Cringely Pans Self-Driving Car Hype, Says They're Years Away (cringely.com) 286
In what may be his final year of technology predictions, columnist Robert X. Cringely argues "I can't say that we're going to see anything beyond more beta tests of self-driving cars in 2019... We simply aren't ready and probably won't be for years to come...."
"The problem isn't with the self-driving cars, it's with the cars that aren't self-driving, cars that are driven by idiots like me." It will eventually happen. Once half the fleet has been replaced with cars that could be self-drivers if we allowed them to be, then there will be a huge financial incentive to get the other half off the street. This will be especially the case if climate change is still accelerating. I'm guessing that most cars from 2020-on could be self-driving with only a software upgrade, which is why Elon Musk is predicting Tesla will have full autonomy by the end of 2019. But notice that Elon isn't predicting Tesla will be allowed to have its cars drive themselves everywhere...
So why is the world talking so much about self-driving cars and full autonomy? Some of it is Tesla hype, some of it is marketing as the car companies try to get us to buy cars that will eventually be self-driving, but probably not until their second owners. And the other reason why we're talking so much about self-driving cars is because Uber is planning to go public later this year...an IPO that will go smoother if the driving public thinks autonomous cars are something that we'll be seeing soon. Uber has a labor problem. If it can spin a story that surly and expensive human drivers are soon to be replaced with electrons, that will be very reassuring to Wall Street. But as I explained, it also isn't true.
The world isn't yet ready -- something Uber and Tesla and all the others will suddenly admit in about a year (post-IPO).
Cringely also argues that the problem isn't just the "millions of drivers who are still controlling their vehicles the old fashion way, which is often in a barely competent fashion..."
"We keep our cars longer because they don't rust and we can't afford to replace them so often. The result is that while we could expect a complete turnover in car technology every decade, now it takes closer to two decades."
"The problem isn't with the self-driving cars, it's with the cars that aren't self-driving, cars that are driven by idiots like me." It will eventually happen. Once half the fleet has been replaced with cars that could be self-drivers if we allowed them to be, then there will be a huge financial incentive to get the other half off the street. This will be especially the case if climate change is still accelerating. I'm guessing that most cars from 2020-on could be self-driving with only a software upgrade, which is why Elon Musk is predicting Tesla will have full autonomy by the end of 2019. But notice that Elon isn't predicting Tesla will be allowed to have its cars drive themselves everywhere...
So why is the world talking so much about self-driving cars and full autonomy? Some of it is Tesla hype, some of it is marketing as the car companies try to get us to buy cars that will eventually be self-driving, but probably not until their second owners. And the other reason why we're talking so much about self-driving cars is because Uber is planning to go public later this year...an IPO that will go smoother if the driving public thinks autonomous cars are something that we'll be seeing soon. Uber has a labor problem. If it can spin a story that surly and expensive human drivers are soon to be replaced with electrons, that will be very reassuring to Wall Street. But as I explained, it also isn't true.
The world isn't yet ready -- something Uber and Tesla and all the others will suddenly admit in about a year (post-IPO).
Cringely also argues that the problem isn't just the "millions of drivers who are still controlling their vehicles the old fashion way, which is often in a barely competent fashion..."
"We keep our cars longer because they don't rust and we can't afford to replace them so often. The result is that while we could expect a complete turnover in car technology every decade, now it takes closer to two decades."
Life is chaotic (Score:5, Insightful)
Re: (Score:2, Troll)
Re:Life is chaotic (Score:5, Insightful)
I shudder to think how many people are going to have to die in utter terror before everyone else understands this
The thing to remember isn't that the goal is zero deaths. The goal is less than what we currently have deaths. You will never have a zero death anything on "open" anything, full stop. So the number of people who die in fully automated cars is always going to be a non-zero number, thinking that it will ever be anything other than that is just not accepting reality. However, in a fully automated environment, we will have a number of deaths that is less than the non-automated environment.
Re:Life is chaotic (Score:4, Insightful)
The problem is that if the automated system can cause a crash, it will look bad. Just look at Boeing and MCAS - we do not know how many crashes it prevented (probably a few, since the plane could not be certified without it), but we know about the crashes it caused.
Every time a self-driving car crashes in a way that only a drunk human driver could (say, empty road, drive straight into the lane divider) it makes it look really bad because people can think "I don't drive drunk, but it looks like the AI could, I want my manual car back".
