Smart Driving Cars episode 408- AV Regs, Waymo and More
Annotations will appear as you listen
0:00
38:02
Company
NHTSA
NHTSA is a government group that makes sure cars are safe to drive in the U.S. They create rules to keep drivers and passengers safe.
Concept
automated driving systems
Automated driving systems are computer programs and sensors that help a car drive itself or assist the driver, making driving easier and safer.
Concept
United Nations global technical regulation
This is a set of rules made by the United Nations to help countries agree on how to make and use new car technologies safely, like self-driving cars.
Brand
Waymo
Waymo is a company that makes cars that can drive themselves. They use computers and sensors to safely take people places without a driver.
Term
remote assistance agents
Remote assistance agents are people who watch over self-driving cars from far away and help if the car gets confused or stuck.
Concept
AV fleets
AV fleets are groups of self-driving cars that work together as a business, like a taxi company but without drivers. They use technology to drive themselves and give rides or deliver things.
Concept
long haul business
Long haul trucking means moving cargo over very long distances, like from one state to another. It needs careful planning because drivers can only work certain hours.
Term
hours of service rules
Hours of service rules tell truck drivers how long they can drive and when they must take breaks to stay safe on the road. This helps prevent accidents caused by tired drivers.
Concept
automation in trucking
Automation means using machines or computers to help or replace drivers in trucks. This can make trucking safer and keep trucks moving more efficiently.
Term
hump yard
A hump yard is like a big train parking lot where train cars are pushed over a small hill and then roll down to different tracks to be sorted and put together into new trains.
LIVE
Welcome back to the smart driving cars podcast. Thank you for spending time with us. I'm Fred
Fishkin along with the faculty chair of autonomous vehicle engineering at Princeton
University, Alan Kornhouser. Hi, Alan. Hey, good morning, Fred. Good morning, marching into March
here. Yes, it's midterm week. Already? Okay. Oh my goodness. Half way through the second half.
That means we're three quarters way through the year. Oh my goodness. And then spring break, I guess,
right? Yeah, I hope so. In our last episode, we focused on your comments to NHTSA on the
proposal for a new United Nations global technical regulation on automated driving
systems. And you're linking in your latest newsletter to the range of comments that
have been submitted and kind of urging people to take a look. Yes, it is. It is a range. I think
there have been at least last time I looked 53 comments. I think they've extended the comment
period from whatever, from February 23rd to March 10th or something like that.
So that I think there's still time for people to put in the comment that you should comment.
Why not? I mean, if I can comment, you can comment. I think more comments in some sense,
the better. And what I did was, you know, just to help everybody, especially, I mean,
you know, one of the reasons I put out this E letter is sort of it's a convenient way for me to
also bring together things that I think that are important for my students to learn along the way.
And I guess the list is what I would help my students to read. Sorry if I didn't include yours.
But of course, there's a link to all of them for them to read others. But I think that these are
not that they're not all solid comments in there. But I thought that these were
making good points. This is still a challenging issue in terms of the public oversight
of transportation, major public oversight is has to deal with safety.
Certainly with when consumers are involved, because as individuals, it's very difficult for us to
as an individual to have an influence and make sure things are good, that if we assemble ourselves
through government, we then can have a better opportunity to make to
have a higher probability that things will be good. So, you know, help out here as an individual.
Most of the entities are presented by corporations or other
groups that have somewhat of a vested interest. But the reason they have a vested interest is
that they at least believe this is something good to do. So, since they think it's something good
to do, they of course should have an opportunity to make sure that that thing that they give is
really good. Let's make it so it's really good. So, great. And we don't know how long the process
is going to be until this gets acted upon, I suppose. Yes, and hopefully that there are at
least some jurisdictions that will say, hey, go ahead and do this while we're getting all of our
ducks in order and in line so that everybody or a greater portion of the folks can have that
available. And certainly, you know, California has stepped out and said, hey, in San Francisco,
in the Bay Area, in LA, in San Diego, and who knows where else. Sure, go do it. Looks like the
folks in Florida have done the same thing. Looks like the folks in Georgia have done the same thing.
I sort of wish the folks in New Jersey would do the same thing.
They've got a lot of things to do. So, whether or not this is going to percolate to the top,
anytime soon, who knows. But at least there are some places that said, my goodness,
the risk-reward trade-off here is, there's a lot more reward than there is risk.
So, those places are saying, hey, go for it, I guess.
