We don't have 100% autonomous trains yet, yet we're supposed to have self driving cars next year?!

not really informed wrote:The premise is that they only have to be safer than human drivers. If there are 10,000 collisions with only human drivers and 3,000 with self driving cars vs 20,000 total collisions the year without self driving cars, then the advantage is obvious. Aiming for perfection is insanity. Increased adoption will also improve the driving capability of the car while simultaneously making the roads more predictable and therefore increasing safety overall.

That premise might be true in a society run by technocrats that isn't responsive to the public and that is much more centrally managed than the United States. Maybe it would work in Singapore or somewhere else, but hypothetically beating the current fatality rate per mile will not be politically sufficient in the United States to allow fully autonomous vehicles in unrestricted circulation around the country.

In the United States, a hypothetical reduction in the fatality rate that still includes a significant number of deaths caused by autonomous vehicles will never be accepted by the people or politicians. The error rate will have to be pretty close to perfect to prevent serious political backlash.

Here's a good example of the kind of randomness that autonomous technology is not yet able to deal with:

https://jalopnik.com/dont-use-teslas-autopilot-like-this-1792896205

Where's my flying car? wrote:

That premise might be true in a society run by technocrats that isn't responsive to the public and that is much more centrally managed than the United States. Maybe it would work in Singapore or somewhere else, but hypothetically beating the current fatality rate per mile will not be politically sufficient in the United States to allow fully autonomous vehicles in unrestricted circulation around the country.

In the United States, a hypothetical reduction in the fatality rate that still includes a significant number of deaths caused by autonomous vehicles will never be accepted by the people or politicians. The error rate will have to be pretty close to perfect to prevent serious political backlash.

You're right, people are generally irrational en masse. That's why I think this will be a bottom-up movement, with cities adopting it first. This approach makes more sense because urban centers are usually less resistant to change, especially when that change offers to fix one of the biggest nightmares of city living.

Driverless Autopia wrote:

I would trust a computer driving a car over a human who may be distracted, drunk, high, texting, running a red light, etc.

Pretty much. People like to talk about various trolley problems. In reality they don't exist in any noticeable numbers. There are very, very few accidents that are not a result of driver error by one or both of the cars. There just are not very many cliffs where kids can pop out in front of you unexpectedly. And in most cases the car will see kids/people jumping out unexpectedly long before a normal driver would.

Right now we are ~5% away from having a self driven car at level 5. But that last 5% is often much harder than the first 95%. It could easily take 20+ years to solve the last pieces of the problem. Short term things like highway driving in any type of decent weather is pretty much a solved problem. Yes the systems will screw up occasionally (see Tesla's crashes on a car that isn't remotely self driving) but it will be a fraction of the human error rate.

DiscoGary wrote:

Could we please just skip to the personal jet packs I was promised many years ago?

For you DiscoGary:

Anyone who thinks hacking isn't an issue doesn't understand how simple it is, and how absurdly costly it is to guard against. I'm sure when a car is sold, it'll have some robust safeguards in place against hacking. But can manufacturers aren't in the game of updating software and patching security holes for vehicles they sold five, six, seven years ago. And overtime, holes are found and exploits open up.

This will be a big issue going forward with all kinds of appliances and household devices that are controlled via internet networks.

And all this stuff WILL be online. Nobody's going to sell you a car with an AI setup that doesn't access the Internet. The data will be invaluable.

We also should remember that the weakest link in the chain is always going to be the user. Something as simple as not updating software can open things up to enormous security concerns. And it may not be the great big road tragedy, but something as simple as you're unable to start the vehicle (the microwave, the oven, your washing machine) until you deal with the ransom ware.

Where's my flying car? wrote:

A good article on issues with autonomous trains.

It turns out we have all the tech needed to make autonomous trains, and we've seen robotrains running in limited capacity since the late 1960s. The problem isn't technology. It's line of sight, and the massive distances trains need to identify and react to obstacles and bring a few hundred tons of steel and cargo to a stop safely.

https://www.wired.com/2013/04/why-arent-trains-autonomous/

This. ^

As I mentioned earlier, computers SUCK at identifying objects and differentiating between a garbage can and a dog.

Add fog and a down pour like you get in the mid west as a storm passes, and you're "sophisticated" sensors break down.

The train article above shows the flaw in trading the human eye and human brain for a computer trying to "see". You can't compete with millions of years of evolution and brain neurons firing to interpret objects far away and react instantly to threats.

On another note, the roads we have today are designed to be safe IF people simply followed the rules. Curves on highways are re-aligned to be more gradual to prevent accidents and ensure that it corresponds to the speed limit of that road. Obviously, people don't follow the rules, I get that.

But here's the thing. Is the goal to make driving safer? or to make our lives more efficient? Or have you considered that companies like Google want to work on it simply because they have a lot of money and their engineers want to work on something "cool" and they just want to sell you something?

If the goal is to make the roads safer, then there are other ways to go about that, rather than investing trillions of dollars and trying to implement an overly complicated computer system and potentially leaving millions of people out of work.

It's called being a responsible individual and not driving drunk, not texting. I'm not a conservative, but I think right-wing minded folks are on the right track when they talk about personal responsibility.

PS Disco Gary is awesome.

bigno wrote:

If the goal is to make the roads safer, then there are other ways to go about that, rather than investing trillions of dollars and trying to implement an overly complicated computer system and potentially leaving millions of people out of work.

