Slashdot Mirror
Elon Musk Rolled Out Autopilot Despite Engineers' Safety Concerns, Says Report (theverge.com)
An anonymous reader quotes a report from The Verge: When Elon Musk announced last fall that all of Tesla's cars would be capable of "full autonomy," engineers who were working on the suite of self-driving features, known as Autopilot, did not believe the system was ready to safely control a car, according to the Wall Street Journal. The WSJ report sheds more light on the tension that exists between the Autopilot team and Musk. CNN previously reported in July that Musk "brushed aside certain concerns as negligible compared to Autopilot's overall lifesaving potential," and that employees who worked on Autopilot "struggled" to make the same reconciliation.
A major cause of this conflict has apparently been the way Musk chose to market Autopilot. The decision to refer to Autopilot as a "full self-driving" solution -- language that makes multiple appearances on the company's website, especially during the process of ordering a car -- was the spark for multiple departures, including Sterling Anderson, who was in charge of the Autopilot team during last year's announcement. Anderson left the company two months later, and was hit with a lawsuit from Tesla that alleged breach of contract, employee poaching, and theft of data related to Autopilot, though the suit was eventually settled. A year before that, a lead engineer warned the company that Autopilot wasn't ready to be released shortly before the original rollout. Evan Nakano, the senior system design and architecture engineer at the time, wrote that development of Autopilot was based on "reckless decision making that has potentially put customer lives at risk," according to documents obtained by the WSJ.
A major cause of this conflict has apparently been the way Musk chose to market Autopilot. The decision to refer to Autopilot as a "full self-driving" solution -- language that makes multiple appearances on the company's website, especially during the process of ordering a car -- was the spark for multiple departures, including Sterling Anderson, who was in charge of the Autopilot team during last year's announcement. Anderson left the company two months later, and was hit with a lawsuit from Tesla that alleged breach of contract, employee poaching, and theft of data related to Autopilot, though the suit was eventually settled. A year before that, a lead engineer warned the company that Autopilot wasn't ready to be released shortly before the original rollout. Evan Nakano, the senior system design and architecture engineer at the time, wrote that development of Autopilot was based on "reckless decision making that has potentially put customer lives at risk," according to documents obtained by the WSJ.
This is MBA behavior, not engineer behavior. I didn't see anything in the newsfeed about it though. I guess the transformation was inevitable though, in hindsight. Someone remove his pocket protector and have him turn in his scientific calculator on the way out.
And people wonder why I don't think arrogant salesmen like Musk should have the power to make OTA software updates without my permission.
.... not disputing that people may have had a valid point here, but anyone who's tried to execute a change or deliver an outcome will always find one or two dissenting voices in any organisation of scale. Once you take away the marketing fluff, what are the actual facts or reasoning behind their concerns? In my mind its' a discussion that runs very similar to:
1) Company develops new anti-cancer medication. Reduces mortality by 85%.
2) In 5% of cases people will die from a non-cancer related reason, as a result of taking the medication.
Musk appears (though I'm struggling to find specific evidence to support) to have suggested that the net balance is in favor of the roll-out, because you can never deliver anything both perfect and safe. Engineers tending to be cautious in nature, in small numbers have reservations about #2. I think some more clear messaging within the vehicle when you engage autopilot would put this to bed:
'You must remain attentive and prepare to take control of the vehicle at any time'.
It already stops you from taking your hands off the wheel for more than a few seconds, after a few folks have abused the system previously (the chap watching Harry Potter and eating lunch springs to mind) - so aside from semantics around marketing, this feels like lawyers trying to put the screws on Telsa for not shipping a product that prevents people ending themselves through idiocy.
Denials notwithstanding, a deep neural net is just a fancy rule-based expert system. The rules are in the form IF A THEN B where A is a pattern and B is a category or label. It's GOFAI redux. Deep nets suffer from the same problem as all rule based expert systems: they are brittle. Given a situation for which there is no rule, they fail catastrophically.
Full driving autonomy will not happen until AGI gets here. Unfortunately for Musk, nobody knows how to do that.
Five magic words make this controversy go away;
Its just like plane autopilot.
Elon, you are off the hook.