Timothy Lee has a piece that you include here in the newsletter headlined.
Waymo just revealed a crucial statistic for scaling its technology. And that is, at any given time,
it has about 70 remote assistance agents on duty worldwide. And that's for 3,000 Waymo vehicles.
And you have a comment here. You're agreeing with Alex Roy that this validates AV fleets as a
business model and indicates Waymo Driver is at least 40x safer than humans? Go into that for us
and explain that. Well, I think I threw in the 40x, but Alex Roy did provide Timothy with that
statement. And it is a very substantive statement, at least to me, because this isn't
whatever number of vehicles operating on some laboratory under some strict tests,
d-d-d-d-da, whatever. And so on, which, of course, part of the comments associated with
what kind of tests, what are the tests that NITS is going to put up, the hurdles that such a thing
has to pass over, such that it can be deemed safe. And of course, it'll never be safe because
nothing is safe. That's in the limit as whatever time goes to infinity. That's an aspiration.
There's risk in everything. And I've said it many times here. I say, my students all the time,
the beauty about the future is that it has risks and uncertainty associated with it.
If it didn't, it'd be boring. I mean, in some sense, it really wouldn't be worth living. I mean,
yikes. But that's what makes it not only exciting, it makes it difficult, it makes it
challenging, d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d-d. And I guess it keeps our juices flowing. Otherwise,
what do you do? Sit there. So yes, there are risks. The issue is, what's the relationship
between the reward and the risk? And that's the beauty of our department in operations,
research, and financial engineering. I mean, as a fundamental, the students are involved in
trading off risk and reward. And I guess what you would like is at least the rewards, at least a
bit. The question is, with all this, my comments have focused on, yes, all the other, not all,
I guess, but essentially a large part of all the other comments deal with the risk part of this
and trying to reduce the risk. But at least philosophically and politically and emotionally,
I feel that we should be trading off risk and reward. You really can't talk about risk unless
you also talk about reward. And of course, you should be comparing apples and apples and
making sure that you're kind of making the right comparison. Because in the end, you're going to,
you know, it's sort of a rank order. Which one's 40, 51, and which one's 49?
And of course, you'd like it to not be 51, 49. You'd like it to be obvious who
rewards much greater risk or the risk is much worse than the reward.
Then, you know, not to do it makes decision making straightforward and trivial.
And I think that the key to what Timothy points out, which was a real embarrassment for
Waymo in the hearings, but it's then substantive to the safety question,
is the oversight control center centers that they have, because these aren't autonomous vehicles,
even though they're called autonomous vehicles. They're told what to do.
They're told, you know, basically nowhere later than go to. Now, how they happen to do that and so
on may have some autonomy associated with them as to how they execute that. But there's oversight on
that. And just like the existing system with people out there, there's a whole host, there's a support
structure in case, oh my good things didn't go right. I got a flat tire. I got stuck.
I ran out of electricity fuel. Who knows, you know, what happens in the future. You get, you know,
so you really need the oversight. Now you might put, you know, an AI agent out there to do the overs.
You're going to have to have people overseeing the AI agent anyway, at least I believe.
Are we really going to go to war and just say, or do whatever the heck you want? I mean, Napoleon
didn't do that. No general has ever done that. No leader has ever done that. So, I mean, it's almost
silly to even think about it. We are not going to relinquish our oversight on machines and technology.
Not as a human being. We just aren't, at least I don't know, sure you can, whatever.
Well, the numbers are significant here right, Alan, because the numbers are significant. The key
is, even though Waymo said, oh my goodness, we're employing these people in the Philippines,
some of them, which was the embarrassment part. The substantive part of that was that they're only 70.
That means that there are 70 humans providing mobility with,
that's provided and delivered with 3,000 vehicles that results in crashes that for all
intensive purposes are essentially the same, likely even less than 3,000, pick a random 3,000
anywhere in whatever the country, pick or whatever, people providing mobility
for driving. 70 versus 3,000, same risk.
70 providing reward of 3,000 vehicles, 3,000 providing mobility reward
of 3,000 vehicles, same risk. So, if you compare that,
the risk involved in the 70 is 140, 147. I mean, if you want to do the arithmetic,
the 18 decimal places on your welcome, it's 40x. Essentially, now, if you want to get the details
of intensity, and sharpen your pencil, you'll get a couple of decimal places on the 40,
but the four out there with another digit before you hit the decimal point,
that sucker is pretty darn solid. So, if you are in a situation in which you can be the underdog,
underdog is being down, you're a real underdog if you're down 40x.