Since we trust software so much, why not have voting over the internet first? I am sure it will be secure and in no way vulnerable to hacking, since humans can write software so well.
Re: (Score:3, Interesting)
Just look at Boeing and MCAS
That said, fact remains that Boeing makes planes that fly millions of miles without a crash. I thought here on Slashdot we were trying to rise above the media hype?
because people can think "I don't drive drunk, but it looks like the AI could, I want my manual car back"
Yeah and people think that vaccines are bad. There's always going to be that group of people who look at one event and use that as evidence that something is bad in spite of the volumes of evidence otherwise. Who knew we lived in such a world?!
Since we trust software so much
Software not so much, but there is a profit motive to ensure that say an automated Semi deliver it's
Re: (Score:2)
Which does nothing at all to change the fact they cut corners to compete with Airbus. Cost cutting that has cost hundreds of lives in less than a year, and could have cost hundreds more if the planes hadn't been grounded around the workd, forcing Boeing to start addressing the problem.
Vaccines don't cause autism. The Boeing crashes actually happened.
Re: (Score:2)
Cost cutting that has cost hundreds of lives in less than a year
Yeah, but you're still focusing down on these two events that killed 347 people. All the while millions of people fly with no problem. That's what I'm getting at here and you just proved my point by doubling down on it as some argument. In 2017 around 4 billion people on this planet flew in a plane and here you are talking about cutting corners that killed 347 people. On average, without cutting corners required, human driven cars here in the US killed that many between when I left work on Friday to whe
Re: (Score:2)
Re: (Score:2)
False equivalency.
Did you not see what I wrote? Believe it: the so-called 'AI' they keep trotting out cannot 'think' at all and will never be up to the job, it will ALWAYS fall short, and people will die needlessly because of it. No fucking way. Never, ever, would I ride in one of those. Full-on general AI (which we are no where NEAR having yet BTW) or nothing at all. Has to be equivalent to a human mind. No compromises.
Re: (Score:2)
Has to be equivalent to a human mind. No compromises.
If we didn't have humans driving today, I doubt that "equivalent to a human mind" would be an acceptable requirement if we were to invent road traffic. I would be scared shitless by the thought of humans piloting these missiles if I wasn't already desensitized to them. In fact I still am scared.
You probably don't want your car to be mad at you because you didn't let it flirt with that BMW at the red light, or spending a majority of its CPU cycles calculating cryptocurrency because it is bored from driving t
Re: (Score:2)
No fucking way. Never, ever, would I ride in one of those. Full-on general AI (which we are no where NEAR having yet BTW) or nothing at all. Has to be equivalent to a human mind. No compromises./em.
so you want an AI to be able to get impatient, angry, tird and distracted? Why on earth do you want that?
Re: (Score:2)
so you want an AI to be able to get impatient, angry, tird and distracted? Why on earth do you want that?
Because that would instantly make them 20% better drivers.
Re: (Score:2)
and people will die needlessly because of it
But fewer people will die of it, that's the point. Fewer.
Has to be equivalent to a human mind. No compromises
How about no. I'm tired of humans rubber necking and rear ending into cars making things worse. I hardly need a machine to do all of that.
False equivalency.
I don't see it as such. Anti-vaxxers literally focus down on the dozen of deaths from vaccines as reason to not have them in spite of the millions of people who literally have zero happen to them other than they become inoculated to a fully preventable disease, which anti-vaxxers additionally use the argument th
Re: (Score:2)
You never, NEVER, EVER force people to use SDCs. EVER.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Nearly all of the advances that could improve automated also apply to assisted driving
Not so, but depends on where your line is here. Level 1 and level 2 automated driving I would say, yeah, you've got a point. But Level 3 and 4 however, I would argue that it would be hard to add things from level 4 and some things from level 3 and still call it assisted driving, but things like assisted driving is just a not so well defined label, I guess I could just simply say, depends...
Fully automated will always be less safe.
That literally makes no sense. A level 5 vehicle which requires zero people inside of it, crashing and only destroyi
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Most deaths on the road occur because of people stupidity like sleeping, eating, looking away, emotional state etc.
I think you need to cite some evidence for that statement.
