around Twitter or politics to credit his past accomplishments, many of which were promised years before they became reality, and it boggles me to see all the people that are incapable of doing so.
Most Jewish people in the US are Democrats, and it's pretty much been the party platform the last ten years to push "hatred against whites" and to immigrate "hordes of minorities."
Apparently the latter is a conspiracy theory against Jewish people as a whole, though I've certainly never heard it. Perhaps Elon hadn't either. To call him affirmatively replying to that tweet antisemitic is a huge stretch. Should someone in his position be commenting on stuff like that? Probably not, but I think it's insane it's as big of a deal that it is.
For what it's worth I just unsubscribed to a particular Patreon because the guy running it was clearly pro-Hamas/anti-Israel from what he's been posting on Twitter. I'm in no way an antisemite myself.
To your comment as whole - no, I was referring to perhaps...5 years ago?
I need to remember the name of the book where I read that, but statistically the people who buy Teslas produce more global emissions than others, simply on the fact that manufacturing a Tesla vehicle produces nearly 2x emissions as manufacturing an equivalent ICE, and the class of people who can easily afford Teslas engage in a lot of environmentally damaging activities like excess consumerism, lots of traveling, etc. versus the people who can only afford to keep using their ICE car instead of throwing them away to buy a Tesla.
It's one of the reasons why despite the boom in EV adoption, global emissions are also skyrocketing with them. So EVs alone won't save this planet. In fact, according to that book, if we were to replace all ICE cars on the road with EVs, the manufacturing, materials and the mining necessarily to extract all the materials necessary for that to happen, would cause such an increase in emissions that would be catastrophicall to the planet.
So the solution is actually less cars for earth, not more cars except Teslas instead of ICEs. It was an interesting analysys that I need to find.
Self driving cars in the form of fleets of autonomous robo-taxis that you hail on demand, would solve that "too many cars on earth" problem since most cars stay idly parked all day anyway, but we're very far away from actually achieve that level.
And those extra emissions are paid for before the car reaches half its lifetime.
The point was that ~~EV~~ train manufacturing is definetly not free of emissions.
Buying an EV instead of an ICE reduces net emissions. Riding the bus instead of buying an EV reduces emissions more, but it is still not zero. Should more people ride the bus? Sure! But I'm not in charge of my city's planning, I can't make them buy more busses and hire more drivers.
I bought an EV when a family member was needing a cheap car, so I sold them my old cheap ICE. I was probably going to buy a new car anyways, so its not like I went out and crushed my ICE and dropped it in the lake and then suddenly willed a new EV into existence creating new emissions. A car was going to get built, I was going to buy one.
> But I'm not in charge of my city's planning, I can't make them buy more busses and hire more drivers.
Sure, but we do live in democracies no? Where we have the power to change those in charge to others who will if our goals if we desire our goals to be betters public services including transportation.
Simply starting with a defeatist mindset of "nothing will ever change from our leadership so it's every man for himself to better his own situation, I got mine you go get yours" will never work for the environment. It does work for selling a lot of cars though.
If we can't change anything collectively, what about those living in less democratic countries?
If the answer is "buy a used car", well, we'll quickly run out of those.
I've seen this play out lots of times.
Public transportation only improves when the vast majority of citizen depend on it because then have skin in the game, otherwise it gets shelved and forgotten and stays run down as a charity to the poor.
Pointing fingers and shaming groups of people we don't anecdotally like for one brand or another, is not how statistics work. Statistically rich westerners, whether they drive a tesla or a Mercedes, all engage in environmentally destructive activities through excess consumerism and lavish lifestyles than less well off people.
For example according to that book, statistically rich westerners, the tesla owning kind, tend to have bigger single family houses, buy a lot more stuff they don't need, throw away more food, travel and fly a lot more or go golfing or have heated swimming pools. Less well off people on the other hand can't afford eco-friendly Teslas but they also have much more eco friendly hobbies, travel less, live in denser housing, consume less stuff, therefore tend to be less environmentally damaging than the Tesla owning class by simple fact that they're poorer.
>but feel free to inhale the particulate emissions sitting in traffic if you'd prefer.
You're pointing out a "false choice". Yes, it would be nice that we all switch overnight to EVs, but the truth is that the global emissions of replacing all the existing cars on earth and replacing their numbers 1:1 with EVs is not feasible at that scale without destroying the planet in the process of mining and manufacturing at that scale needed to achieve the conversion.