And if you're trying to make a decision of which side do I want to be on, where basically it's,
is there any disagreement that you would want to do the 40, not the one?
Which means these are safer and providing the same mobility.
By 40x, I don't know. That's the way I, fundamentally in my mind.
And if, and I think that if you go there and sharpen your pencil,
you're still going to be at about 40. You aren't going to be at 4.
You aren't going to be at .4. Chances are that are, you know, 10 to the minus many.
Oh my goodness. I mean, hit the nail on the head here.
That's really what I think Timothy Lee and, and Alex through his comment and so on have done here.
I said, as Alex said, game over, you know, this looks like this. Let's scale it.
Boom. So that's why, that's why I probably, if it wasn't for all the good testimony and
all the work that people did putting the testimony, I would have put that as the lead,
as the lead article here. But, you know, the amount of work and amount of effort and amount
of diligence that went into the, of the people that, that put in those testimonies that are
worth reading there. They aren't, they aren't clickbait for who knows what of, you know,
most of the articles that appear in your Instagram or, I mean, it's just
debilitating. Next. Terrific. You have a link in the newsletter to a piece
in American affairs, a headline. Why are American passenger trains slow?
Well, I mean, I, that came across and there are so, well, I shouldn't say that. I love Timothy's
article, you know, and, and understanding AI and, and what did I love this one? And, and I guess
a lot of people that I have a lot of respect for also liked it because I guess I've played with,
you know, I'd like to say that I, I basically dealt with Conrail from its conception to its
internment. So, you know, I, I, I did choo choos at one point in my life and so I got choo choo
back here. I love choo choos and then I just thought, I just thought, what a wonderful article that
just made it perfectly clear why. And it's a wonderful story. And, and I won't, I won't,
I won't give you the punchline. It's, it's, it is so well worth rereading. I encourage you.
So, I won't give the punchline. You'll, you'll, you'll hit you over that. Wow. If you don't say
wow, then write to me and say, man, this is James, what the hell are you recommending? I mean,
you're really screwy. I don't know. Let me know. I just, I, I didn't know that I'm just lost. I mean,
I might finally losing it. Some people would say you don't want, never mind.
Terrific. Then you have a link to an article with the headline,
the future of long haul isn't coming. It is here.
Yeah. So Craig Finder, who is a former ALK employee, love him dearly,
Princeton gold tender when he was a student here, has been in the industry for a very long time,
sent me this because he was at a conference in LA in Las Vegas in which torque was displaying
their truck. And apparently there was a long line and he thought, oh my goodness, you know,
these guys, you know what they are. Yeah. There is really, it's really a big hit. And so I, you
know, I immediately wrote back, well, why was it a big hit? I don't know. And then of course,
it became obvious from their, from their, from their very simple ad that they had,
which relates to the previous article. And I don't think I'm giving away the
punchline of the previous article, but it says, keep the trucks moving.
And if you look at the transportation or logistics challenge, it is about flow,
not speed, flow, flow is really speed over time.
And it just got me to reflect on what is the fundamental value of, of automation of the driving
function. It is a little bit to be able to remove the driver and save the cost of the driver.
Because then that, that's your productivity numbers of that asset and, you know,
revenue or net revenue per whatever they do improve, because you can remove
that, that labor cost in that. But actually, it's not really that.
Okay. It's that you can keep your trucks moving.
So if you're in the long haul business and you have
a driver, then the ours of service rules basically say that this asset that you have
of a truck in its cargo can go fast for only part of the time, whatever it is,
10 hours a day, whatever the details are, that thing. But a day is 24 hours long.
The rest of the time, it has to rest or whatever.
So it's ESOP's fable.
The tortoise and hare.
And in some sense, the opportunity that the automation gives you
to have that asset flow 24, 23 or most of the 24 hours in a day, as opposed to just 10.
Wow.
A long time ago, I was, I was given a tour of the Russian railway system.
The head of the Russian railway system was right after the wall fell.
And they were showing me their, their control room in which they
oversaw part of the railroad operation. And then I was taken out to one of their major hump yards
and proudly shown the hump yard in which there were all kinds of cars,
wagons, all of the hump yard. And I really wasn't trying to be cute or whatever.
