Full Autonomy All But Confirmed for 3, "We're doing the obvious thing"

[ohmman]

Once you have full autonomy then there's not really a need for better sensors or faster processing.

There's always an opportunity for improvement. Maybe the first autonomous sensor suite can see as much as humans can. But there's a case to be made that we've got a pretty limited view. Sensors that have better light sensitivity could improve on our vision. Being able to improve on seeing through dense fog or snow would also help. Analyzing the roadway for upcoming ice, nails, etc.

Full autonomy can mean "drive like a human" but it can also mean "drive a hell of a lot better than a human." I think we're shooting for the latter, and that's a moving target.

 

[Topher]

I think we're shooting for the latter, and that's a moving target.

Even ten times fewer traffic fatalities is 3,300 dead every year.

Thank you kindly.

 

[zenmaster]

What I'm saying is that I don't see why... there comes a point where the sensors will be able to provide all the necessary information and anything is useless.

That mythical point will always be on the horizon as far as the expectations for this tech. Until then, there will be many paradigm shifts which will change the context of the discussion.

As for processing, once you achieve real-time or even just acceptable time (much faster than humans) then it'll allow fully autonomous driving. Once you have full autonomy then there's not really a need for better sensors or faster processing.

The curve to true driverless autonomy processing is going to be asymptotic as it approaches actual full 100% go-anywhere capabilities. You'll see a fairly rapid advance to that easy 90% which is what Musk is likely trying to achieve in the short term. However, the remaining capabilities will take decades of research and people will always want more.

There will still room for improvement on camera sensor sensitivity and dynamic range for the next few years though.

Not "a few years". Try decades. For example, drivers often have to rely on only their hearing to time exits from fairly dangerous blind driveways, where current radar tech would not even get a bounce. Dirt roads - new dirt roads. No visible driveway due to leaves. Single lane bridge with two way fast traffic - apparently these exist in England. Suffice to say there is a lot of processing heavy AI involved - whole environment recognition not just simple object recognition will be required.

 

[Umair Siddiqui]

Really excited for my model 3 after reading this: Elon Musk on Tesla fully autonomous car: 'What we've got will blow people's minds, it blows my mind... It'll come sooner than people think'

 

[alseTrick]

Really excited for my model 3 after reading this: Elon Musk on Tesla fully autonomous car: 'What we've got will blow people's minds, it blows my mind... It'll come sooner than people think'

Electrek is getting ahead of itself there.

Tesla is a long way from full autonomy. They will implement features like full on-ramp to off-ramp highway driving and the ability to stop at traffic lights and stop signs first (and many other various features are required, too). They are NOT going to jump straight from the current Autopilot into full autonomy. No way.

 

[JeffK]

Electrek is getting ahead of itself there.

Tesla is a long way from full autonomy. They will implement features like full on-ramp to off-ramp highway driving and the ability to stop at traffic lights and stop signs first (and many other various features are required, too). They are NOT going to jump straight from the current Autopilot into full autonomy. No way.

Judging by some of the demos on youtube from even over a year ago and judging from how Google is doing, I'd say we're extremely close to full autonomy.

In the call he practically guaranteed that the hardware will be there, so all we need to do is wait for the software updates.

Keep in mind Tesla can test out DNN models any time using data already collected. There's nothing stopping them from moving full force.

 

[3Victoria]

Agree, I think we will be surprised.

 

[JeffK]

Agree, I think we will be surprised.

... but by Elon's admission it'll be the obvious thing.

It's like knowing about your own surprise birthday party.... it's still pretty exciting! I can't wait to see how it performs.

 

[TslaIsFuture]

We will have the hardware capable for full autonomy in the Model 3 but it will probably be a few years for the software to catch up and of course government slows everything to a snails pace so who knows when they will approve.

 

[alseTrick]

Tesla is not going to jump straight from its current autopilot into full autonomy.

 

[JeffK]

Tesla is not going to jump straight from its current autopilot into full autonomy.

You'll see

 

[3Victoria]

it is possible that it will be 'full' autonomy where 'full' means we will need to be ready to take over while it learns in aggregate. They may need to be selective in who 'teaches'.

 

[zenmaster]

"full autonomy" is such a BS word - it is meaningless.

 

[zenmaster]

Judging by some of the demos on youtube from even over a year ago and judging from how Google is doing, I'd say we're extremely close to full autonomy.

In the call he practically guaranteed that the hardware will be there, so all we need to do is wait for the software updates.

Keep in mind Tesla can test out DNN models any time using data already collected. There's nothing stopping them from moving full force.

If you change the definition of "full autonomy" and change your expectation of what "close" means, then why yes, yes we are indeed. But what is that actually saying?

 

[KeRaSh]

We will have the hardware capable for full autonomy in the Model 3 but it will probably be a few years for the software to catch up and of course government slows everything to a snails pace so who knows when they will approve.

That's all I really want. I'll drive my Model 3 until the day it dies. Plenty of time for them to add the software as long as I have the hardware installed.

 

[jdevo2004]

The day a fully autonomous vehicle even bumps a pedestrian is the day all the newscasters in the world will run a full week of headlines about the danger of autonomous cars and the fact that they are fallible. People will freak out and be suspicious of autonomous cars for a decade and will endlessly try to sue the pants off of autonomous vehicle manufacturers for running over their foot or something.

 

[KeRaSh]

The day a fully autonomous vehicle even bumps a pedestrian is the day all the newscasters in the world will run a full week of headlines about the danger of autonomous cars and the fact that they are fallible. People will freak out and be suspicious of autonomous cars for a decade and will endlessly try to sue the pants off of autonomous vehicle manufacturers for running over their foot or something.

That's why I'm sure all autonomous cars will have cameras and advanced sensors on all sides to exactly log what happens in case of an accident.

 

[ecarfan]

The day a fully autonomous vehicle even bumps a pedestrian is the day all the newscasters in the world will run a full week of headlines about the danger of autonomous cars and the fact that they are fallible

Breaking News, circa 1905:

TERRIFYING "AUTO MOBILE" DEATH MACHINE KILLS INNOCENT BYSTANDER
Mechanical Malfunction Causes Horrible Fatality: Are These New Contraptions Safe?
Bystanders Are Shocked, Some Yell Out "Get a Horse!"

 

[dsvick]

"full autonomy" is such a BS word - it is meaningless.

That's two words, which one don't you like?

 

[ecarfan]

"full autonomy" is such a BS word - it is meaningless.

I find that phrase (it's two words, not one) quite clear: it means a vehicle that is capable of driving from start to destination, as determined by the human sitting in it, without any human input other then specifying the destination.

That is what it means to me. What word(s) do you use to describe a vehicle with that capability?

Google is operating cars that are capable of "full autonomy" right now, so it is not a theoretical concept.