Re: (Score:2)
Re: (Score:2)
Re: (Score:3)
Re: (Score:2)
And deterministic systems don't handle chaos very well. Even if all the cars are automated, you have people, dogs, deer running into the street. You have blowout tires. You have road damage, debris, snow. Given that understanding a left-hand exit is problematic today, how does it handle a deer dashing on to the road? Or snow covering all lane markings - and heavy enough to block GPS signals?
I think the fact that making a left turn is hard is more of a testament to how much humans are just winging it, I'm sure the computer actually has more detailed information on the opposing traffic than I do. I've absolutely made left turns where I had to stop dead due to an unexpected pedestrian and I'm relying on the opposing traffic not to ram me because I'm still in their lane. In any case, if that was the only hindrance we'd put up turn lights in addition to the straight red-yellow-green. As for everyth
Re: (Score:2)
"As for everything else, I'd be happy with an AI that detects when it's out of its depth."
How does it know that? Sure it might get a weather report over the network. But maybe during a storm the network isn't reliable. So now its relying on its sensors? What if it doesn't recognize a flooded lane as 'too deep to cross' or that its flooded at all instead of just flat? And it only discovers its 'out of its depth' after its literally in too deep?
"My commutes are 99% boring, if it one day it says there's a bliz
Re: (Score:2)
That's why self driving AI is not deterministic (Score:5, Insightful)
And deterministic systems don't handle chaos very well.
That is why we hadn't been close to getting anything like a self driving car until the re-emergence of practical neural networks.
Neural networks are up to the task because they are fusing info from dozens of sensors and models to determine every second where the car should be moved. The people working on self driving car tech today are building systems can handle any surprise because fundamentally the car is going to try (A) not to hit anything, and (B) go somewhere else if it has to override some laws of the road to do so, in an emergency.
It can basically make a car do anything within the ream of physical ability which is way, way better than 99% of human drivers can do.
how does it handle a deer dashing on to the road?
About a billion times better than a human can, that's for sure. It can see in infra-red, radar and sonar - in 360 degrees, not just straight ahead. It can react and brake quicker. It can steer off the road almost a billion times better than any human would, since it would also be continually considering that semi truck behind you that you totally forgot about when you saw the deer and that cannot possibly avoid hitting you if you stay where you are...
Re: (Score:2)
But so far, they can't handle a parking lot or a divider in the road or heavy rain. People can do that a billion times better than a computer can.
I bet these vehicles couldn't handle London with it's constantly changing roads due to road and building works.
A Waymo vehicle can go 11000 miles on average before human intervention is needed, I bet this would be more like a few hundred miles in London. They did manage to double that distance in the past year, but can they do that again or are there diminishing r
Yes, they can (Score:2)
But so far, they can't handle a parking lot
Wrong as of now [teslarati.com]
or a divider in the road or heavy rain.
There are more incidents with humans being unable to handle dividers than Teslas, so already you are wrong there.
As for heavy rain that also causes problems for humans, again self driving cars have more sensors so in the end they will handle heavy rain lots better than humans mostly do.
Unlike a human, self driving cars will also be much more inclined to drive at a reasonable speed for conditions, because they f
Re: (Score:2)
My current car has sonar. That’s what feeds the beeps to tell me I’m about to back into a lamppost. :-).
Re: (Score:2)
Re: (Score:2)
Smart sounding bullshit. For a start, chaos is deterministic.
That's randomness, not chaos.
Re: (Score:2)
Re: Life is chaotic (Score:2)
Thatâ(TM)s only because you havenâ(TM)t bothered looking.
After a few seconds on Google, first hit:
https://www.theregister.co.uk/... [theregister.co.uk]
Re: (Score:2)
It all has to (Score:2)
Re: (Score:2)
Re: (Score:2)
Pry my manual transmission pickup truck from my cold, dead fingers. Never give up, never surrender.
Good for you. I can't wait to not have to drive because I'm a lousy driver and have other things I'd prefer to do than baby sit a machine during my daily commute.
As Cringely points out, we have to share the roads. Hopefully our preferences (and vehicles) don't collide.
Re: (Score:2)
Re: (Score:3)
it all has to start somewhere sometime. Might as well start now as the hardest part seems mindsets not electromechanical elements.
The hardest part is the software, which has not seen much improvement in the twenty years or so that we've had SDCs. We threw 1000x resources at the problem since the mid-nineties, and only got a marginal improvement.
