Elon: "Feature complete for full self driving this year"

[mongo]

They still advocate waiting until self-driving cars are better than humans! Does anyone seriously think that Tesla is close to having FSD drive better than a human?

Which humans? The ones that get into at-fault crashes or the ones that don't?
And are we talking about the idealized, "I wouldn't have gotten into that crash" after the fact level of driving ability; or the vigilant but checking the blind spot at the wrong moment level of ability?

What if we put FSD on cars where people have an at fault accident, DUI/ DWI, or other driving infraction first? It that more or less safe?

 

[diplomat33]

Right now based on my limited experience, before even considering bumping up to the next SAE tier, I think they need to work on:

1) Predictive course steering input. The car starts to turn in too late. I could not figure out why AP kept shutting off autosteer, and then I realized I was giving the wheel input before AP sees the curve. Just like human advice: "Look farther ahead".
2) Adding a seat shaker so bad drivers will also be warned of danger.
3) Addressing gaps in the line markers such as found lane marking transitions. "Look far ahead".

I think just these 3 would have prevented the Mountain View fatality and perhaps others. If the car was human, I'd say it has object fixation. But in reality, they aren't processing predictive motion, and they aren't doing enough to alert distracted drivers.

Those are good suggestions to start with. But Tesla has to do a lot more than that before they can even consider the next SAE level. Remember that there is a huge gap between L2 and L3 because L2 is a driver assist whereas L3 is the first level that is considered fully autonomous.

 

[Daniel in SD]

Demanding a higher level of proof for automated driving technology than we require for cars that don't have it is another form of putting in place hurdles that slow down progress. The data that Tesla has released shows that Teslas with and without Autopilot get in far fewer accidents than average cars. If you want more data, don't buy the car. If you want safer highways, force other car brands whose cars get in accidents at far higher rates to adopt and improve their automated driver assistance, so the fools driving them stop cutting me off and drifting into my lane. And force them to get cracking on stopping for red lights, as running red lights is a major cause of serious accidents.

They don't even define what they consider an accident. In order for my car to get into 6 times fewer accidents than average it would have to avoid almost all accidents that aren't even my fault. It just doesn't seem plausible. If the 6x number is really true then Tesla should be raking in crazy profits on their insurance business. My skepticism is not about systems lie AEB, emergency lane departure assist, and emergency braking for red lights and stop signs. My skepticism is about features that cause complacency in too high a percentage of users.

Which humans? The ones that get into at-fault crashes or the ones that don't?

The average (mean) human.

What if we put FSD on cars where people have an at fault accident, DUI/ DWI, or other driving infraction first? It that more or less safe?

I suppose you could replace below average humans earlier than above average humans. Practically speaking I don't think that's going to happen.

 

[EinSV]

.... In order for my car to get into 6 times fewer accidents than average it would have to avoid almost all accidents that aren't even my fault. It just doesn't seem plausible.

This post vividly demonstrates why it's pointless for Tesla to release more data. Tesla critics and skeptics discount any data that suggests automated driving technology reduces accidents or saves lives. More data would only generate more points of skepticism/criticism.

If we cared about saving lives, we should be focusing on:

(1) how best to speed up the development of this technology
(2) how to get it into cars that don't have it; and
(3) when should we require automated driving technology to be mandatory so the unsafe human drivers on the road stop causing so many injuries and deaths

 

[mongo]

They don't even define what they consider an accident. In order for my car to get into 6 times fewer accidents than average it would have to avoid almost all accidents that aren't even my fault. It just doesn't seem plausible. If the 6x number is really true then Tesla should be raking in crazy profits on their insurance business. My skepticism is not about systems lie AEB, emergency lane departure assist, and emergency braking for red lights and stop signs. My skepticism is about features that cause complacency in too high a percentage of users.

The average (mean) human.

I suppose you could replace below average humans earlier than above average humans. Practically speaking I don't think that's going to happen.

Yeah, I don't think staggered replacement would happen either, but that put is back in the realm of "you can't get better than zero accidents" on a per driver basis.
There can't be 6x safer than zero. But there can be one sixth the overall accident rate over the fleet. And then you are in total accidents per total miles which then skews things if certain drivers are responsible for multiple accidents.
For drivers with zero previous accidents, who's to say whether they will have fewer in the future with FSD vs themselves? For sure, if they have an accident on FSD, they are more likely to feel FSD is less safe then they are.

I agree complacency is a leading human factors issue with the current level of AP. Similar thing exists on our non-Tesla's TACC that won't slow the car below 30 or so. (Speed) control.. untill there isn't.

 

[DanCar]

I suppose you could replace below average humans earlier than above average humans. Practically speaking I don't think that's going to happen.

I suppose there could be some amount of self selection. Who wants to drive while they are drunk? Who wants to drive while they are tired and sleepy? Who wants to drive when they just got a message?

 

[MXLRplus]

Those are good suggestions to start with. But Tesla has to do a lot more than that before they can even consider the next SAE level. Remember that there is a huge gap between L2 and L3 because L2 is a driver assist whereas L3 is the first level that is considered fully autonomous.