The secret to the environment is less cars overall, not more cars but EVs. Therefore less consumerism in general. But tell rich westerners they need to stop owning/using cars see how that goes.
Hence the ever increasing global polluting despite the growing number of EVs.
However, since rich westerners did the politically realistic move of increasing emissions standards and EV's, we are actually reducing our emissions[0], instead of doing nothing but hand-wringing that we "should" have less cars and get nowhere because 3/4 of the US won't have it.
Per capita US emissions are down 25% since 2007 and peaked back in 1979.
We help the environment by finding the wins whereever we can politically. If you get too idealistic, you halt progress because the rest of westerners aren't in for the ride.
Because to me it seems emissions went down to tougher broader environmental policies, especially on coal, rather than just people buying EVs now, as ICE cars were only 9% share of emissions anyway, yet people tend act as if they're the main culprit for it.
I imagine very few people are taking their older ICE and crushing it and shoving it into a landfill once they buy a Tesla.
Buying the Tesla is (usually) offsetting the sale of a different new ICE and selling/trading in the old care to add to the used market, not demanding a new car that would have never been built to come into existence and then destroying the old car.
This line of argument has nothing to do with Tesla or EVs. If you aim to make the case that the wealthy produce more emissions, that seems fairly clear, whether EVs exist or not.
> It's one of the reasons why despite the boom in EV adoption, global emissions are also skyrocketing with them. So EVs alone won't save this planet. In fact, according to that book, if we were to replace all ICE cars on the road with EVs, the manufacturing, materials and the mining necessarily to extract all the materials necessary for that to happen, would cause such an increase in emissions that would be catastrophicall to the planet.
> So the solution is actually less cars for earth, not more cars except Teslas instead of ICEs. It was an interesting analysys that I need to find.
Obviously, things would be better if people just stopped buying cars. Obviously, things would be better if everyone had good public transport available and dense development and didn't need cars. Obviously, things would be better if just less people existed and didn't need to be transported. None of this contributes in useful fashion to the here and now. I can go to my city development planning meetings and argue against car-oriented development and seas of parking until I'm blue in the face (and I do), but cities take generations to change. Cars, on the other hand, can be largely cycled out within a generation and a half. And like it or not, cars aren't going anywhere.
But there were electric cars before the model T.
IIRC the idea of reusable rockets was mostly considered possible just risky and pointless when the only real customers (at the time) tend to have seemingly infinite budgets if you drag it out of them. SpaceX bet on the reusable rockets and the gamble paid off handsomely, and showed there is quite a market for things when you bring the launch costs down 5-10x. It turns out a lot of people would like to put things in to space at that price point.
A lot of these big achievements are largely having the willingness to throw capital at a potentially risky but otherwise technically seemingly possible ideas. Self-driving cars are not in the same category as EV chargers when it comes to a "can we do this" kind of thing.
Uhm...what?
First of all there were Chevy EV1s then. In fact it was Chevy cancelling the EV1 in 2003 that was part of the inspiration for Tesla's founders to found Tesla. (Musk was not yet involved).
Second there was AC Propulsion's tzero. The Tesla founders had a test drive in one before they founded Tesla and urged AC Propulsion to start manufacturing them, but AC Propulsion wanted to be a designer and licensor of electric car technology for others rather than a car maker. The Tesla founders than founded Tesla.
Musk came into the picture when he test drove a tzero, and also urged AC Propulsion to go into manufacturing. Instead AC Propulsion directed him to to Tesla where he became an investor and then chairman.
What Tesla did was become the first modern car company to go all in on electric cars that were intended for ordinary car buyers (although it first it was just luxury sports car buyers). I think they were also the first to go with lithium batteries.
If Musk had been a "good leftie" he would have been carried on arms both here as well as in the media because of his work in furthering the electrification of transport. Bezos does play that role and as such is not targeted like Musk is even though his companies have had a detrimental effect on retail employment opportunities, even though working conditions in Amazon warehouses are know to be less than optimal. It is also why there is no criticism on Bezos buying the Washington Post since he runs it as a "good leftie" would.
