I hope it is heat tolerant
It can hit 45C ambient in Melbourne. The inside thermometer in my last car would top out at 55C, I suspect it was 5-10C higher than that.
Nvidia revealed its new Tegra X1 system-on-chip processor, and hardware to power autonomous cars, at “day zero” CES press conference. The new silicon features the Maxwell graphics engine, juiced up with eight CPU cores and 256 GPU cores, all available while consuming ten watts. Formerly known as “Project Erista”, the Tegra X1 …
Not even trying. When I was working in this field, back in the 1980s, the chief engineer of Caterpillar told us the underbonnet temperature in tractors could reach 150C in the Southern states in the summer, which made the design of ECUs rather difficult. Despite this, just about every modern Diesel outside a few small applications is now electronically managed.
Obviously there are mitigation methods that can be used, but there is no real problem in designing electronics for 110C operation.
See the comment from Cuddles.
Higher temperatures do require things like thicker oxide and a process that doesn't create small bottlenecks in tracks which cause IR heating and atom migration. But modern semiconductors start off a lot more consistent than they did back in the day when I was working with JAN stuff. As I recall there was a scandal when it turned out Nat Semi had been sticking commercial ICs in mil packages and selling them to Uncle Sam, and not only was nobody the wiser till it was noticed that the paperwork had been fiddled, but that year they won a quality and reliability award from the government.
ECUs do have CPUs, 32 bit and a few tens of MHz, but they aren't usually liquid cooled. There are all kinds of mitigation I can think of, from putting them in the inlet air path in Diesels (which draw in a lot of air), to the liquid cooling used in the Prius.
One thing I do know is that electronics are far more rugged than the current wetware used to steer cars, which starts to malfunction outside the very limited ambient of 0-30C and is very prone to EDI - emotion and drug interference.
A standard Nvidia Geforce GPU on factory settings has its fan set to keep the temperature at or below 90C (although Maxwell is a lot more efficient so the newer ones may try to keep to lower temperatures). The danger temperature not to be exceeded is generally around 110C. So no, 55C is not anywhere near enough for any kind of worry. In fact, it's closer to normal idling temperature than anything else. As others have mentioned, actual engine temperatures are likely to get higher than that so it's not quite simply a case of sticking a bog standard consumer GPU in there and assuming it will work, but "my car sometimes feels a bit hot" is really not any kind of sensible worry about whether in-car computers will have problems.
@Cuddles. Under a car bonnet (hood for the WSOPs) the temperature can reach over 200C with ease - all you need to do is drive around a bit and then park in tarmac car park on a calm but sunny day. The heat rising from the tarmac combined with the heat from engine and exhaust manifold easily raises the air temp under there. I've returned to my car twenty minutes after parking to find the cooling fan still running flat out in blighty before now.
Woah, 3 thumbs down. I wonder why? Guess the general level of interest in interesting IT has gone down the toilet and has finally arrived in the web circle of hell, populated by zombies emitting noises like "Hurrr AJAX WIDGET Durrrr". Oh well.
100 GFLOPS/W? Isn't that leaps and bounds ahead of all CPUs and GPUs out there?
Apparently Quadro K6000 is capable of "5.2TFLOPs/sec for single precision (FP32)" consuming 225 W. So about 23 GFLOPS/W. Double precision for the K6000 card would be 1/3 of that (7.7 GFLOPS/W). November 2014's Green500 #1 most efficient supercomputer achieved 5.27 GFLOPS/W.
So the real story here is the power efficiency of Arista/Erista/X1, even if it's mainly for graphics performance, mobiles, and tasks like vehicular vision rather than supercomputing. If they could combine X1's power efficiency with 1/3 FP64, you could build a 33 MW exaflop supercomputer (not really though). Some cards out there also do 1/2 FP64.
Edit: According to Anandtech it is 1024 GFLOPS of FP16 operations, 512 GFLOPS of FP32. I don't see 10W anywhere, but I see what looks like 64 "single precision GFLOPS/W normalized" on page 1. Feel free to crack the code.
This is one half of a teraflop as we know it. NVIDIA's slick marketing got them a slick headline. It's also called "Erista" not "Arista".
"11:12PM EST - Confirmed that it's Erista
11:12PM EST - NVIDIA is proclaiming it a 1 TFLOPS GPU, though this is at FP16 as opposed to the more normal FP32 metric for TFLOPS"
My thoughts exactly. Then add in the memory, cameras and screens they were talking about and wow, you've probably got a $10k-$15k bump in price after the Automakers get finished with the profiteering.