What makes you think that this is a near-solvable problem? It clearly isn't.
As an occasional away drinker (Score:3)
What does it add to the mix if we consider the vast number of intoxicated drivers removed from the roadway if autonomous vehicles become ubiquitous?
I am reluctant to drink much when out at social gatherings because I feel the need to remain coherent enough to pass a random LEO roadside interview when called upon to do so. I would certainly consume an additional adult beverage or two if the task of driving home safely was out of my hands. Perhaps we could we get the powerful alcohol lobby behind the implementation of inhuman vehicle piloting.
Re: (Score:2)
I would certainly consume an additional adult beverage or two if the task of driving home safely was out of my hands.
it already has been. It's called Uber and Lyft and there are 0 excuses to drive buzzed let alone drunk when all it takes is a few taps and 7 or 8 bucks to get home safely without putting a ton of other peeps on the roads at risk.
Re: (Score:3)
it already has been. It's called Uber and Lyft and there are 0 excuses to drive buzzed let alone drunk when all it takes is a few taps and 7 or 8 bucks to get home safely without putting a ton of other peeps on the roads at risk.
I remember reading some studies on this, and the reduced cost of Uber/Lyft has indeed dropped drunk driving. However, rm does have a point. Taxis have been available for centuries, starting with horse and even human drawn carriages. Let somebody else get sweaty hauling you to your location. The question is cost and convenience. A ride-share ride is a bit more than $8 for me, and in some ways even $8 is pricy when you realize that I need to spend it twice - getting home is one leg, of course, but either
Re: (Score:2)
> I remember reading some studies on this, and the
> reduced cost of Uber/Lyft has indeed dropped drunk
> driving.
More than that, I'd wager that, if you were to run the numbers, Uber/Lyft would also be shown to be good for bars and nightclubs. That same $8 Uber home from the Castro used to be a $45 taxi ride. And that's if you could successfully hail a cab in the first place; because their dispatch number is something between an exercise in futility and a sick joke. With Uber and Lyft, I go out a
Re: (Score:2)
7 or 8 bucks?
It would take 60 dollars to get back to my place, probably more than that at midnight.
that's full on taxi prices down here. I have yet to have to pay more than 12 bucks ever on either service here in south texas.
Re: (Score:2)
That'll never happen because autonomous cars will never be fully autonomous as to not need a driver behind the wheel to take over in case of an emergency. The insurance lobby will make sure of that.
I think you're onto something with the influence angle of the insurance lobby, but unlike some studies, the insurance industry uses scientifically generated actuarial tables [lifeexpect...lators.com] to measure risk and premiums. If autonomous drivers don't rapidly eclipse humans in categories like measurable risk, I'll get your car fare and pop the infected-looking pimple on your maid's back.
Re: (Score:2)
For things like insurance, the accident-with-damage rate is more important than the fatality rate.
The number of automobile fatalities per year in the US is already so small that every dollar we spend to reduce a US road fatality returns less than a dollar in benefit.
The cost of non-fatal accidents, though is enormous - especially for continuing medical costs. So, yes, I could see the cost of insurance pressuring people into purchasing autonomous vehicles if the non-fatal crash rate drops significantly, eve
Re: (Score:2)
Wonâ(TM)t solve the problem; if youâ(TM)re not sober enough to drive you wonâ(TM)t be sober enough to read the inevitable t&cs required before getting into your self driving vehicle
I don't know... isn't it likely a greater percentage of /. posters read the summary and article than total users who bother with perusing the terms of service?
Second owners, Right! (Score:4, Insightful)
In the future most cars won't have second owners. They will just be tech trash after their initial owner.The tech will be obsolete, no parts available to fix, no right to repair even if one can and want to.
Convenient transportation will just become one of those things only the elites that can afford to have. Everyone else will be required to live with sub standard/inconvenient public/private transportation providers
But I am not worried. I am 63 and I might see a full scale deployment, Maybe!
And I am a target audience, but one thing I know, it will be more expensive and much less convenient than having your own vehicle. And it will not happen in the next 5 years, except maybe small test runs in a few urban areas.
Just my 2 cents
Re:Second owners, Right! (Score:4, Interesting)
In the future most cars won't have second owners.