Here's a thought experiment: Imagine Musk buying the New York Times and changing the editorial board to use it as a counterweight to Bezos' WaPo. He'd bring back the likes of Bari Weiss, get rid of the identity politics crowd and see to it that the clear left-wing bias the NYT has been burdened with is would be undone. All the news that's fit to print would again be the motto. No more 'DEI' nonsense, nothing of the sort, back to basics. Now there would be the Bezos-owned WaPo doing the whole DEI/oppressor/oppressed/identity-politics thing versus a NYT which attempted to critically [1] report on the state of the city/country/world. What would the reaction to such a take-over be, who would get the support of the like of CNN/MSNBC, who would you support?
[1] as in the 1980's dictionary sense of the word 'critically', not as in the 'critical theory' deconstruction-type criticism.
Just like developing a product, the first 80-90% of the functionality is easy and is done in 10-20% of the time. But that last 10-20%... oof. Sometimes that's when you discover that the problem may not even be fully solvable and you either cut back the requirements, give up, or slowly go insolvent.
I view the AI hype timeline a little differently. Things really took off with the "Attention is All You Need". Since there there have been a ton of concrete accomplishments in areas that were kind of stuck: image recognition, the game of Go, protein folding, etc. I'm still very bullish on what can be accomplished as current AI techniques are applied to more problems.
The 5% remaining might just be the most difficult problem mankind has ever tackled technologically.
These companies need to go in cities with bigger roads and less crowded streets. They also need to incentivize citizens to ride these cars for training.
Here are places I can see this tech gaining a foothold and improving its data
- Walt Disney World Resort. I hate waiting for busses or taking the transit options. Disney had (don’t know if it’s still there) a brief collab with Lyft having earmarked cars for ride share between the parks. Make the cars some sort of upgrade package for visitors. A Cruise/Disney partnership would definitely benefit testing those new Cruise cars they put on hold.
- Las Vegas. Big open roads. Lots of people and construction. Plenty of room to move around. Lots of data to learn.
Over time they can’t go towards more condensed roads and traffic.
Ideally I could see Disney being a partner with Cruise to get those Cruise cars on the road for real world testing.
This is a far easier problem to solve than self driving vehicles
The technology that has been put out so far is amazing and the engineers behind it should be very proud of that work.
But there are too many "but if..."s for driverless to be mainstream just yet.
We need a system that can really think like a person can. Something that has very high-level abstractions and understands context, situational state and human behavior.
It's not a problem that can be just solved by slapping more and higher-resolutions sensor packages onto the car, though that will absolutely be necessary. (And until that becomes MUCH cheaper, that's a major roadblock as well)
How accurate do you think your tech predictions from the early '90s about today were?
This actively hurts funding when promises fail!
Who's going to want to invest in self-driving cars when their "full self driving" Tesla that they bought 5 years ago will never be "full self driving" while they own it?
Who's going to believe the promise that self-driving-cars are safer when Cruise's and Uber's cars hit passengers? It's clear that their cars weren't refined as much as their leadership believed.
Self-driving cars are a problem that might take another decade or two to perfect; so acting like they are going to come out in the next few years is counter-productive.
Sure, in the sense that's the idea behind lots of things. But there is an awful lot of handwaving about how to concretely get there.
The overall goal of using self-driving technology to reduce human deaths and injuries is clearly admirable. What's less clear is a) if it is technologically achievable and b) if any of the current efforts are scaled appropriately to get there over a decade or two.
The early self-driving hype was pretty depressing to behold. It was obviously an Apollo sized or bigger effort, and people were actually getting traction with the laughable idea that we'd "get there" in 2-3 years with a few 10k of engineers (or less, I'm estimating conservatively).
Of the current efforts, the only that I have heard very positive things about is Waymo.
Cars are safer than they've ever been; looking at some statistics, we have halved the fatality rates since 1980. So I'm not exactly sure claiming "it's safer" is really much of a selling point.
GP is correct that the money has never really made sense, other than in the context of logistics (automating semi trucks maybe?). Given that even the safest of transport modes (say, the metro in London or Paris) still needs to have a conductor (for potential emergencies), it's hard to see the fully driverless dream ever materializing.
Not if you are outside the car. Pedestrian and cyclist fatalities have increased steadily in the US over the past ten years [0] due to ever increasing car sizes and speeds [1].
[0] https://www.cdc.gov/injury/wisqars/fatal/trends.html
[1] https://www.cdc.gov/transportationsafety/pedestrian_safety/i...
