Hacker News Reader

Comment by KellyCriterion | Hacker News Reader

Very impressive video! What you can clearly see:

If you scale this over the next years, "manual driving is over" :-D

This is just the very very early beginning, like the first seconds after big bang, we haven really started this whole thing: If you put just more ressources here, we will have giant parking houses for just self-driving-cars, like "coming home over night and recharge"

And this technology will come on a mass scale, Im pretty sure - there is nothing that can stop this?

How will this affect transportation jobs?

If you are today a big insurance like Munich Re, and you see already today that self-driving produces already much less accidents (90% or so I read days ago?), and the tech is really new & "not 100% reliable" and you believe that this tech will be rolled out - one day you will start lobbying politicans that manual driving needs to be forbidden, except some rare cases.

Lets come back in 10 years! :-)

View on HN

> If you scale this over the next years, "manual driving is over" :-D

I ride a motorcycle. Unless you make this illegal, then no, "manual driving" is not over anytime soon. Some people actually _like_ driving, although, none of them appear to frequent hacker news.

> there is nothing that can stop this?

We learned the lesson of over automation in aviation very quickly. The best solution is very often, turn off autopilot and autothrottles, then hand fly the plane. You don't want pilots with their heads down in a computer making changes when they can, and should, just fly the plane. It also tends to reduce over all competence and harms safety particularly during equipment failures.

> and you see already today that self-driving produces already much less accidents

The utility of these statistics are harmed by the fact that "self-driving" does not occur in all the circumstances that "manual driving" does today. They're barely past the foundation stage, in a limited operational scope (cabs), and in a limited operational area (where it never snows).

> Lets come back in 10 years! :-)

I think this technology will eventually be prevalent. I think hacker news constantly gets the time lines wrong. You're going to need 25, minimum, and possibly more like 50.

> You don't want pilots with their heads down in a computer making changes when they can, and should, just fly the plane.

There is a famous video on youtube, 'Children of the Magenta Line', an extract from an American Airlines course decades ago.

Well worth the watch: https://www.youtube.com/watch?v=5ESJH1NLMLs

As someone who loves driving, I still use full self driving (FSD) on the Tesla all the time since I got a new model Y Juniper.

And occasionally I switch to manual just because I enjoy it.

So when you talk about limited operational scope, I disagree. It’s being used by many as daily drivers (pun intended), like commutes and soccer practice, and road trips, and it’s here (from Tesla FSD v13 and definitely v14). Not 10 years away.

I want both tbh, I want the people who have no concern for learning how to drive safely that just want to commute to use these, as I believe they will be statistically safer over large numbers.

On the other hand we'll never see a full self driving 911. And I plan to drive one until I can no longer safely drive myself. It's the best therapy I've paid for.

Good points.

About adoption I want to add that at some point a whole generation will turn 16/18 (legal driving age) and just not do a driver's license anymore because they will buy an autonomous car anyway. And IMO from that point on adoption will be very fast.

>If you are today a big insurance like Munich Re, and you see already today that self-driving produces already much less accidents (90% or so I read days ago?), and the tech is really new & "not 100% reliable" and you believe that this tech will be rolled out - one day you will start lobbying politicans that manual driving needs to be forbidden, except some rare cases.

Why would insurance companies lobby for that? 90% reduction in accidents means 90% reduction in premiums, which means 90% reduction in profits.

Do insurance companies have a history of lobbying for safety regulations?

Oh sure, they have!

e.g.: https://www.insurance.ca.gov/01-consumers/200-wrr/Safer-from...

And sure, insurances are fine with making profits, but if they can get less catastrophes, they are fine with that as it reduces overall risk etc.

Feel free to check the corp publishing section on https://www.munichre.com/en/insights.html

That 1st link is the CA government, not an insurance company. Also, it doesn't seem to be a regulation requiring people to do something. It's something people can optionally do, and if so, get a discount on insurance. That's not insurance companies pushing for regulation, it's insurance companies offering a competitive price to both high-risk and low-risk houses, similar to how car insurance companies base their rate based on how risky of a driver they estimate the person to be.

But fire insurance is different from auto insurance. Insurance companies want uncorrelated risk. Insurance companies want a high rate of car crashes, but the exact same rate of car crashes each year, because that makes planning easy. If there's a risk that in a some years way more crashes will happen than other years, that's correlated risk and makes planning difficult; they don't want that.

For cars, there's not much correlated risk. For fire insurance, there is correlated risk due to wildfires. So to reduce correlated risk, insurance companies do likely want to reduce wildfires, while still wanting to increase non-wildfire fires.

Self-driving cars will increase correlated risk, because there could be some software update with a bug that's pushed out and causes a ton of crashes. (That risk does also exist with cars today, due to the various software in cars, but self driving increases the risk.)