In the future, most cars won't have first owners (unless you count the automated-fleet service's assets department). Most cars will spend their entire services-lives operating as driverless taxis. As for how long they will last before becoming unusable, it will probably be quite a long time, assuming their battery-packs can be economically replaced. Competition will see to that; nobody running a driverless-taxi fleet is going to put up with cars that don't yield a good return on their investment.
This asshole can go fuck himself (Score:2)
old fashion way (Score:2)
Don't you mean "old fashioned way"? Silly old fart.
Retrofit (Score:4, Insightful)
My car is 21 years old, still under 100K miles and runs great. There are a lot of older cars out there. The rich tend to forget that most people aren't on 3 year leases.
The only way I can see every car on the road being self-driving before 2050 is if the tech to retrofit an existing car with self-driving features gets so cheap that it can be subsidized for the poor. If it costs say $500, then some equivalent of the cash for clunkers program could pay people to go self-driving. If we can't cheaply retrofit existing vehicles, then we're going to have to wait at least 30 years.
Re:Retrofit (Score:4, Interesting)
Average miles driven per year is ~13k in the USA, so if your car is under 100k, you're driving around a third of average. Even a '90s era car could expect to do 100k miles without major issues, so "runs great" is assumed as long as it wasn't exposed to excessive environmental problems.
That said, you are correct. It will take a "long" time for the shift. I tend to rate it in milestones. We're currently at "testing on public roads". Some other milestones:
First self-driving commercial vehicles available for sale/lease - taxis, delivery, intra-company transfers, areas where a certain amount of limitations are acceptable, and more attention can be made for specialized maintenance requirements. Looking cool matters less than practical.
Next, self-driving car available for public sale to private individuals/parties.
Self driving cars become dominant, IE over 50% of sales
Self driving cars become exclusive - only special duty vehicles aren't self driving.
Note, these milestones are basically at the dealer - it isn't what is actually on the road.
so you realize that the average age of cars on the road today are 11.7 years, so if half of new cars sold are self driving, that doesn't mean that half of the cars on the road are. Maybe 5% are.
So, if each milestone takes only 5 years(them moving to dominate could happen very quickly, but we're lagging on the introduction itself), that's 20 years before virtually all cars sold are self-driving. Then, another 20 years to get most of the remaining human driven cars off the road. Time for virtually all cars on the road, short of the occasional dude taking his Model-T out for a spin? 40+ years.
The way you get there (Score:3)
The only way I can see every car on the road being self-driving before 2050 is if the tech to retrofit an existing car with self-driving features gets so cheap
It's pretty easy to get to a point where 99% of cars on roads are self-driving - you just reach the point where it's cheaper to subscribe to and use a low end car service that takes you door to door, than it costs to maintain and gas a beater car.
Unlike others though I don't see a future where most people use a car service like that, it's nice having
Re: (Score:2)
That, frankly, is an impossibly high bar. For a significant percentage of us (far more than 1%), even public transit subscriptions to get most of the places we want to go (ignoring the fact that it doesn't go everywhere we want to go) costs more than maintaining a
Re: (Score:2)
Personally my car costs average $100/mo
Your experience certainly isn't the norm. If you purchased that car new, it's impossible. Your commute is so short you put less that 5k miles per year on your car, but so long that it spans multiple counties. Your car is much older than average but requires fewer repairs than average. A lot of people pay more than than in just fuel costs each month. Heck, some people's insurance alone is more than that. I have a very similar car and even though I have had no major repairs in all the years I've owned it
Re: (Score:2)
cheaper to subscribe to and use a low end car service that takes you door to door, than it costs to maintain and gas a beater car.
You have missed the Tesla lesson. Whether electric, autonomous or whatever, the uptake of these technologies will be at the high end of the market. Not the low end. Cars are, and will be for the foreseeable future, status symbols. Because of that, 'gassing up' a car (beater or otherwise) is a non-issue for rich people. In fact, the big, low mileage vehicles are a demonstration of wealth specifically because they are expensive to run.
Re: (Score:2)
You are poor.
Odds are he is rich. Older used cars are in pretty high demand and, as a result, expensive. How about a 40 year old truck that sold (new) for less than $10K and now commands $40K to $80K? If you think 'cash for clunkers' is going to get those off the road, you are delusional. Registration won't either. In many cases, older cars are exempt from inspections and annual renewals. They are here forever.