Cars are inherently unsafe but we refuse to disengage from this failed model despite the solutions (ie. public transit) being right there for decades and decades.
Self driving public transport in contrast is a solved problem. Vancouver's driverless Skytrain has been serving customers since the early 1980s (as you say, requiring a driver only in the rare exception of some emergency or technical malfunction).
London's system is particularly old and deep, and lacks space for emergency walkways.
There are certainly exceptions, often well publicized by the companies, of drivers who figure out how to exploit the algorithms to make decent money. But when too many figure this out, the companies adjust the algorithms to contain costs, leading right back to the above.
Once upon a time there was probably someone who said it was silly to spend a couple million dollars to do the job of a secretary making 50 cents per hour, but now here we are, typing these comments with computers.
Self driving is not even claiming it would have much of an effect of the world. A few points more efficient, cutting a bit of costs - not changing the human experience at it's core.
Right now the best way to reduce car fatalities is reducing the number of cars by eliminating single family zoning mandates, building safe cycling infrastructure, improving public transit frequency, allowing mixed-use buildings and walkable neighborhoods.
Boring stuff, yes, but unlike self-driving cars, it is something we already know how to do.
Instead we get companies doing R&D of unproven technologies on public roads because no one wants to build more rail.
For some eye-popping numbers, check out the automotive deaths/100,000 people for the US (12.9) vs. Denmark (3.4)[1]. Denmark doesn't have some magical AI that makes all their cars safe, just better infrastructure.
[1] https://en.wikipedia.org/wiki/List_of_countries_by_traffic-r...
For example, Seattle area just spent tens of billions on light rail expansion, then built the outlying stations next to the freeway. Half the walkable distance is wasted. Then they build parking garages and zero amenities. They're surrounded by dangerous parking lot and multi-lane roads. It's not designed to actually be used by enough volume to matter! And even if people did come (which they won't) it can't handle more than a small portion of the existing traffic because it's too slow and can't be expanded and can't have express routes.
This is the best we get, for tens of billions!
If US drivers had to meet the same standards as Danish drivers in order to get a license, our roads would be a lot safer.
It's frustrating to see traffic fatalities trotted out as a "the ends justify the means" justification for self-driving R&D budgets and practices as if there are no other solutions to the problem.
It probably wouldn't. A VERY high proportion of the most risky drivers on US roads are already driving illegally. For example, nearly 1/3 of drivers in Mississippi are driving around without insurance.
https://www.iii.org/fact-statistic/facts-statistics-uninsure...
If people are given a choice between driving illegally and going without basic necessities, they choose the former.
A lot of people end up in this situation because they are bad drivers with bad records, and can't afford the insurance, so they drive illegally. Functionally this is no different than if they weren't able to get a drivers license. They are still on the road.
Lax licensing is not the cause of bad drivers in the US. Both the bad drivers and the lax licensing is caused by a lack of transportation availability.
Self driving vehicles could be the innovation that makes public transit usable in non-dense areas. We could do dynamic routing based on the demand of the riders.
There are other solutions to the problem - I'm sure you can come up with a few that are more satisfactory than changing all cars to being computer controlled.
It would also massively increase traffic compared to taking the school bus or walking, and compared to parking at work.
- Car ownership could end up eliminated except for those who really need dedicated work vehicles. - Goods and delivery costs plummet. - Cross-state/country trips eliminate short/medium haul flights (wake up at your destination). - People get thousands of hours back from commuting (in theory).
So many other knock-on effects similar to how smart phones changed every aspect of our lives.
Also, highways with human drivers are highly wasteful of material and space. A smoothly flowing highway requires 80%+ of the space to be empty in order to account for the high latency in human decision making. This is expensive.
This thread, and this entire topic, is a discussion about things that currently don't exist and aren't currently possible. We are all aware that self driving vehicles are future technology, not an existing one.
If you think that this upside potential is not very well defined, entirely hypothetical, and isn't even clear that it will even happen, you're entirely right. Investors call this "risk".
Either way, it was envisioned from the beginning that digital computers could replace human computers along with a lot of other office disciplines. There once was a big "computers are going to make office jobs disappear" scare.
That's a lot of ifs, so I'm not at all saying its right around the corner, but it's within the set of possible outcomes. V12 is looking quite promising, but then again so was V11 and V10
Isn't this true of all the other players as well.
