Hacker News Reader

Comment by smsm42 | Hacker News Reader

Why federal government needs to invest in this? I mean, when people say "government support is necessary for fundamental research which can not be financed by private means", ok, there's an argument there. But companies like Google - not exactly short of money - already investing into this. Why feds need to intervene? Of course, every company would gladly take gifts from feds to advance their private research - but why do it? There are a lot of fields that could use $4B and are not a focus of attention of a half-trillion-dollar corporations.

Also, they claim simultaneously that self-riving cars are safer and that they would be exempt from safety rules. How that works? If they are already safer, they should be able to satisfy more safety rules, not less?

View on HN

Did you read the article? Such a monumental change in behavours is going to bring in massive changes to the legal framework, safety standards, insurance practices and road markings and upgrades to name a few aspects.

$4B seems like a pittance if that's the goal.

Legal framework is what the Congress and the state lawmakers are supposed to do anyway, and they don't need $4B from federal budget to do this. As for insurance practices, etc. - that's the business of insurance companies, and I'm sure once Google contacts them offering the opportunity to insure world's first self-driving car, insurance industry would be more than happy to jump on that, and they are not the poorest industry either.

It sounds to me like the funding is being allocated as a response to private investment. Companies are going to create self-driving cars - the feds want to be ready for them instead of a roadblock.

Yes, but that much money could be used to feed, clothe, and take care of a lot of people.

Why can't they instead spend 1B on buses, which would serve more people than support for driverless cars, and 3B on the homeless?

Because that's not how funding works. This is research money, you can't just move money around willy nilly.

>>Yes, but that much money could be used to feed, clothe, and take care of a lot of people.

You could make this argument for literally anything that isn't feeding, clothing or taking care of people, which means it is a silly argument.

Not a silly argument at all. Every time we say, "lets spend $x on y", it is fair to question whether that money can be used to feed and clothe... or defend.. or fix the infrastructure, etc.

> whether that money can be used to...

Of course it can, there's no reason it couldn't. That doesn't make it a proper challenge. One would be "instead of $4B, $X should be invested on this, because of such and such". X == 0 without details is a big red flag.

Because if the govt. gave the underprivileged all the money, they wouldn't have any incentive to work hard.

Self driving cars have the potential to solve the leading cause of death amongst young people, and save a significant percentage of GDP http://www.who.int/mediacentre/factsheets/fs358/en/

Seems worth investing in?

Valid from this perspective. There is also the negative impact to GDP from increase in unemployment, which might be inescapable though.

You are assuming perfection. Have you considered that these cars may cause more deaths? What if a solar flare affected GPS and caused all of the navigation systems to injure or kill almost everyone driving at the same time? While that may be only a remote possibility, there are a myriad of other problems navigation systems may run into that they are unprepared for.

> that they are unprepared for

Certainly the people working on this are clowns that know nothing about safety countermeasures, and don't care about their whole industry going belly up if major accidents happened.

The same clowns that run software industry and regularly allow major security incidents costing millions of dollars to happen. It's not like nobody knows making a backdoor in a major networking software is not the best practice. It's just that people are people, and they are imperfect. I'm sure people that make those cars are as imperfect as the rest of us, no more, no less.

It's acknowledging the imperfection of people what makes software an engineering field instead of a hard science. It's not news to anybody.

Continuous testing, code reviews, defense in depth, etc. are practiced to deal with that. The engineers that design cars (the "hardware") are as imperfect as anybody else. So are the engineers that design planes' autopilots. Nobody relies on an engineer being perfect, and there's no need to.

I suppose you think that we shouldn't fund NASA too because that money could be used to feed starving African children too, right?

> spend 1B on buses

Driverless busses would allow for:

* Every dollar spent on busses to go much farther

* Vastly improved public transit, which would allow people to live in cheaper areas and still have access to work.

Do you know how much it costs to buy a bus? It is not unusual to pay 400k+ for a city bus. Let's say driverless cars will cost $20k each (I have no idea), that's 20 cars per bus.

I suggest the only reason taxi trips are so much more expensive than bus rides is the cost of medallions and driver wages.

Once driverless cars arrive, public buses will not make sense anymore.

http://publictransport.about.com/od/Transit_Vehicles/a/How-M...

> Once driverless cars arrive, public buses will not make sense anymore.

How is that exactly? Buses have routes and schedules. How could you plan a schedule with cars if you don't know where everyone is going? You would have to have excess capacity, that's how. And those extra cars will be more expensive than the bus.

Please prove your point with examples.

If the bus is full, I'm sure you're correct. I think there will always be place for mass transit along popular routes.

In the UK there are a lot of buses that drive around with only a handful of passengers. It's a subsidised service that's important to lots of people, but ultimately very inefficient. I'd imagine at least 60% of bus movements (not passenger miles) could be replaced with self-driving cars.

This is only an issue during peak times. Most of the time during the day, buses are fairly empty.

If driverless car companies use something like uber's surge pricing, so that it is more expensive to travel 1/2hr before/after peak, people will respond, flattening out peak demand.

Intelligent routing software will assist, allowing people to car pool to reduce the required number of cars.

Finally, let's say if automated cars remove half the riders from buses, the costs of providing buses will need to be spread across half the number of riders, making the economics of running a bus system twice as expensive per passenger. At that point, it makes sense even more sense to replace the buses with cars.

The only question will be how long it takes for people to adapt due to existing habits having momentum.

About 2000 elderly die of starvation every year, and about 30-40,000 people die in traffic. There are many cases in which the government doesn't have its priorities straight, this is not one of them.

It was government funding via DARPA that led private companies to work on self-driving cars. Google simply hired to two winning DARPA Grand Challenge teams. The Internet is another example showing that the government is quite capable, and often necessary in driving technology forward.