Because, you know, the "high tech" car keys/fobs they came out with in the 1990's still cost $200+ to replace...
>Huang said he expects mirrors could eventually become smart displays
My first reaction was "No! Wing mirrors are effectively a consumable, like tyres or brake pads", since on the lanes around here they don't last indefinitely.
However, if one were to replace a £20 mirror with a £100 display, it wouldn't have to be placed outside the vehicle. There would be other engineering challenges too - such as visibility in bright sunlight - but they can be overcome (e.g, by using a honeycomb filter aligned to the driver's eyes).
> A camera can be stuck at any convenient point
Indeed, the driver-side camera can be placed to eliminate the 'over the shoulder' blind-spot. I've been witness to a collision (thankfully nobody physically hurt) caused by a car driver moving into the faster lane of a motorway without looking over his shoulder (and thus he didn't see an articulated lorry).
I believe Volvo or Toyota tried using radar - giving vibration feedback to the driver through the indicator stalks - to warn drivers in this same situation. Unfortunately, it relied upon the driver actually using their indicators...
I rather read this as keeping the mirrors as is, but adding HUD to them so you look in your wing-mirror and it highlights the fact the car coming up behind you is an Audi travelling at 108mph so you can get out of the way! Combining real optics and extra information rather than just putting in a camera... so if there's a fault you can still see.
"Combining real optics and extra information rather than just putting in a camera"
So loads more development & homologation costs, higher complexity, and more expensive spare parts, just to draw attention to things an attentive driver should be able to see? I can't see much benefit myself, since the inattentive will remain that way. I think this is about a desire to eliminate the drag of wing mirrors that is widely quoted as being in the range 3-6% of total aerodynamic drag.
"As a former commercial pilot, please allow me to recall my human factors training and tell you that, if that is a reflection of your attitude, you are a dangerous driver."
How marvellously high handed of you, Biggles. Like you'd know anything about my driving (or I about yours).
Let's be clear about the benefits of automation. Take stuff clearly and completely out of a meatsack's hands and you often improve safety. Trivially automate some check, or half automate something, and you have the makings of an AAIB investigation. Look at AF447, where automation appears to have trained the pilots not to be able to fly. Or Kegworth, where automation to simplify the controls resulted in the crew shutting down the one working engine. Many accidents have been caused by crew over-riding safety warnings, or switching off automated systems.
So, going back to the post, how is additional info from "smart mirrors" to be usefully provided to the driver? A distracting text message flashing on his dash? A big, panic-inducing klaxon? A flashing amber light on the mirror?
Apparently you didn't even understand the pilot's post, you just took offense to his characterization of you as a dangerous driver since you, like about 95% of the population, believe yourself to be a good driver.
When you talk about "things an attentive driver should notice" it sounds very much like you are assuming that others see what you believe you would see in their situation. You don't practice defensive driving, but instead get upset when you sit in someone's blind spot for a bit and they attempt to change lanes directly into your car and you have to take evasive action. Since you were watching that guy and successfully took evasive action you think that makes you a "good driver".
A good driver avoids (to the extent possible) putting himself in situations where the actions of bad drivers will affect him or cause him to take evasive action in the first place. It seems some people like to put themselves into a position where they need to take evasive action, or get cut off, etc. due to the actions of poor drivers so they can honk or flip off other drivers and feel superior about themselves and their driving abilities. Sort of a passive aggressive version of road rage, I suppose.
> Apparently you didn't even understand the pilot's post,
> you just took offense to his characterization of you as a dangerous driver
Well, I didn't explicitly call him a dangerous driver. I did say that someone showing that attitude is dangerous, the reason being that we humans are very fallible, unreliable machines.
Driving (or flying) are not the sort of activities we are good at, since at a basic level they require a level of concentration that cannot be maintained for very long, causing our attention to fluctuate between the pure mechanics of driving (or flying) and the tactical picture. Sooner or latter our attention is going to be directed at the wrong subject and we are going to scare (or worse) ourselves and/or our passengers.
It is for this reason that anything that reduces our workload and allows us to take a higher level view of the situation (the sort of things humans are good at) is a welcome advance (of course, this should not be taken to exonerate the driver of any amount of responsibility--you are still at the wheel).