Sam Vines boot theory (Score:5, Interesting)
You also have the Sam Vines boot theory. A good set of boots that will last you for life might cost as much as 5 sets of boots that each last a year, especially if you take care of them, but it's the rich people who will buy the good boots, while the poor stay in the hole buying a new pair of boots every year or less, despite it costing more over the long run.
It's expensive to be poor.
In this case there's a good chance that Gava bought a good car, though I'd argue his car is lasting mostly because he doesn't drive it much. Less than 100k miles on a car that old is extremely low mileage, less than 5k/year, when the average is 12-15k.
Re:Sam Vines boot theory (Score:5, Insightful)
Cars are the absolute reverse of Sam Vines boot theory. The more you pay for a car, the more it's going to cost you every year to maintain it. If you buy a cheap car you tend to find it's cheap to maintain -- because it doesn't have all the fancy parts like power windows or self-driving to break, and because parts are easier and cheaper to find. Manual transmissions are the cheapest of all to buy and maintain, of course.
Personally, knowing I drove about 5K mi/yr, I spent $4000 on 10 year old (at the time) Ford Escort that had 45K miles on it with the knowledge that it should last me 10-20 more years. Your ideal option may vary.
Most things don't obey boot theory. Cheap stuff can actually last ages, if you take care of it. Cheap PCs can last decades, a cheap flip phone can last decades. Even cheap boots can probably be repaired to last quite a while -- I've made $10 walmart shoes last quite a few years with shoe goo.
The way in which being poor is expensive is almost entirely debt. If you're low income but have savings (like me), you're okay -- if you're higher income without savings, your money is going to burn from the loans you take out.
Re: (Score:3)
Personally, knowing I drove about 5K mi/yr, I spent $4000 on 10 year old (at the time) Ford Escort that had 45K miles on it with the knowledge that it should last me 10-20 more years. Your ideal option may vary.
I can't quite put my finger on why, but - I really like this guy.
Re: (Score:3)
Most things don't obey boot theory. Cheap stuff can actually last ages, if you take care of it. Cheap PCs can last decades, a cheap flip phone can last decades. Even cheap boots can probably be repaired to last quite a while -- I've made $10 walmart shoes last quite a few years with shoe goo.
Some cheap stuff lasts for ages, deending on how you treat it. Some cheap stuff is utter junk.
Cheap scredrivers, for example. Not only do they get the ends mushed up quickly, they also mush up the screws too making your
Re: (Score:2)
Re: (Score:2)
My '98 Escort really isn't in demand and never will be, so no, I'm poor. There are tens of millions of us in the USA, and tens of millions more who simply don't have any interest in buying another car until they have to. Any politician who proposes banning us all from the road would be writing a ticket to instant humiliation and early retirement.
There are also the rich people you mention with classic cars, but those aren't a problem for self-driving since the owners have plenty of money to pay for a retrofi
Re: (Score:2)
They're not exempt from mandatory liability insurance. If Google is telling anything remotely resembling the truth about their Waymo division; once the data finds it into the hands of the insurance companies' actuaries, it will quickly become prohibitively expensive to drive human-operated cars, except perhaps occasionally as a hobby.
Re: (Score:2)
it will quickly become prohibitively expensive to drive human-operated cars
That's not how insurance works. I have to cover my liability, which is governed by my ability to drive safely. Autonomous cars may have lower liability. But that doesn't drive my costs up. Just theirs down.
And in my state, I can buy a bond to cover my liability. If the state says I have to have $100K of liability coverage, I buy a bond for that price. Oh, and I get the interest the bond pays. So I'm actually making some money off that deal.
Re: (Score:2)
Re: (Score:2)
Mint condition FJ40.
Only true for level 5 autonomy. (Score:5, Informative)
It's important that distinctions be made when referring to self-driving cars. There are five levels of autonomy and I can only say that this claim that it's "years away" only applies to level five because Waymo has already demonstrated a level four autonomous vehicle.
While it may or may not be true that level five autonomy is still be years off, that doesn't negate the fact that we have level three already deployed and level four in development.
Level 5 for full deployment, 4 for limited (Score:3)
I'd argue that a "strong" level 4, something that can handle something like 90% of the tasks a human driver would be expected to be able to handle, but 99%+ of daily tasks, would still be useful for taxi and similar services. It's more for personal owners that you need full capability. Even a city is an artificial construct, and could handle some modification to better suit self driving cars.