I'm sure when the self-driven cars are finally made, everybody would know the government created them, just as everybody now knows the government created the internet.

Are you high?

It's cheaper and easier to get food, clothes, and care to people over a fleet of self-driving, shared cars than it is over a hodge-podge network of human-driven, privately-owned cars.

If there are any problems with getting food, clothes, and care to people, it's ignorance, apathy, and greed. Not whether cars can drive themselves or not.

Ready in what way (that needs $4B investment)? I understand that you may need to create some laws about who's responsible if there's an accident, etc. but that doesn't look like $4B project.

Did we read different articles?

It doesn't read at all like the US Govt is investing in developing self-driving car technology, of course plenty of private automakers are already doing that, it looks like US Govt is investing in self-driving car public infrastructure and guidelines.

Really?

"U.S. Transportation Secretary Anthony Foxx said the president's 2017 budget proposal includes $4 billion over 10 years for pilot projects, including a program to test self-driving cars by investing in technologically connected roads. He provided few other details on how the money would be spent if it comes to fruition."

That sounds to me like they are try to pick technologies to invest in. If that's the use of funds, I think it's a waste. Let the market choose the best tech.

You think investing in improving infrastructure tech is a waste?

The government is incapable of spending money wisely. The best we can hope for is that the chosen contractor who receives the funds doesn't find it offensive to actually perform the duties contracted.

"...including...investing in technologically connected roads"

The private sector can't install a network of communication technology to wirelessly link roads and self-driving cars on public property (plus you probably need to coordinate with the FCC). From my reading, it seems like they're investing in things that only the government has control over and at the request of private sector companies like Google who want clear national guidelines formed and are probably pushing for a new "connected road" infrastructure.

But if you still can't see that reading of it, then let's just agree that there aren't enough details released yet so let's just wait for more.

Autonomous operation might require updating infrastructure. That falls under the scope of government.

Why federal government needs to invest in this?

Generally I agree with the sentiments you express here, but I just remembered the murderdeathwar on which many in the federal government would prefer to spend $4B. Compared to that, even if this is a total waste, it's not that bad.

That's the thing.. government never chooses A instead of B, they choose to do A, B, and then C just for good measure.

This isn't going to take money from other wasteful or bad areas. It's just going to be more debt.

That's naive. The government, like any big entity that handles money, prepares a budget choosing to do A instead of B.

And wait! Government officials will be glad to take credit for the driving revolution that is about to happen :-) For some reason, people in power think that it was by their hand that the world turns.

It should be noted that the Federal government has been investing in autonomous research for decades.

When you hear stories about dudes building self-driving cars in their garages, you might think the "market" will solve it. But we're no way near having self driving cars.

Because cars by themselves are NOT GOING TO WORK anytime soon. Rain, fog, sleet, snow, road work, garbage on the pavement, etc — all those things can't be handled by today's self driving car tech. One way to get around some of those things is make the roads assist the cars with things like machine readable guidance markers in the pavement, rfid, and so on. Those things will need to be developed and deployed on federal roads for there to actually be viable autonomous cars. That costs money and takes time. I'm just as pissed about the government throwing away taxpayer money on all kinds of shit, but let's face it, this is less than four days of our military budget.

> all those things can't be handled by today's self driving car tech

Some "can't be handled".

Just because Google or Honda's implementation (based on a camera) is geared toward a specific implementation, doesn't make it the best. Cruise Automation has the right idea with sensor arrays that make fog and rain a non-issue. There are lots of options.

No, it's an issue. Heavy rain and fog are even problematic for humans. Current systems use image of the pavement for precise vehicle localization. Pavement not visible? You're fucked. Pavement markings unclear? You're fucked. Road crew painted on temporary markings alongside existing? You're fucked. And so on and so forth.

If the pavement isn't visible, I'm fucked if I'm driving myself.

But you can at least try. You can make slow progress. You can get out, have a look around and continue slowly if needed. An automatic car will just go "no sensor reading, not moving amigo". Bird shat on your lidar sensor? Sorry, have to call assistance to get it professionally cleaned, otherwise the car is not moving. One of the lighbulbs burnt out? Sorry, the car is legally not fit for driving so you are not going anywhere, or alternatively the only destination you can enter is an approved garage.

I personally think this is going to be a huge problem with self-driving cars. They can't go anywhere if any of their sensors are faulty. And they need to have hundreds of them to work in all conditions. Ideally, you would want to have sensors in every wheel to tell you the pressure, sensors that tell you if there is ice on the road, sensors in literally every part of the suspension to monitor if you haven't suffered a broken coil or damaged bearing that could cause an accident - and those things break. Even such a simple thing as pressure monitors in modern cars are notorious for breaking down. Anyone who has driven a 5-10 year old car will know that there always is something broken. In cause of automatic cars pretty much every problem is a breaking problem that immobilizes the vehicle completely, and I think people will get tired of that pretty quickly.

What are your thoughts on the millions of miles self-driving cars have already put on the road without a sensor in every little spot?

Notice how all those miles are in California where it almost never rains. It's still a hard problem, but it's much easier than driving in Alaska or Seattle. When I see millions of accident-free miles from those places, that's whe I'll be truly impressed, even if it's done with adapted infrastructure.

My thoughts are - all self driving cars at the moment operate knowing that there is a human at the wheel, so they are likely to notice a broken suspension coil, rattling clutch, or deflated tyres(even if the sensor reports they are fine, which happens). For "true" autonomous cars the car has to know it itself - especially if it is to allow passangers to literally go to sleep, or even drive without them - so you can summon it from the other side of the city for example. The computer just has to know 100% of its own state, and that requires a huge number of sensors.

That's not entirely true: http://www.engadget.com/2016/01/11/ford-is-testing-autonomou...

Reply on news.ycombinator.com