Your point about defensive driving is spot on, btw. I believe that defensive driving should be taught as part of the driving school curriculum, and regularly tested.
Wing mirrors are where they are not for ergonomic reasons but for line of sight optical reasons - a tradeoff between where the driver is, where the windows are, and where you need to see.
I would assume that the final display will be across where the centre mirror now goes and will be a combination of several cameras, with some way of highlighting potential dangers. I just look up slightly as usual and I have a 180 degree view around the back of the car.
As an emergency backup, two small folding door mirrors (no need for streamlining) that normally fit within the car perimeter, and a flip down mirror to act as a conventional centre mirror.
If someone like me can see a workable solution, the thousands of very bright vehicle designers and engineers can surely come up with a better one.
> I believe Volvo or Toyota tried using radar - giving vibration feedback to the driver through the indicator stalks
"I believe Volvo or Toyota tried using radar - giving vibration feedback to the driver through the indicator stalks - to warn drivers in this same situation. Unfortunately, it relied upon the driver actually using their indicators..."
I have that in my 2008 Mazda. Works well, but yes, to get the audible warning you do need to use the turn indicators so the car knows that you are trying to change lanes. There is a visual passive warning that shows up in your side view mirrors, if you glance at them before you make the lane change.
> Unfortunately, it relied upon the driver actually using their indicators...
I have a Dodge that uses radar to check blind-spots, and I think its means of signaling are an improvement over what you described. It always gives a visual indicator on the wing mirrors, and will sometimes give an audible alarm (if you use the turn indicators, or if you're in reverse). However, that arrangement means a distracted or slack driver can still miss all the helpful warnings. For examples...
> I've been witness to a collision (thankfully nobody physically hurt)
> caused by a car driver moving into the faster lane of a motorway
> without looking over his shoulder
...My brother had a driver try to move into the lane behind his large diesel pick-up truck without looking over her shoulder. She apparently only kept an eye on the truck beside and in front of her and thus changed lanes once she past his rear fender. This was unfortunate because he was towing a 9-meter boat on a sizable trailer, which is about as hard to overlook as an articulated lorry. (Thankfully the boat was unharmed, though the trailer lost a tire and the woman's car gain a soiled driver's seat.)
It's exactly the sort of situation that would've rendered current blind spot warnings moot: she didn't look at her mirrors and didn't signal her lane change. You can make warnings more intrusive to reach less attentive drivers, but the false positives or positives the driver expects are annoying - like Vista's overly-sensitive user account control.
> Wing mirrors are effectively a consumable
A modern car's wing mirror has got more electronics in it than an entire 20 year old car.
> However, if one were to replace a £20 mirror with a £100 display, it wouldn't have to be placed outside the vehicle
As I understand it, what the quoted chap had in mind was not replacing the mirror with a display, but embedding a display within the mirror, for displaying info on it (say for example, distance to the car behind or some such).
> A modern car's wing mirror has got more electronics in it than an entire 20 year old car.
a motor, a heater and some lights in the most high tech mirrors on the planet? my 44yr old car has all those, and even if you want to count individual electronic components, you still fall far short.
you know nothing about cars, and very little about electronics.
This post has been deleted by its author
Yes, one needs camera input as well to deal with things like fallen rocks and pedestrians that don't come with transponders. (I rather anticipate a future with self-driving cars when sane pedestrians and cyclists *will* carry transponders, possibly integrated into their mobile phones).
But consider optical illusions. Despite several hundred million years' evolutionary honing, there are still situations where either you cannot quickly work out what you are seeing, or you totally mistake what you are seeing. (The latter largely accounts for the "invisibility" of pedal cyclists to car drivers. "Think bike" has limited effect on the lower levels of our vision processing. I've seen two cyclists collide because "invisibility" applies equally to the cyclists themselves )
Do we really think that a computer vision system will be better than the human vision system after a mere few years' development? I doubt it.
"Why, yes, I guess we do."
Please don't expect us to take anything Ray Kurzweil says as being even within spitting distance of objective reality. The man has been absolutely full of it for years.
Recognizing an object is "A Car" rather than "A Table" is a damn sight easier (though admittedly still not easy) than recognising what make of car it is, model, maybe even year, any user added additions, number of people inside, state of its paintwork, windows, lights, which way its front wheels are pointing etc which is what you and I and any non blind human over the age of about 5 can do in fractions of a second.