If we had truly capable self-driving cars, we could probably get rid of >50% of all cars on the road. Most cars spend upwards of 90% of their existence parked and doing nothing, waiting for their owners to need them. If cars could be operating full-time 24/7, ready to be dispatched anywhere as needed, it would mean the destruction of vast amounts of transportation demand.
I've seen this argument before but it doesn't make sense to me. In some subset of situations, yes, this would be the case. But 8-9a or 5-6p there is a surge of demand in every major metro in America.
Developing self driving cars is not a fruitful endeavor IMO. We need to push alternative modes of transportation (transit, e-bikes, etc) which are way, way more efficient and cost effective.
The US is incapable of building cost-effective heavy transit though. Just building a few miles of subway in the LA metro area, for example, costs on the order of double digit billions and takes a decade plus. In huge areas of the country where it's snowy in the winter, e-bikes are a non-starter.
I'm not saying every self-driving vehicle needs to be a single occupancy car. We could have fleets of self-driving minivans that are somewhere between buses and cars that can move the dramatically more people than solo drivers. Even self-driving buses would be a win, a very big chunk of the operating expenses of public bus fleets is the payroll for drivers.
Then don't build trains - busses work fine.
> In huge areas of the country where it's snowy in the winter, e-bikes are a non-starter
I live in Denver - it's fine.
I don't hear feasibility concerns in your arguments - I hear convenience concerns. How I interpret this is that you, and many other people in the world, prioritize convenience above all else.
I understand why, but we must come to the realization that we're taking the side of convenience over everything else: decrease in safety, increased electrical consumption, increased congestion, increased pollution, increased noise, etc.
> we could probably get rid of >50% of all cars on the road. ... If cars could be operating full-time 24/7, ready to be dispatched anywhere as needed
Wow, you really don't get it, do you?
No one wants to get in their car to find it a mess from an inconsiderate customer. No want wants to deal with the extra wear and tear. Most people drive during rush hour, so assuming you need your car to get you to and from work, your car won't be doing chores for you when you don't need it.
Really, robotaxis may cut down on lower-use cars. IE, if you live in the city, and don't use your car everyday, you might prefer to use a robotaxi. It might even be "worth it" if you live where parking is sparse. (Or not, because your self-driving car could park itself 15 minutes away.) Robotaxis might make it so that multi-car households can cut down on the cars in their household.
But turning every car owner into an entrepreneur who rents out their car when they don't need it? You're letting your imagination run away!
And here's the problem: So many people are letting their imaginations run away regarding self-driving cars and robotaxis that it's actively hurting development. The pie-in-the-sky promises like yours never come true; so the people who need to believe in self-driving-cars to keep investing loose faith and cut funding.
I'm glad we agree on this, individuals renting out their own car is a wacky idea.
I picture it more as a subscription service. Pay x dollars per month for an on-demand vehicle, or just pay per use.
But the (uhm) unpredictability of what you get may make this less desirable for everyday use:
If you have a family, you may find it easier (and cheaper) to own your own vehicle that your stuff can be left in.
If you live in a rural area, the customer density may make it uneconomic to rely on a robotaxi.
If you're of the wide middle class you just might like the fact that the space that you spend 1-2 hours a day in is yours, and won't be defiled by someone else.
You live near the beach and get tired of cars randomly showing up filled with beach sand or wet seats.
---
But I want to stress a very important mindset: The interior of a car is your space. People leave stuff in their cars for all sorts of reasons. You could be doing a certain activity and leave some backup equipment in your car. You could be of a certain profession and always carry certain equipment in case you get a call from a customer... You could be running errands and need to carry different things with you for different stops.
This becomes significantly harder with robotaxis when your car is no longer your private space. Don't assume that robotaxis will suddenly make people change this behavior.
Instead, assume that you can't fully predict how markets will adjust to robotaxis and self-driving cars.
---
Honestly, I think the early markets for self-driving cars will (obviously) be robotaxis that are cheaper than current taxis / ubers / lyfts.
But, I don't think that will be the first market. I think the first market will be people who really need self-driving vehicles and have the money to pay for it. They will be able to jump in and smooth over situations when the car can't drive itself. I suspect these are people who have access to money and have some form of impairment that makes paying for self-driving "worth it."