The current hardware doesn't have side facing cameras (or lidar or radar) on the front sides of the car. There are cameras in the side front fenders (in the logo) but they are looking backwards. They need to have side cameras close to the front of the car and high up as possible because before proceeding onwards from a Stop sign you need to see what's coming at you from the left or right side. The camera mounted in the middle post of the windows doesn't have an adequate view.
through no fault of his own
Are you referring to that idiotic youtuber? Through what leap of logic does one conclude that he wasn't at fault?
Ezekiel 23:20
Can you please focus on making your electric cars and the batteries that run them better and more affordable instead of getting sidetracked with these bullshit features that nobody wants?
Thank you,
Everyone
"Nine times out of ten, starting a fire is not the best way to solve the problem." - my wife
Your mom is just a fancy rule-based expert system.
Ezekiel 23:20
Human's aren't that great in situation's for which they don't have experience either...
That the fountain of BS that is Elon Musk allegedly put personal profit ahead of people's lives?
It seems incredible to me that any engineer who values their career would do any work for him in any shape or form.
I've heard him described as "a latter day Edison".
Unfortunately, Musk knows as much about engineering as Sonny Bono did about downhill skiing.
I think he's a naturally slothful person, sluggish and indolent, a dawdling flaneur, content to waste his life spread eagled on pillows forever indulging himself in the pleasures of the palm.
TL;DR: A complete wanker.
The Machine stops.
We are amazingly great in new situations. In fact, most of the patterns we experience on a daily basis are encountered for the first time. We know how to generalize.
If the potential pay out from the crashes that may happen is less than the money gained from the boost in sales due to this feature then musk made the right choice..
after all you must break a few eggs to make an omelet.
and for those that are going to hate all over elon for this, why arent you grabbing your pitchforks and going after all of those pharma CEO's? the side effects of some of the medications i have heard about are ridiculous.
Simple fact, it is impossible to make a 100% fool proof autopilot given the technology we currently have, either we take a risk and learn from the mistakes we are bound to make or we perpetually put it off until it is perfect (which is effectively never)
capcha: quagmire
Really, who isn't?
There is always people that will say something is unsafe and risky. It's also true that everything has some risk. Doing nothing and sitting in a cave has risk.
When you buy any vehicle you assume some risk. If you are buying a vehicle and believing it to be 100% risk free you are an idiot who shouldn't be on the road.
Trump?
Ezekiel 23:20
There's been more people than those mentioned here who left Tesla. Chris Lattner is one.
What about the patterns we experience while driving? I think they are fairly common. In exceptional circumstances, stopping or swerving to safety sounds reasonable, and there's no reason to believe we need AGI to reach human level capabilities there...
His " full self-driving car" (by Musk's own admission) drove into the side of an 18 wheeler in the middle of the day with perfect weather conditions.
Seems pretty clear to me.
If it doesn't autonomously pilot the vehicle, it's not autopilot and should not be called such.
And autopilot for planes is far, far, far more advanced, capable, robust, and reliable than the shit Elon is selling. The people using it are explicitly trained on what it can and can't do. People buying Teslas only get the marketing speak shoved in their face.
Human's aren't that great in situation's for which they don't have experience either...
Like pluralizing thing's?
It won't matter because AGI will mark the extinction of humans.
That's hardly the only time. There was an incidental in China where autopilot drove full speed into a road sweeper that it apparently couldn't see.
Tesla seem to have admitted it doesn't work as originally advertised, by repeatedly increasing the amount of effort it makes to keep the driver alert.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
Wrong!
He is very disciplined, he has the best rules.
Better yet don't call it autopilot. Even if you try to explain it, there are plenty of fools who won't get it. Call it "drive assist" and people might be a little less foolish. Some assholes will still misuse it, but you can't stop someone hellbent on stupidity.
Actually, I have a lot of experience with grammatical mistakes.
Case in points
I can come up with a few examples that I STRONGLY suspect don't meet that standard... Mostly they drive cars around here, but some of them are in public office...
"File to fit, pound to insert, paint to match" - Aircraft Maintenance 101
We know how to generalize.
Hmmm... Ain't that the truth...
"File to fit, pound to insert, paint to match" - Aircraft Maintenance 101
You know, a country with money's a little like a mule with a spinning wheel. No one knows how he got it and danged if it knows how to use it.