The only time that kurtweil won't have exaggerated something is when he's been replaced by an AI copy.
Please don't expect us to take anything Ray Kurzweil says as being even within spitting distance of objective reality. The man has been absolutely full of it for years.
From which I gather that whatever appears on Kurzweil's website is automatically bogus, regardless of whether he's just reporting on research done by others?
I guess we should warn someone at MIT, then. Their Department of Brain and Cognitive Sciences has just been outed!
Also, I am curious: when you say "us", in whose name do you speak?
Recognizing an object is "A Car" rather than "A Table" is a damn sight easier (though admittedly still not easy) than recognising [a whole list of things mostly irrelevant in the context of autonomous driving].
Why, yes, and you could make pretty much the same point about how chess-playing computers lag so hopelessly behind humans in a host of cognitive tasks. Didn't do Kasparov any good, though.
"(I rather anticipate a future with self-driving cars when sane pedestrians and cyclists *will* carry transponders, possibly integrated into their mobile phones)."
Yes. And the Government are probably already considering it. "We couldn't get ID cards through... but what if we sell the idea of an electronic ID card system, that we could use to track everyone? We won't say that's what it is and that's what it's for, of course, we'll claim it's being introduced for the sake of road safety: a small device that everyone needs to carry so that in-car sensors in modern vehicles, not to mention self driving ones, can see them."
And the Government are probably already considering it
If you carry a mobile phone you are already doing everything you fear. They know whose phone it is, where it is, where it's been. It's quite likely they can use it to bug you. Even if you try to conceal your ownership of it, they can still cross-reference it to everyone you ever called with it and find out who you are with a minimum of further interrogation of their records.
A transponder could be far more anonymous than a phone (and cheapest possible would not be personalized).
I'm rather dubious about cars being able to discern the wide variety of road conditions anytime soon. Puddles? Boards or tire shards lying in the road? Golf ball or softball-sized rocks? (God forbid your car ignores larger ones than that) Broken glass? Ice or slick spots where water has come across the road? Ruts or potholes?
There are lots of things out there that I'd be pretty concerned about my car deciding that it could charge through at the posted speed limit.
Maybe in 10 years, but I don't see it for awhile. Plus there is the "hacking and tracking" risks.
@Marketing Hack - That's kinda the point of a learning computer - over time it will learn more and more things that it will recognise.
I don't know what that weird object on the road is until I'm either standing over top of it or have picked it up and looked at it. A computer is no different.
The human eye is notoriously easy to trick.
The most important application of this chip seems to escape notice. It can save lives. Lots of them. Imagine cars in the future being able to avoid accidents no matter how how impaired their drivers are. This chip foresees a time when the car will correct the mistakes of the intoxicated, the elderly, the texting and the just plain inexperienced drivers on the road. The other features, while notable, most certainly pale in comparison. We may never get bad drivers off of the road. But, a car that refuses to be driven badly is doable.
For starters I don't want or need an autonomous car and I suspect I'm not alone in that.
Secondly while I'm driving I don't need ANY visuals other than the dashboard and what I can see out of the windscreen. I don't need a 4K display refreshing at 60Hz for starters not do I need any other bells and whistles that car makers are shoving into cars in the mistaken assumption that most new cars are bought by 15 year old boys.
Manufacturers - how about you just provide us with cheap vehicles that are reliable with good fuel economy, reasonable performance, have a 10 year warranty and don't cost a fortune to fix due to your artificially inflated parts prices which can only get ever higher the more tech that goes into a vehicle?
This post has been deleted by its author
This post has been deleted by its author
> That's what you want. I want a car that can get me home from the pub, safely and legally.
Ahh, you aren't the guy I'm worried about.
I'm worried about the guy at the pub next to you that has less concern about safety and legality. I want him to have a self-driving car.
I do cringe a bit when I read stories such as this that say 'TeraFLOP' in the headline. I think about scientific computing, where 64 bit floating point is the standard. And this chip is far from a 'teraflop' by that metric.
But *in the context of this presentation*, it really is a teraflop. The presentation was mostly about what applications can use a chip like this. Those applications can (and do) effectively use FP16. How many bits per pixel does it take to recognize a speed limit sign? About a single bit, with a really big dynamic range. (Half joking here, but only half.)