For professional services, if the taxi runs into a problem it can't handle, there could be a central office with tr
Re: (Score:2)
Yes, exactly. Even as Cringely publishes this piece, Tesla is racking up billions of miles worth of real-world data to train their AP system. But all we hear about in the news is the occasional accident or near-miss, we don't hear about all those safe miles... nor are these incidents put in context of how much safer the autopilot is than a human driver in most circumstances.
That said, however, I think the "hype" Cringely's talking about is that true Level-5 autonomy that allows driverless taxis long-haul tr
Re: (Score:2)
You can train your NN with as much data as you want and you will never achieve autonomous driving. That is what Tesla-nuts don't understand: NN work NOTHING like the human brain does. NOTHING. The fact that they are even called neural nets is a scam.
Re: (Score:2)
You can train your NN with as much data as you want and you will never achieve autonomous driving.
The entire driving system is not a big ass NN. The NN's form components of it like object detection and semantic segmentation.
That is what Tesla-nuts don't understand: NN work NOTHING like the human brain does.
It doesn't really matter what the tesla-nuts understand, it matters what the engineers at Tesla understant.
The fact that they are even called neural nets is a scam.
Oh get over yourself. They've been call
Re: (Score:2)
You can train your NN with as much data as you want and you will never achieve autonomous driving.
That's not true at all. It's likely to reach a point where it's better than a human driver and yet there will still be collisions. What you are expecting is perfection which is far beyond human capability.
That is what Tesla-nuts don't understand: NN work NOTHING like the human brain does.
You seem to have made the false assumption that the human brain is the most optimal mechanism for driving. In case you hadn't noticed, neural networks have greatly exceeded human capability at performing some tasks already.
self-driving, self-landing, bah-humbug! (Score:5, Interesting)
Cringly is a familiar old idiot. [wikipedia.org]
Watson & Crick predicted it was going to take 30 years to sequence the human genome. Venter did it in a fraction of that time, because he thought outside the box. [wikipedia.org]
Just a few years ago I had the pleasure of being surrounded by crusty old defense contractor types who prattled on about how Elon Musk and his stupid little rockets were literally nothing but a flash-in-the -pan publicity stunt. They insisted that self-landing re-usable rockets were not feasible or we'd already be doing it....
Same argument here. Self driving cars are stupid, they don't work well enough, they'll get people killed, they are years away from being practical, Tesla sucks, blah blah blah.
Re:self-driving, self-landing, bah-humbug! (Score:4)
On the other hand, we've been only 5-10 years away from a practical fusion reactor for - what - 50 years now?
Your examples don't remotely prove that Cringely is wrong. They only demonstrate that sometimes even an expert's predictions on a subject they know well can still be blazingly wrong... and Cringely is no expert by any stretch of the imagination. But you can't say "Mr. A was wrong about topic B, so Mr. C is also wrong about topic D".
Re: (Score:2)
Self landing rockets isn't any better than the alternative, which is why no one else continued making them since they were first introduced 40 years ago.
I guess somebody better tell the ESA they are wasting their money, then. https://www.popularmechanics.c... [popularmechanics.com]
Re: (Score:2)
I never claimed SpaceX was the first to launch a self-landing rocket, although they are the first to use such a rocket to actually reach orbit and deliver a payload. So please, before God and this vast court of public opinion, please illuminate my lie...
As for your assertion:
Self landing rockets isn't any better than the alternative, which is why no one else continued making them since they were first introduced 40 years ago.
Based on this: SpaceX Falcon 9 Capabilities and Services [spacex.com]
and this:
ULA Atlas V costs [wikipedia.org].
ULA's cost summarized (emphasis mine):
ULA suggests that customers will have much lower insurance and delay costs because of the high Atlas V reliability and schedule certainty, making overall customer costs close to that of using competitors like the SpaceX Falcon 9.
Close. Not equal, not less. Just "close".
Thus, SpaceX's re-usable rockets continue to beat ULA in price, despi
What does ... (Score:4, Insightful)
Re: (Score:2)
your cost to use it could be lower than your cost per mile owning your own vehicle
Poor person detected. I don't care what my mileage is. I can afford to have enough cars sitting in my garage so that my use factor is less than 1%. If cost is an issue, then pass right by the self driving fleet and jump on the bus. That's even cheaper.