Heh-heh, mule.
The name's Musk, Elon Musk. And I come before you good people tonight with an idea. Probably the greatest... Aw, it's not for you. It's more of a China idea.
Now, wait just a minute. We're twice as smart as the people of China. Just tell us your idea and we'll give you subsidies for it.
All right. I'll tell you what I'll do. I'll show you my idea. I give you the Tesla Autopilotl!
I've sold autopilots to Plymouth, Oldsmobile, and Studebaker, and by gum, it put them on the map!
Well, sir, there's nothin' on earth like a genuine bona-fide electrified one-car autopilot! What'd I say?
Autopilot!
What's it called?
Autopilot.
That's right! Autopilot!
Autopilot, autopilot, autopilot, autopilot, autopilot...
I hear those things are awfully new.
It's user tested, but not by you.
Is there a chance the car could crash?
It's not your life, so splash the cash.
First adopters must be brave...
They'll be given early graves.
Will this venture fund new green jobs?
No, good sir, I'm the new Steve Jobs.
We've killed off our whole space program.
Fund my SpaceX, my good man.
I swear, it's the country's only choice! Log in to PayPal and raise your voice!
Autopilot
What's it called?
Autopilot
AUTOPILOT!
But the economy's still all fucked and broken.
Subsidies, this man has stolen!
Autopilot
Autopilot
Autopilot!
Autopilot!!!
Auto...*CRASH*
Except those words are false. It is NOT just like plane autopilot. Your Tesla on autopilot going down highway 101 is in a vastly different environment than an airplane flying on autopilot in the middle of the Pacific ocean at 30,000 feet altitude.
Plane autopilot is safe because it is well understood by all parties that it is not to be engaged in an environment containing tractor-trailers and pedestrians in your path. Telling people Tesla autopilot is just like plane autopilot is dangerous snake oilmanship.
If it doesn't autonomously pilot the vehicle, it's not autopilot and should not be called such.
And autopilot for planes is far, far, far more advanced, capable, robust, and reliable than the shit Elon is selling. The people using it are explicitly trained on what it can and can't do. People buying Teslas only get the marketing speak shoved in their face.
When any autopilot specifically informs a human driver to put their hands on the wheel and pay attention, and the fucking idiot behind the wheel ignores that, I'd say it's pretty damn obvious who and what is at fault.
Airline pilots are intelligent and highly trained individuals. That is why they are not found on every street corner, and are worth more than a dime a dozen. The main mistake Elon made was assuming that the average driver wasn't a fucking idiot.
No, he was instructed not to do what he did. And he did it anyway. Seems pretty clear to me.
Ezekiel 23:20
I was being sarcastic. I guess I'd better go ahead and state that now
And thus she earns a label?
Browsing at +1 - no ACs, I ignore their posts. So refreshing!
I don't have the exact words Musk has used, but I distinctly remember that he said that all Teslas will come equipped with the HARDWARE necessary for fully autonomous self-driving (computer power, sensors), but that the actual functionality would depend on a future software upgrade.
Now you and I, as software-related techies most of us, know that that will have to be one MASSIVELY COMPLEX and not really invented yet by any stretch of the imagination software upgrade, but technically, what he said is not false.
Where are we going and why are we in a handbasket?
"Is it safe?" is the wrong question.
The real question is: "are the fatalities, injuries and accidents that occur per passenger mile greater or less when compared to a human driver?"
"Safe" is a subjective and unmeasurable term, unless you define it in unreasonable terms (for example if your definition of safe is zero accidents).
The real "Libtards" are the Libertarians!
I've been in engineering organizations releasing new products that had life saving or threatening potential. It is always an agonizing, scary hard call as to when you've passed the threshold of risk.
There is a bell curve with a peak. You rarely hit the peak. If you make the call too late, you cost the lives of those you might have saved - too soon, you cost lives of those who might have saved themselves.
Even if you hit the peak perfectly, you'll always be able to truthfully argue that some people are being saved who would have died and some are dying who would have lived. The peak is a point of balance between the two - not a perfect elimination.