Working in FP16 is a major advantage for a mobile chip. It's not so much that FP64 floating point units are larger or take more power. They do use a little more power and space, but only a little. The major issue is moving the data around, especially writing it to memory, take a lot of power and bandwidth -- far too much for a passive heatsink chip with only one or two memory devices.
> I don't want or need an autonomous car
> I don't need ANY visuals other than the dashboard and what I can see out of the windscreen
> Manufacturers - how about you just provide us with cheap vehicles
You don't need any of that, or you just can't afford it? :-)
"You don't need any of that, or you just can't afford it? :-)"
I can afford most cars on the cheap side of a ferrari right now thanks. Thats moot however since no doubt in 10 years all this crap will even be in the cheapest cars on the road. Even now touchscreens seemed to me making their pointless presence felt in almost every new car around despite the fact they're the worse possible interface for eyes-off operation. But then only a driver would appreciate that, not the kids who drool over them or the kids who design them.
The tens of thousands of deaths on the roads of the world due to human error would seem to show that you aren't really on the right side of the, err, road here.
Automation of cars will reduce road deaths. I trust a well trained computer more than a human who might be tired, inebriated, distracted etc...
Many a time on the motorway have I faced shame and embarassment on finding myself holding up a Very Important Person by using the outside lane to briefly overtake another minion like myself.
I shudder to think of the chunks of time I've taken out of a Very Important Person's hectic schedule by simply being in their way. Time they can never get back.
The cumulative effect of Very Important People not being able to get to their important meetings a few seconds earlier because of idiots like me must be utterly profound... probably a bit like the World Hunger problem.
And now... the future offers me the possibility of a self-driving car which will automatically get out of the ****ing way when I next find myself inadvertently in danger of being a hindrance to Very Important People coming up behind me in their powerful, aspirational and muscular vehicles.
The future really is brilliant!
"Many a time on the motorway have I faced shame and embarassment on finding myself holding up a Very Important Person by using the outside lane for over a mile to amble past another minion like myself at a 1mph speed differential not once looking in the rear view mirror afterwards to see the 200 metre queue of cars I'm holding up because frankly the speed I'm doing is good enough for anyone since I'm the Worlds Best Driver so sod the lot of them."
I would guess, if it's a Self Driving Car vs. a Human, then Human will lose every time, as the associated Lawyers will be very keen on protecting there product.
If it's Self Driving Car vs. Self Driving Car, then that will be interesting to see (but 3rd party Human will get the blame!)
This, rather than the technology itself is what will hold up adoption of the self driving cars. And once the legalities have all been worked out, public perception has to be managed and trust gained.
Most commentards here will understand that statistically machines that drive themselves should be far safer than meatsacks, but the average driver actually thinks they are really good at driving despite the fact that daily traffic accident stats say otherwise. Can you imagine when driverless cars go mainstream and there is the first fatality involving an autonomous car? It wouldn't even need to have been the fault of the car, the pitchforks would come out.
Foggy round our neck of the woods yesterday, quite surprising the number of cars that either have only side or no lights on at all. Note, sidelights are generally inadequate round our area as they get muddy very quickly and are pretty useless in that state.
Pretty sure a car that turns lights on automatically would be of immediate benefit given the parlous state of the driving around here.
Quote: If a self driving car *does* have an accident then who is to blame?
Clearly the other driver as the self driving car will have tons of data both visual and sensor based showing that it tried it's very best to avoid the idiot driver in the other car but when stupidity comes into play even the best of technology cannot overcome that.
> Quote: If a self driving car *does* have an accident then who is to blame?
> Clearly the other driver
Err... When I'm flying an aircraft as commander, it does not matter a iota whether I am physically acting on the controls, or someone else is, or the autopilot is doing its thing: I am responsible in any case.
Not all the time, no. If the on-board computer simply dies, severing all fly by wire controls, the pilot is pretty much useless. Its not like there are hydrolics from the pedals in a Dreamliner cockpit to the rudder or ailerons.
If the pilot does something wrong then they are held responsible. If the plane's systems do something wrong, it is either the fault of the manufacturer, or the ground crew who were maintaining it.
Now, this is all beside the point as the actuation method has nothing to do with who/what is giving the control inputs. The vast majority of autopilot installations are on non-FBW aircraft and conversely, FBW does not require the presence of an autopilot. In fact, flights can be dispatched with an inoperative or unusable autopilot on every aircraft and company I know of.