And hopefully it didn't, it got translated properly. Otherwise I probably would have been shot
right there or something. I said, I'd be more impressed
if there were no cars here.
And I actually meant it.
Because I think even at that point in time, I realized that if there are cars here,
that means they're not moving anywhere. They're standing here.
And if we're talking about transportation, especially logistics,
it's all about flow. It's all about moving the stuff.
I mean, if I go into a warehouse and there's a lot of stuff in the warehouse,
I'll make the same comment. There should be very little stuff in the warehouse.
There should be stuff coming in and going out, tough because I'm going and going out,
stuff coming in and going out. Stuff coming in and going out. It should be stuff coming in,
get stuck there, stuff going out, then stuff coming in, getting stuck there,
stuff going on. It's all about flow. Because otherwise you have to have inventory.
Inventory is just overhead. You're stuck with it. It's doing no good. It's a
waitin. You're paying the bank.
You got to flow it. And if you're really in logistics, it's the thing that has to
be in the back of your mind. All that is the flow of this stuff. Oh my goodness. If
it just reminded me of all this yesterday in class with Spencer Lucian of From You
Flowers where he's talking about flower business and what needs to happen on Valentine's Day
and Mother's Day and so on and whatever. And we talked about, I mentioned our Malini brothers who
basically run tandem drivers. Has for years. I think I hope they still do.
From Miami Airport to New York. So they can flow those flowers in 24 hours.
Tandem. One sleeps while one drives. So they can get the stuff there before it wilts.
Because damn, there, if you're not flowing those flowers, you lose them.
And that's not just true in a flower business. It's true in anything.
And all of a sudden now you have a machine that can flow that.
And you say to yourself, I've been paying a truck driver in long-haul
trucking. Not for 10 hours of operation actually. Because that trucker, she's given me 24.
She's given me her life that whole day to get the stuff from LA to Chicago or whatever.
Now, I got away with just paying her for 10 and said, hey, in the other, you can do whatever you
want. How are you going to do in a truck stop? Making sure that you don't violate your hours of
service. Now with automation, I can ask no more from you that I've asked from before.
But I double my productivity because I move and instead of stomping, I just keep moving.
Whoa. Hey, maybe I could get the whole thing out of you by, you know, not having you in there at all.
My goodness, I probably need to because there's all sorts of other information. There are all
sorts of other things you could do for me while you're there, plus you're taking care of them.
And if something really happens, you can fix the stuff that I anticipate. And boy, you earn your keep
on demand when I need you as opposed to, and then, and then,
you know, whoa, that's what those folks must have seen.
That's what they, if they saw that.
Whoa. Whoa. If torque, that's what torque is selling. Great. Aurora, if you're selling that,
great. Sorry, I didn't pay attention to what you were doing before. It's fantastic.
And now, you know, put some great accommodations in that cab so that, you know, the person can have,
you know, some suede v. Well, the machine is doing keep me between two white lines and don't
crash. Keep me between two white lines and don't crash. Keep me. Could you imagine what somebody
has to put in their body to be able to do that for nine hours and not die?
Perfect thing for a machine.
How do you like that one? Very good. I guess we can make that part of the midterms.
I don't know. I mean, I'm just, I noticed the students get it. They get it.
You know, you know, why is Fermi and Flyers been so, why? Because, boy, they get stuff in motion.
Who's really important? Those things.
But I guess everybody knows that that's in the business. So I'm not telling anybody
anything they didn't know. It's just, no, but really the, the risk is not, is not all that,
you know, it's like 40x on your side. Man, we as a society should take advantage of that because
the productivity improvement and the logistics leads the, leads the winds on the, on the, on the
asset utilization side, which means that the cost of goods, of good sold reduces, which means you
can lower the price or make more money to do with competition. And everybody gets the, everybody
wins. Wow. He create more jobs. Seems intuitively obvious that the most casual observer to me,
but of course I'm Mr. Waco.
Well, we know, we know you've been tied to the trucking business and the trains. And now let's
talk about space. What am I supposed to go to the full gamut, right? NASA news, changing the
Artemis plan, shifting the first lunar landing of the Artemis era to Artemis four. You know,
I think I read the Elon Musk is shifting his goals to the moon rather than Mars too. That's
every once in a while you've got to pivot. There's nothing wrong in pivoting as you learn.
As you learn and you say, you know, maybe I should do this. I mean, the interesting thing
and, and, and, and as he makes, he makes the argument, I mean, it's absolutely interesting
looking back at the launch cadence, you know, going back almost to Vanguard.