Another thing: If cost and safety are important, then why aren't we seeing proposals and prototypes for self driving city buses?
Three seperate issues that Cringely combines into (Score:4, Insightful)
Those three are: 1) Legal Liability. 2) Fear of the new, and 3) AI not correctly predicting human stupidity.
Issue 1 is a political lobby away from being removed. Some company, like Tesla, develops the technology and pays a political lobby to get Congress to allow the company to self insure all the vehicles they sell - either in the new market or the 2nd hand market - if Tesla sells the used car. Boom, that issue vanishes.
Issue 2 is irrelevant. The first AI vehicles will be commercial vehicles, not cars. Long Haul trucking, Garbage Trucks, Public Buses, will be the innovators. Their companies will all pay attention to long term cost - including the cost to hire drivers, and ignore the human fear. Or they will be beaten in the market by companies that do.
Issue 3 is the only problem without an obvious solution, but technology is ALREADY better than a human in almost everything except this issue. The benefits of slower driving and fewer stupid human first hand drivers already outweigh the lessened ability to predict what a stupid human in the other car will do.
Re: (Score:2)
The benefits of slower driving and fewer stupid human first hand drivers already outweigh the lessened ability to predict what a stupid human in the other car will do.
Or the stupid human cyclists. Or pedestrians. We get those off the roads and maybe autonomous cars will have a chance.
It will happen in Scandnavia first (Score:5, Insightful)
Re: (Score:2)
Re: (Score:2)
Self-driving... no (Score:2)
Humans after 100 or so years are incapable of driving safely, so why would one expect said humans to be able to build a self-driving car that drives safely?
Re: (Score:3)
Re: (Score:2)
Look until a self driving car can get fired, got to a bar, get shitfaced then drive home at 90 miles an hour then there's no way I'm getting in one.
He's not wrong: (Score:3)
"The problem isn't with the self-driving cars, it's with the cars that aren't self-driving, cars that are driven by idiots like me."
I've been saying for a while now that real adoption of self-driving cars will not be driven by early adopters, or even legislation. If Google/Waymo is even half right about the reduced accident rate of self-driving cars, the watershed will be driven by actuaries. Because a reduced accident rate also means a reduced insurance payout rate. Once enough of the data are in, the insurance companies will do the math and start raising the rates on human-driven cars. Once that drives a few more people to self-drivers, they'll have even more supporting data. And the insurance premiums will wind up set such that it will be prohibitively expensive for humans to drive their own cars; except, perhaps as a weekend hobby. But certainly, once the actuaries do their thing, none of us will be manually driving for our daily commute or errands.
Re: (Score:2)
When people no longer accept traffic deaths (Score:4, Insightful)
Re: (Score:2)
We accept this stupidly high number of deaths and injuries because we don't see an easy way out.
Or we as a society accept it because we consider driving our own cars to be a form of freedom. Freedom always has a cost, and today we are willing to pay this cost. The question in my mind is if society will change to the point where the cost of (self driving) freedom becomes too expensive (in terms of accidents) and we decide to give up this freedom. Personally I place a very high value on this freedom - the idea that a government or company at a whim could prevent my car from going where I want it to go m
No (Score:2)
No. The problem is not the human operated cars. I've been in a Tesla on autopilot, and it would be fucking road armageddon if that was all cars.
Rather obvious (Score:3)
That he even needs to point this fact out just speaks to the general climate of stupid cheering for things people do not understand. Full self-driving is still at least 10 years in the future and general availability more like 15...20 years. May also take quite a bit longer.
Re: (Score:2)
Re: (Score:2)
He is. I think he knows that but thought this had to be pointed out again.
self-driving busses on the other hand (Score:2)
They're turning them loose in crowded areas near me!
https://wtop.com/dc-transit/20... [wtop.com]
The areas described here are both several city-block "town square" type of deals,
bustling with pedestrians on small streets and sidewalks, with restaurants, shopping,
movies, and (at the onewith the subway stop) also apartments. Street and garage
parking, random double-parking and curb pickup. Traffic is normally about 10-15 MPH
and is signed for 25 MPH; these two areas are about 2 miles from each other and are
connected dir