However, if the autopilot does screw up, it is still my responsibility to maintain safe control of the aircraft.
With that said:
> If the on-board computer
What on Earth is the "on board computer"???
> simply dies, severing all fly by wire controls, the pilot is pretty much useless.
It is a certification requirement (FAR/EASA) that the aeroplane be controllable in the event of loss¹ of flight control signalling. Different solutions have been adopted, e.g., on Airbus (320 series at least) rudder and trim have mechanical linkages. The 787 has redundant hydraulic / electric linkages to the flight control surfaces. If things go wrong enough, the flight deck still have direct control of the stabiliser and a pair of spoilers.
> Its not like there are hydrolics [sic] from the pedals in a Dreamliner cockpit to the rudder or ailerons.
See above (and actually, the rudder on the B787 *is* hydraulically actuated, although there is no direct linkage to the pedals).
¹ As in complete loss. A simple degradation may not even ground the aircraft.
You just yourself said that the 787 doesn't have a direct link between the peddles and the flight control surface. That means that when you press something, or use a pedal, or use any of the flight controls, those actions are being translated into the movements that the plane actually performs - via computer. It isn't a little bloke in there going "oh, he wants to go right, lets move the rudder" and then pumping hydraulic fuel into the system... Its a computer. It might be many different individual computers, with redundancies, but they're still computers.
If the wires that fly your plane break - meaning the systems that run those hydraulics are not working, that means YOU as a pilot are not legally responsible. There is absolutely nothing you can do. You can't go climbing on the wing to jump on the ailerons. Imagine it this way - you're flying along happily and your wings fall off. Is that the pilot's fault? No!
You can do your damned best to control the plane, of course, but if the plane crashes and it all ends up in court, the fault won't be "the pilot didn't do the impossible! He didn't fly the unflyable plane!". The blame will be placed, legally, on the engineers (ie. the company) or the manufacturer.
I did not mention "autopilot" at all.
> when you press something, or use a pedal, or use any of the flight controls, those actions are being translated into the movements that the plane actually performs - via computer
That's when everything is going according to Plan A. When Plan C (aka direct mode, always talking about the 777/787) comes into action the PFCs get out of the way. If your day keeps getting more and more interesting and now you're down to Plan D (aka forgot to pay the electricity bill), the stabiliser and a couple of spoilers are mechanically linked to the cockpit controls, which gives you pitch and roll authority (+ yaw via differential thrust and as secondary effect of roll).
I am afraid the rest of your post is too far out into nonsense territory to merit a reply.
Go read some NTSB reports. You'll see pretty quickly that pilots don't get the blame for everything. You seem dead set on saying that a pilot is legally responsible for all and every crash and fault that affects fly by wire planes. That is what is nonsense, and you know it.
How about this site, which lays it out quite clearly about legal liability - http://injury.findlaw.com/torts-and-personal-injuries/aviation-accidents-overview.html (that being a US site).
Or look at some crashes. China Airlines Flight 611 - crash was due to failure of maintenance (therefore, the pilot isn't responsible). American Airlines Flight 96 - faulty design (therefore, the pilot and the maintenance team aren't responsible). In neither case was the pilot legally responsible.
Those same concepts apply to *all* systems on-board a plane.
Its a question that we already have answered. If you have your car repaired by, say, a mechanic and he fails to tighten the lug nuts and a wheel later flies off - the mechanic is responsible.
If a person doesn't have their car regularly maintained and it a bit falls off on the road, it is that person's responsibility.
If a car develops a fault with its ECU and locks the throttle on, causing a crash, it is the manufacturer's responsibility.
There's a pile of case law that already exists on it. An automated car isn't so much removed from the "fly by wire" type vehicles we already have.
Audi are already running self-driving cars on test in California and Nevada - and they don't have any external indication that they're autonomous
Granted, these only do it in stop-start traffic and on highways, but Audi are already paranoid enough that they're forcing all testers to go on a driving course for familiarisation with the vehicle.
The course is 2+ days, with actual vehicle familiarisation being about 10-15 minutes. The rest of the time is making sure the human is competent, because if the car has decided it can't cope then things are going very VERY wrong.
Aside from this being a quite entertaining video for racing nuts like me, it's relevant to this discussion in that the display showing the rear view also includes additional information to aide the driver. https://www.youtube.com/watch?v=Kf0hSVL6y9A