I can remember many times hating my bosses when they released a product that I didn't feel was ready. As an engineer, I have to be over-focused on the problems and stand no chance of seeing when I am perfectly perched on that probability peak. They had to pry the projects from my hands to get them out the door. I actually begged in tears once. But, in retrospect, I can't think of any case where my bosses weren't right in releasing the product that I was concerned about releasing.
What we need to force progress is for attorneys to get smart and start figuring out how to file more effective suits for lack of progress toward autonomy. How many are dying today because we don't have it? We need to focus hard on that.
artificial deep neural nets know how to generalize too. What's your point?
Where are we going and why are we in a handbasket?
And autopilot for planes is far, far, far more advanced, capable, robust, and reliable
That's mostly because it is also much simpler. The comparison with it is correct, it just omits the part where you have to additionally recognize roads, vehicles, traffic signs, pedestrians...all the things airplanes don't have to worry about. Airplane-level capability just isn't enough.
Ezekiel 23:20
While not entirely surprising, it's disappointing that Musk has promoted it in the way it has. It's obvious to most people that it shouldn't be though of as a fully autonomous system, for several reasons.
When the car can drive in snow, heavy rain, fog, or around road construction let me know. As it is you can't even compare autonomous cars to people it's like gangsters to space stations as the only miles on self driving cars are sunny freeway miles and carefully planned city routes. No study I've seen compared them in the same situations.
I have a better plan.
Kill Musk, take his money, keep it.
licensed engineers may be need for autodrive software or something like it.
The FAA does code audits on autopilot software.
If it doesn't autonomously pilot the vehicle, it's not autopilot and should not be called such.
It autonomously pilots the vehicle to about the same extent an airplane autopilot does. In neither case is the pilot or driver free to stop paying attention. The pilot is expected to monitor the behavior of the autopilot, which follows simple instructions such as, "fly a track to this radial to this VOR", or "climb to FL30 and level off, increasing speed to mach 0.81". It can chain those together and fly a preprogrammed route, but no matter what, the pilot must maintain situational awareness. Autopilots can and do fail, or develop problems such as entering a descent when level flight is commanded. Autopilots can and will disengage if something happens they aren't prepared to deal with.
Autopilot here means the roughly what it means in airplane land, which is why the name is tolerably apt. "It will help you with the drudgery of steering the vehicle around, but you the human must pay attention to what's happening and be prepared to intervene".
It's a lot simpler than that:
This article is bullshit.
Sorry to be so blunt, but it's journalistic malpractice. The author is confusing Enhanced Autopilot (EAP) with Full Self-Driving (FSD). To be clear:
* Some safety features related to autopilot, such as automatic braking and the like, are available to everyone for free.
* EAP is an optional add-on available today ($5k on the Model 3 if purchased at the time of buying the vehicle, $6k as an over-the-air upgrade) which provides lane-following (requires hands on the wheel and driver attention) and driverless summon features (very low speed, "back out of / into a tight space / drive down the parking lot" stuff without a driver in the car). More to the point, there are two entirely different versions that have existed over the year. AP 1.0 was used on earlier vehicles, based on software and hardware from Mobile Eye. Tesla and Mobile Eye split in a contract dispute. Mobile Eye claims that Tesla wasn't using their hardware right. Tesla says that Mobile Eye found out that Tesla was working on an in-house Autopilot system and demanded that they stop as a precondition to get to continue to use their hardware. Mobile Eye says they knew about Tesla's internal work but didn't feel threatened by it. Regardless, Tesla was forced to switch to their internal version, AP 2.0, which was a step backward. AP 2 is just now catching up to the features of AP1.
* FSD is Tesla's current goal, where the vehicle can drive itself without you having to have your hands on the wheel or paying constant attention. You cannot use FSD, even if you buy it. It costs $3k on the Model 3 ($4k as an over-the-air upgrade later). The article is talking about FSD being rolled out before engineers think it's ready. To reiterate: you can only buy FSD right now, you can't use it until it's ready. Tesla apparently tried to clarify this for the author:
The author apparently nonetheless still failed to understand what that means. You Cannot Use FSD. Period. If engineers are complaining about FSD being rolled out too soon, they're complaining about Tesla selling something that its drivers aren't going to be able to use for too long of a period of time. And you know what, I fully agree with the engineers in that regard - I think it's wrong of Tesla to sell something that there's a big question as to whether they'll be able to get it working reliably enough or pass the serious regulatory barriers in its way.