Certainly through the whole Gemini Apollo projects and so on, the launch cadence
by NASA was, you know, every three months, launch cadence on Artemis is every three years.
And so the question is, you know, as he puts it, you know, if we lost muscle memory in between,
because we're not, we're not launching it, we're not, you expect to learn things out of all these
cadences and you don't expect to necessarily solve everything. It'd be nice. Hey, one shot,
done. Hey, yeah, right. I mean, even with AI, we're not that good and there's no AI anyway,
so, you know, we're not that good. In fact, we learn, we tend to learn more by failing than we do
by success. So in some sense, we have to build failure into the system because that's how we're
going to, that's the tuition, that's the tuition we pay. Students on exams learn more from questions
they missed than what they, what they may, of course, they already knew what they got right. So
therefore they're, you know, almost by definition, you're learning only through the mistakes,
otherwise you don't learn anything. That's why I got to make the exam so that they don't,
you know, get them all right. They're going to make sure they get 50% of them, right, just
deal with their heads so they don't think they're total losers, you know, so 50% is a giveaway.
Now, now we start the second half, whoo. And then the last piece, whoo, you got to be really good,
whoo. That's when we have real fun with them. But anyways, and I guess the next one is,
is that I put in there is his interview on CBS. It's just,
boy, it's nice, you know, for someone being that's interviewed as someone that's forthright
and bright and just answers the questions and, and is as opposed to reading some AI-prepared
press release. I mean, it's just, it's just, so I'm so pleased with the leadership of
NASA that Jared got the job. And I guess we should all be happy, whatever taxes are being put in that
one. And please don't look at who's head of Department of War, for instance, Brad. Never
mind. Well, if I get arrested by the ICE, you know why. That's right. I'm a foreigner. So, you
know, I'm done. I'm gone. Alan, you mentioned your students. We're anticipating a compendium
of findings by your students from, I think last semester, right? Yeah, I know. I mean,
they haven't finished it. I keep putting that in there. And then, you know, they haven't finished
it then. That's the next semester. And now it's last semester. And, you know, students,
problem with students is they disappear on you because they then, you know,
have other professors they have to kowtow to. So I may be a loser. I may have to pull that one
at some point. Well, also, in the newsletter, there's a new image at the bottom that people may
notice. Well, yeah, I don't know. I decided you got to put something new in it. So I decided to put
a picture of an old guy giving another old guy a ride. So, you know. That relates to the update
on handy rides, I suppose. Well, I don't know if it's an update on handy rides, but Catherine
Ford will certainly appreciate it since she's been doing that very well for 30 years. I'm just,
you know, learning my way here. And it's amazing how when we get to this point,
how much we need help and just getting, you know, from A to B. And, you know, while driving may have
been fun and easy a while ago. At least for me, it ain't no mo. And at least I have the good fortune
of being able to say, you know, I'm not driving anymore simply because I got a machine there that
in so far has really worked for me. And so, I don't know. And it doesn't
better than I do it. So, and maybe that machine will then start doing these podcasts and these
new casts. And then I don't know what's going to happen. But, you know, we'll have AI do it,
call that, we'll do it.
On that note, Alan, that's going to wrap up this edition. You can find more and the newsletter
at smartdrivingcar.com. You can find my tech reports at textination.com. Thank you again for
taking the time to watch or listen. And please stay safe. Thank you for it. Have a great day, folks.
About this episode
The discussion centers on recent developments in autonomous vehicle (AV) regulation, particularly the global technical regulations proposed by the UN and public comments on them. Alan Kornhauser highlights the importance of balancing risk and reward in AV deployment, using Waymo's operational data to illustrate that their fleet is significantly safer than human drivers. The episode also touches on the role of human oversight in AV operations, the slow progress of regulatory acceptance, and insights into transportation topics like passenger train speeds and long-haul trucking futures. The conversation blends technical analysis with policy and industry perspectives.
Original notes
More responses arrive to NHTSA's request for comments on Global AV Regulation. Waymo validation of business model. Why are American passenger trains slow and the future of long haul. Join Princeton's Alain Kornhauser and co-host Fred Fishkin for episode 408 of Smart Driving Cars and remember to subscribe! And catch us on The Transportation Channel.
Smart Driving Cars Podcast
Fred Fishkin/Alain Kornhauser
0:00
38:02