It's called being a responsible individual and not driving drunk, not texting. I'm not a conservative, but I think right-wing minded folks are on the right track when they talk about personal responsibility.

PS Disco Gary is awesome.

Of course personal responsibility is the biggest issue with vehicle safety today.

How do you mandate that?

In today's world you wait UNTIL AFTER a wreck has occurred and someone has died. Why not do something BEFORE a wreck occurs and someone dies. People are not going to get MORE responsible. With more and more distractions people are simply going to be more irresponsible behind the wheel.

There will be wrecks with self driving cars, everyone know this, but if you have a majority of self driving cars on the road you will see less accidents and less deaths.

bigno wrote:

If the goal is to make the roads safer, then there are other ways to go about that, rather than investing trillions of dollars and trying to implement an overly complicated computer system and potentially leaving millions of people out of work.

It's called being a responsible individual and not driving drunk, not texting. I'm not a conservative, but I think right-wing minded folks are on the right track when they talk about personal responsibility.

PS Disco Gary is awesome.

And abstinence is a foolproof way to prevent unwanted children. Its a nice thought but it is not going to happen.

Like any new major change it will happen in stages. Just like the move from the horse to the car was not immediate so will this occur. The first popular phase has already started with the technology just as a better cruise control. From there little bits of driving will be shifted to the technology - perhaps highway lanes for automatic mode. Insurance discounts for certain types of software to encourage use, etc etc.

The abilities of the tech will always far ahead of the popular use. But I think an adoption rate of about 5 years for each new incremental advance is about right.

Phase 5 may be 30-40 years away but by the time we get there it will hardly be noticed.

Remember, 30K lives are lost to traffic accidents a year and that does not include injuries, lost work time etc etc. There is a substantial societal cost that can be saved and that will give government an incentive to allow this service to be rolled out in the form of incentives, legal protections, etc.

Overall I think it is inevitable.

Driverless Autopia wrote:I would trust a computer driving a car over a human who may be distracted, drunk, high, texting, running a red light, etc.

But would you trust an autonomous car over driving a car yourself?

Studies show that 88% of American drivers believe that they are above average drivers.

Of course the math doesn't work out.

You have to convince the 88% of drivers who think they are better than average that they are safer with autonomous cars, not that an autonomous car is better than a drunk, high, texting drivers. We're a long way away from that as a political and social matter, on top of the improvements in technology and infrastructure that will be needed.

How many people will sleep good knowing cars are driving themselves. What if one of those cars crashs into your home and kills your family, are you going to put a computer in prison. Most of you posters must be smart kids, too much school, too much breast feeding.

welcome to the US...where 70% of adults are overweight. 38% obese. healthcare cost at 1/6 of GDP. and we're looking to make our lives more..."efficient"

we should be riding bikes and walking. using our bodies to get around. but if you want to stream netflix in the backseat for another 8 minutes...

self-driving car promoters just another wave still getting us hooked on the automobile. more addictive and costly than cigarettes!

apackeddirttrack wrote:

welcome to the US...where 70% of adults are overweight. 38% obese. healthcare cost at 1/6 of GDP. and we're looking to make our lives more..."efficient"

we should be riding bikes and walking. using our bodies to get around. but if you want to stream netflix in the backseat for another 8 minutes...

self-driving car promoters just another wave still getting us hooked on the automobile. more addictive and costly than cigarettes!

Can't come soon enuf, sez I.

I'm looking forward to being chauffeured to Walfart, Boiger King and thence to the lickker store in my own 1975 Dodge Dart.

Tech progresses exponentially ...

We won't just be "there" all-in 30 years from now, EVERYTHING will be different 30 years from now. Auto cars, auto trucks, auto planes, auto trains, drones like you won't imagine,

Here's a good take on the way fans of autonomous cars are overselling current capabilities.

http://jalopnik.com/stop-saying-dumb-shit-1795939663

Hurl wrote:

Tech progresses exponentially ...

No it doesn't. Tech progresses at best logarithmically. Progression along one direction is big in the beginning but then dies off.

moores law, biatch

wejo wrote:

The question is when do we get to self-driven cars being safer than humans? You and the guy I talked to last year don't think that is anywhere near. Elon Musk thinks it's very soon.

Most of the people who say we will soon have self-driving cars, have a financial interest in promoting that idea..

A self-driving car will need AI that is functionally equivalent to a general-purpose intelligence. That is a hard problem. Moravec's Paradox dates back to the 80s when I was working in AI - things that are hard for humans are easy for computers, but things easy for humans are hard for computers. The second half of that is not generally understood.

As a computer scientist I am not yet convinced that self-driving cars in uncontrolled environments can work.. level 5 cars have always been 10-20 years away, and I suspect always will be.

See

and

and

Why do people ask Lets Run, when they can just ask google and get a better answer?

Oh, it's a brojo. Must be trying to pump the site after learning that college coaches (let alone elites) don't read the site:)

The answer according to wired is the time it takes to stop a train (up to a mile), the the difficulty spotting something on the track in the distance).

The answer according to motherboard is economics: the number of conductions per/passenger is so low the it's not worth the investment to finish the technology required to make a safe self driving train (to determine there's an obstruction a mile ahead on the track).