But if engineers are complaining about FSD, then it's not complaints about EAP. Because the two are very distinct things. EAP isn't perfect, don't get me wrong - and the 1.0 / 2.0 switch was a big setback (they still don't use all of the cameras on the vehicle). But it also pesters drivers enough if they show signs of not paying attention to the road (e.g. not holding onto the wheel) in order to overcome its imperfections (the level of pestering was significantly increased after AP1's fatal accident, in which the driver was apparently watching movies during most of his trip).
Ever since, I've been suspicious of Jesus and very careful around chlorine.
This is a prime example of what I'm talking about when I say so-called 'self driving cars' are being rushed to market.
NONE of them are really ready and won't be for quite some time -- if ever.
Meanwhile people really don't want them anyway.
There's also the motorcyclist in Norway that was run over from behind by a tesla that couldn't detect it.
Chris Lattner wasn't there long enough to get started. We don't know why be backed out.
Personally I could never see why a compiler guy was being hired as head of one of the most complicated AI projects anyway. Different field.
Electric cars cool, self-driving cars bad.
If Musk really wants to "save the planet", drop the self-driving crap already. It makes the car more expensive so less people can afford one, meaning they keep their old polluting car or even buy a brand new polluting car.
#DeleteFacebook
"he was instructed not to do what he did"
The car just plain didn't see anything, and your talking about a much more recent version of Telsa's software/hardware, and as AmiMoJo said:
"Tesla seem to have admitted it doesn't work as originally advertised, by repeatedly increasing the amount of effort it makes to keep the driver alert"
Neural Nets are very specifically NOT rule based. They are trained.
GOFAI was pretty much a phrase invented to label stuff that IS NOT the neural net approach.
Autonomous vehicles do not need AGI. It's very much a single domain system. You don't need your autonomous car to be able to diagnose diseases for example.
Musk "brushed aside certain concerns as negligible compared to Autopilot's overall lifesaving potential,"
Popular /. talking point right there. No wonder so many fanboys. Musk is the Donald Trump of tech.
I don't see how explaining that Telsa is now officially selling something that doesn't yet exist helps their record of misleading the public on "auto-pilot".
Oddly, planes don't autopilot to the terminal. Why? It's just parking. That's a solved problem.
Now I understand.
When Musk opposes autonomous killing machines, he's just trying to reduce the competition. He wants a monopoly!
No, he was instructed not to do what he did. And he did it anyway. Seems pretty clear to me.
Yes, two things are very clear;
1) The driver was irresponsible and legally at fault
2) Tesla autopilot was not good enough to stop the car from plowing into the truck on its own.
And autopilot for planes is far, far, far more advanced
This comment demonstrates an glaring ignorance of how plane and automobile autopilots work.
And autopilot for planes is far, far, far more advanced, capable, robust, and reliable than the shit Elon is selling.
That's total bullshit. Autopilot for airplanes has been around for many decades now. It just maintains a heading and altitude. It's roughly analogous to cruise control on cars in technological terms, and maybe automatic lane-keeping in actual functional terms (since cars have to follow a road, planes don't; of course, technologically, lane-keeping is far, far, far more advanced than the autopilot in a typical Cessna).
Yes, there are very advanced autopilots in today's newest passenger planes like the 787, but the term is not exclusive to those, and there's countless decades-old Cessnas and Pipers out there with autopilots that are quite primitive.
No, autopilot in planes does not autonomously pilot the plane. It doesn't take off or land, it doesn't fly around bad weather, it doesn't check METARs and PIREPs, it doesn't watch for other traffic, it doesn't handle radio calls to ATC when you cross into class B airspace, it just keeps you flying straight and level.
The only thing in your post that's correct is the bit about pilots being trained to use their equipment. That isn't true for Teslas, but it also isn't true for any other car either. How many drivers on the road today got explicit training to use the cruise control in their car? Cruise control has been around a few decades now too. Or how about the more advanced features we have not, like adaptive cruise control, lane-keeping, and operating the infotainment system? Every car is different, with different controls and different quirks. Airplane pilots aren't even allowed to fly a plane (solo) unless they've been specifically trained for that model and received a rating for it. Perhaps we should do that for cars....