The 2nd link is an insurance company, but it as well doesn't seem to be advocating for regulation.

You have a point. At the same time insurance companies are getting out of insurance because they can't afford it. Apparently (I didn't verify this), the cost to repair cars has gone up so much that the economics don't work out. So they're leaving the market. Maybe less crashes would bring them back?

Pick a source: https://www.google.com/search?q=car+insurance+companies+leav...

I believe in CA, the reason for pulling out is that CA limits the amount the companies can increase premiums[1]. So less crashes is one way to bring them back. Allowing them to increase premiums more is another way to bring them back.

[1] https://old.reddit.com/r/explainlikeimfive/comments/1c7lm5b/...

Insurance is cheaper on safer vehicles.

A 90% reduction in accidents is a 90% reduction in _paying out_. That reduces operating costs.

Insurance companies aren't a monopoly. They're in competition with each other to offer lower rates. So if there's a reduction in paying out, they'll need to reduce their premiums to stay competitive with each other.

They can charge companies more, as they drive more miles, for profit.

Good point that more miles driven might increase both accidents and premiums and thus increase insurance company profit.

However, how many more miles will they drive? Double? If there's a 90% reduction in accidents as KellyCriterion alluded to, then the total number of accidents will still go down. That means total premiums will go down, and total profit will go down.

Reduction in premiums doesn’t mean reduction in profit.

It generally does. If accidents (and payouts) drop by 90%, revenue will ultimately drop and profits will follow. Profit margins may increase, but total profit $$ will likely drop.

Aren't insurance profits capped to some percentage of revenue in most jurisdictions by law?

Yes, this is true - but it’s still beneficial enough to have fewer claims. Claims incur cost in many ways and running a business with fewer claims would be more predictable and likely worth the minimal trade off.

> If you are today a big insurance like Munich Re, and you see already today that self-driving produces already much less accidents

Insurance companies don't really care that much if accidents are more or less likely, what they care about is that they can be underwritten effectively.

The most interesting shift here as it pertains to insurance is the fact that losses might be correlated in brand new ways. Today there's no simple way for 100,000 cars to all crash a few minutes apart, worldwide, for the same reason. In the future that may in fact be plausible.

Unless it comes down massively in price, it's not going to displace manual driving for anything other than Uber/taxi. It's far cheaper to drive your own car in most of the us. If parking lots disappear and the parking that remains becomes similar in price to that of major cities and insurance rates skyrocket, then maybe that will change, but only by increasing the overall cost of transportation.

Self driving taxis should absolutely be cheaper to use than current regular cars.

Because they don't need a steering wheel, backseat, and many other things. They can also run close to 24/7.

How "massive" the price drop becomes I don't date to guess though.

> They can also run close to 24/7.

In practice, if you provide enough cars to meet the demand to travel from suburbs to city at 7am-9am and back again 8 hours later, I’d expect a large fraction of them to be idle outside of peak hours. And even at peak hours, to be spending half their time empty, as almost all the demand is in one direction.

And if you don’t provide enough capacity to meet rush hour demand? Good luck convincing people to give up their cars.

You're right that travel demand is impossibly uneven for a 24/7 taxi fleet.

My best answer is that they can deliver goods/packages in the off hours.

That may even end up being the main business.

The accident rate isn't even the biggest advantage over human drivers.

The big game changer is that the computer doesn't need sleep!!

Human truck drivers in the US are only legally allowed to drive 10-12 hours per day. Which usually means the truck is parked half the time. With a computer driving, trucks can run non stop 24/7.

This is revolutionary change! Twice the delivery speed at half the cost. Should improve a lot of things across the economy. Unless some luddites find the power to stop it.

> How will this affect transportation jobs?

The human long distance truck driver will fade away as a career. It will probably take a decade or two, so people will have time to adjust.

The transportation industry should end up substantially larger, which normally creates a bunch of new jobs/professions, but I won't pretend to know how that will play out.

> you see already today that self-driving produces already much less accidents (90% or so I read days ago?)

This is like saying skydiving is safer than driving because you're looking exclusively at accidents and ignoring the total number of events. There are way more cars than there are skydivers. Now in the realm of autonomous vehicles: there are way more manually driven vehicles than there are autonomous ones, so of course there's going to be significantly more accidents from manually driven vehicles.

The safety stats are per road mile, so it is normalized.

https://waymo.com/safety/impact/

I'm sure that an insurance company would reject larger premiums and make itself obsolete, assuming that claims you make are true.

Hint: Normies outside the tech world are not hell bent on making themselves unemployed. That phenomenon only exists among software engineers.

Reply on news.ycombinator.com