Airline pilots are intelligent and highly trained individuals. That is why they are not found on every street corner, and are worth more than a dime a dozen.
Actually, this isn't true. Pilots start out their careers as instructors ("those who can, do, those who can't, teach"), and make peanuts. After that, they might get a job as a copilot for a regional jet company. Last I heard, the starting salary for one of these guys is $18k. Yep, barely above minimum wage. It takes many years for them to work up to any kind of decent salary approaching 6 figures. Then, when they hit 60 years old, they're forced to retire.
Being a pilot is for people who are independently wealthy (e.g. trust fund, or has a spouse willing to support them), or for people who love it so much they're willing to sacrifice everything just to have that job.
The one that happened before this, using Tesla's prior "auto pilot" technology? And who ignored repeated warnings from the system to put his hands on the damn wheel and take control before a crash occurs?
They implied that Tesla is currently having people drive something that its engineers deem unsafe. This is simply not the case at all. If the engineers were complaining about selling FSD, they're not complaining about anything that consumers are actually driving.
Everyone who buys FSD does so on their assessment of how likely they think it is that Tesla will actually deliver. There is zero confusion among anyone who buys it about the fact that they can't use it right away; the option always includes the "you can't use this until it's finished and legally approved" disclaimer next to it. It all comes down to how optimistic or pessimistic you are about the technology. I'm a pessimist, and will not be buying it. Some of Tesla's engineers working on it are apparently also pessimists. I'm not surprised. It's a crazy-hard task, and very different from human-supervised autosteer / EAP.
Ever since, I've been suspicious of Jesus and very careful around chlorine.
What if there are two people on different parts of the road in front of the car, and swerving to avoid one will mean hitting the other? The car needs to be able to diagnose which of the two people has a terminal disease, in order to select to hit that one.
This space intentionally left blank
Adversarial patterns prove that you are wrong.
Actually that's false when it comes to new *dangerous* situations. Research has shown that most people's minds take them to what they know when their lives are on the line, even if it runs counter to survival.
If the complexity of the tax code keeps increasing then yes, AGI will mark the extinction of Americans, at least, due to how long it will take to actually compute that number.
The author is confusing Enhanced Autopilot (EAP) with Full Self-Driving (FSD).
Every single person I talk to, who only casually observes Tesla-related news, assumes that EAP == FSD. Every friggin single one. Usually in knee-jerk comments, when I mention something about having to drive my car somewhere. (I have a Model S /w AP v1... Love it in stop-and-go highway traffic and long highway trips, but I'm quite well aware of its capabilities and limitations.)
Might be useful, actually. You pull over and the hooker gets in. Car scans for disease and automatically hits the eject button before you pay.
Aeronautical autopilots are not safe, precisely because they have significant nuances and problems which have resulted in numerous crashes through misuse and over reliance. You do not turn on an aircrafts autopilot and then sit back and relax for the rest of the flight.
Which is pretty apt considering what we are discussing and what is being claimed on all sides...
Oh, trolly problem time. What if the terminally ill person has several years to live, and the other person is suicidal and wants to be run over?
LOL.... and what does the training do? Nothing more than program a set of rules which we are not able to even understand or decompose (that's why NNs can't be used in some applications, where the consumer must understand WHY the specific outcome).
Whenever someone starts getting sanctimonious about safety I ask them if they fit the best possible tyres to their car for the next journey. If not then they are prepared to sacrifice safety for cost or convenience.
^--- This. My brother is a pilot. My parents spent about $100k sending him to Embry-Riddle 25 years ago, then spent ANOTHER $100k renting planes for him so he could get his 1,000 hours of multi-engine time and actually become employable by an airline.
The thing that really sucks about being a pilot is that you can't independently afford the training & certifications you need to get hired, but airlines won't pay for the training & certs until after they hire you
It's kind of like the catch-22 film students had prior to cheap HD video... nobody would fund your films until you were good, but you couldn't afford the gear, film, and services to BECOME good in the first place unless you and/or your parents were independently wealthy.
3rd article questions if Tesla can see motorcycles:
"Tesla Model S with Autopilot rear ends motorcyclist because the system didn’t “see” the rider"
I think the "on it's own" it really the thing.
If a driver can detect 95% of hazards, and the car can detect 70% of hazards, then there is improved security, because they both would have to miss a hazard. But switching completely from the driver in charge to the car in charge decreases security.
Kinda reminds me of some people I encountered in IT back in the 90's. "Hey, I don't need to backup anymore, I have RAID now"
Better yet don't call it autopilot. Even if you try to explain it, there are plenty of fools who won't get it. Call it "drive assist" and people might be a little less foolish. Some assholes will still misuse it, but you can't stop someone hellbent on stupidity.
Agreed, this was a marketing fiasco of Tesla's making. The technology is good, but they marketed it as something it wasn't, and at least one guy died.
Fast Federal Court and I.T.C. updates
Agreed that Neural Nets are not "rule based", but I do think they need far higher level of intelligence than anything we've seen to date. The car is going to need to identify traffic police and follow instructions, they need to interpret human instructions. If they get on the loud-speaker and say "pull-over" vs. "pull-over at the next exit", they'll need to understand that. Not saying it's impossible, just many years off.
Fast Federal Court and I.T.C. updates
What if there are two people on different parts of the road in front of the car, and swerving to avoid one will mean hitting the other? The car needs to be able to diagnose which of the two people has a terminal disease, in order to select to hit that one.
What if the two people are healthy, but the car can quickly identify which of the two has a higher net worth, so it can hit the poorer one. No moral problems there at all.
I know it's a currently an absurd result, but eventually, the computer will have to decide, or decide to ignore their net worth and use other factors, or to just flip a coin and hit one randomly. Every decision comes with it's own set of moral issues.
Fast Federal Court and I.T.C. updates
It won't matter because AGI will mark the extinction of humans.
I'm pretty bearish on AGI. To get it in our lifetimes, it'll require continued exponential increases in processing power, i.e., it requires Moore's law to continue. However, many signs currently indicate that Moore's law is dead. It's entirely possible that our civilization may be hitting a technological plateau and that real AGI is not a few or dozens of years away, but hundreds or thousands.
Fast Federal Court and I.T.C. updates
If the complexity of the tax code keeps increasing then yes, AGI will mark the extinction of Americans, at least, due to how long it will take to actually compute that number.
The tax code is large and opaque to most, but it has a bit of organizational beauty to it. Most people don't understand it because they never studied it. But if you can write software, you can understand tax law.
Fast Federal Court and I.T.C. updates
licensed engineers may be need for autodrive software or something like it. The FAA does code audits on autopilot software.
It's not exactly a code audit, it's more like the FAA certifies code to a certain level of robustness. For commercial airline software to get certified, they generally have to prove that every line of code has been covered by tests, and every branch has been taken and not taken. There are even higher levels of certification (usually for the OS), where the code must be symbolically expressed, and mathematically proved to be correct. Not an inexpensive undertaking.
Fast Federal Court and I.T.C. updates
Autopilots in planes have been able to land and exit the runway for quite a while. I'm pretty sure they should be able to take off as well if the plane is well aligned at the start of the runway, but they can't taxi (drive the plane on the ground to the start of the runway).
If they are well trained on a dataset that covers all the relevant situation, they can interpolate, but they cannot really extrapolate all that well. No known machine learning technique know how to deal with a completely new and unexpected situation.
In other words we don't know.
More specifically, they are indeed rule based, but the rules are learned, not engineered.
Agreed, this was a marketing fiasco of Tesla's making. The technology is good, but they marketed it as something it wasn't, and at least one guy died.
Actually, they marketed it as what it was, and that was confusing for people, and at least one guy died expressly ignoring the warnings of the manufacturer, of which he literally could not possibly be unaware since he had to attend a safety lecture before he was allowed to use the feature. He ignored the warnings of the people who produced the system, and that is what killed him.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
I'm a bit worried by their FSD technology. The current auto-steering/speed control needs to use map data to work, i.e. its maps tell it where sharp bends are so it can slow down, that sort of thing. For full self driving it needs to be able to work without accurate maps, e.g. there might be some roadworks or a new road layout that isn't on the map yet.
At best the car would be forced to stop, possibly stranding the passengers if they were unable to legally drive it. At worst it might not slow down in time.
Maybe they will get rid of this requirement before release. The fact that they are selling the feature right now is worrying though, because either it's screwing people over with something that won't be released for years, maybe even before they sell/scrap the car, or because they are planning to release it way too early.
const int one = 65536; (Silvermoon, Texture.cs)
SJW, n: "Someone I don't like, and by the way I'm a fuckwit" - AC
Autopilot in planes is far simpler than what is needed in cars, that's why it came decades ago. If a car had a plane's autopilot it would have cruise control and would drive in a straight line.
I think the "on it's own" it really the thing.
If a driver can detect 95% of hazards, and the car can detect 70% of hazards, then there is improved security, because they both would have to miss a hazard. But switching completely from the driver in charge to the car in charge decreases security.
Kinda reminds me of some people I encountered in IT back in the 90's. "Hey, I don't need to backup anymore, I have RAID now"
Agree. That last part, the human factor element, is the hard part. As people feel safer they naturally take greater risks. We all do it, some more than others. Its those 'some' that do it more that we have to worry about.
Well that's debatable, and I would come out on the other side of that debate.
But it's neither here nor there, because GOFAI is specifically done with Human readable rules. And there's no one suggesting Neural Nets can be interpreted as that. That is not debatable.
Autopilots in planes have been able to land and exit the runway for quite a while.
Some can, many can't. It's just like cruise control in cars: some can detect cars in front of them and slow down ("adaptive cruise control"), most aren't at all that advanced.
I'm quite sure that all the people with 60s-era Cessnas do not have autopilots capable of auto-landing.
Nah, you just hit both of them. Line up on the most open path and hope for the best.
Actually I think you just need to be able to recognise emergency vehicles that have their flashing lights on. Pretty easy as they have lots of markings in addition to their lights.
When one of them is behind you, as a human driver you find a safe place to pull over and slow down so it can pass. If instead of passing it keeps following you, then you know you are the one they're after and if you;re law abiding, you come to a stop.
Automated driving can do the same.
Being a pilot is for people who are independently wealthy (e.g. trust fund, or has a spouse willing to support them), or for people who love it so much they're willing to sacrifice everything just to have that job.
Or for people who are willing to do a stint in the military. That's the best route to becoming a commercial pilot: Join the Air Force and let Uncle Sam pay for your training and flight hours. If you're smart, go for cargo jets rather than fighters, because after an eight year hitch as a C-5 pilot (for example), you'll have lots of heavy multi-engine hours which are very expensive to buy.
Note to ACs: I usually delete AC replies without reading them. If you want to talk to me, log in.
Wow, wait until SpaceX's man-rated rockets.
Customers of those do not take recklessness well.
Kriston
Exactly. Deep learning experts can jump up and down and claim otherwise but they are still doing symbolic AI aka GOFAI. They just figured out a way to automate the production of the rules with fast computers, back-propagation and gradient descent.
"No known machine learning technique know how to deal with a completely new and unexpected situation."
Neither do humans, at least in car-operation time.
Try turning your head upside down and looking at the world. Hard to interpret, isn't it.
Where are we going and why are we in a handbasket?
Maybe the musk could define what his autopilot is?
Also various critters.
Of course let's ignore that people drive all the time in snow, rain, fog and road construction conditions when they are told it is unsafe to do so. Look no farther than Texas where hundreds if not thousands of cars are trapped underwater, when they were told to stay off the roads. Every time there is a big snow storm, with road conditions that are too hazardous for people to drive in there are hundreds, if not thousands, of people who ignore the warnings, drive anyway, and have accidents, some of them fatal.
People speed through construction zones and every year dozens of workers are killed.
A self driving car that refuses to move because the environmental conditions are too bad for it to drive is actually better than a manually driven car that allows a stupid person to race down a road that is too hazardous to be driven on. Typically we are on the cusp of having technology that is better than all but the best professional drivers, and it will not be able to drive under conditions that people insist on driving under. Conditions that are too hazardous for them to be driving, but of course they know better.
Actually, they marketed it as what it was, and that was confusing for people, and at least one guy died ....
If a company markets something that confuses people, it's the company's fault.
Fast Federal Court and I.T.C. updates