I will venture a layman's guess that the current 2020 processor lacks the muscle necessary. It seems to lag in all it's responses both steering and braking. It will need to be quick enough to handle 140 mph. That would be 70 mph on a two lane road.

 

[mongo]

I suppose there could be some amount of self selection. Who wants to drive while they are drunk? Who wants to drive while they are tired and sleepy? Who wants to drive when they just got a message?

So rather than prevent system use based on driver monitoring, FSD kicks in if the driver is not paying attention.

I will venture a layman's guess that the current 2020 processor lacks the muscle necessary. It seems to lag in all it's responses both steering and braking. It will need to be quick enough to handle 140 mph. That would be 70 mph on a two lane road.

Sort of. The car being controlled is still only going 70 from a control physics point of view. Oncoming speed is a factor in time to act and required look ahead distance.
As to response, the system does not have that much lag on the perception side of things. It occurs at the frame rate of the cameras. It is likely highly filtering the action requests to smooth out responses. For gross example, it's not that it takes a second to notice that it needs to brake, it's that it waits for a second's worth for brake requests before braking.

In a general sense, I'm a terrible passenger and always think the human driver is lagging.

 

[diplomat33]

I will venture a layman's guess that the current 2020 processor lacks the muscle necessary. It seems to lag in all it's responses both steering and braking. It will need to be quick enough to handle 140 mph. That would be 70 mph on a two lane road.

I am not sure I get what you are saying. The only time you would be going 70 mph would be on a divided highway where the incoming lane is separated from your lane. So the car would not need to worry about the incoming lane. So I don't see why it would need to be able to handle a relative speed of 140 mph.

 

[mongo]

I am not sure I get what you are saying. The only time you would be going 70 mph would be on a divided highway where the incoming lane is separated from your lane. So the car would not need to worry about the incoming lane. So I don't see why it would need to be able to handle a relative speed of 140 mph.

Michigan has 65 MPH two lane roads. Montana has a 70 MPH limit in the daytime on two lane.
Speed Limits

Speed limits in the United States - Wikipedia
Other 70 MPH states:
Idaho, Nebraska, Nevada, Oregon, Wyoming

75 states: Texas, New Mexico,

Twelve states have 65 MPH limits.

 

[Daniel in SD]

This post vividly demonstrates why it's pointless for Tesla to release more data. Tesla critics and skeptics discount any data that suggests automated driving technology reduces accidents or saves lives. More data would only generate more points of skepticism/criticism.

If we cared about saving lives, we should be focusing on:

(1) how best to speed up the development of this technology
(2) how to get it into cars that don't have it; and
(3) when should we require automated driving technology to be mandatory so the unsafe human drivers on the road stop causing so many injuries and deaths

If your data doesn't stand up to criticism maybe that's the problem...
Insurance companies certainly don't seem to believe that my Tesla gets in 6 times fewer accidents than the average car.

 

[Daniel in SD]

I suppose there could be some amount of self selection. Who wants to drive while they are drunk? Who wants to drive while they are tired and sleepy? Who wants to drive when they just got a message?

Maybe it is irrational but I doubt the public would accept an autonomous vehicle that is less safe than the average human. It's very hard to see a company successfully arguing that they're only selling to bad drivers. Obviously their liability losses would be through the roof, low asset human drivers get to have absurdly low insurance liability limits, companies are putting all their assets at risk, and they would likely face punitive damages as well.

 

[MXLRplus]

I am not sure I get what you are saying. The only time you would be going 70 mph would be on a divided highway where the incoming lane is separated from your lane. So the car would not need to worry about the incoming lane. So I don't see why it would need to be able to handle a relative speed of 140 mph.

Using the visualization screen, at 45 mph the processor is 2 car lengths late.
If you were on a two lane highway with an oncoming car closing at 140, it would hit the brakes or steer out of the way after impact. Not a desirable outcome. When the center island is a curb, my car does not react quickly enough at 45. It begins to turn about 2 lengths (30') late, which is enough to avoid the curb, but shows what the reaction time of the steering system is. Ditto for someone aggressively braking at a red light, at 45 my car brakes 30' later than it should. The instant the lead car brakes, the trailing car should engage the retard system.

 

[EinSV]

If your data doesn't stand up to criticism maybe that's the problem...
Insurance companies certainly don't seem to believe that my Tesla gets in 6 times fewer accidents than the average car.

Again, you said you needed to see more data but you are raising the same type of red herring argument that could and would be raised by Tesla critics if Tesla presented more data, which again shows why its pointless for Tesla to do that.

Why is it a red herring?

Many reasons, the most obvious of which is that Autopilot is only used a relatively small fraction of the time in cars that have AP enabled. Also, insurance companies' Tesla rates are well known to vary wildly from company to company, and it is common for people who shop around to get lower rates on Teslas than much less expensive vehicles, which is consistent with reduced accident rates and greater safety.

Bottom line -- you don't believe Tesla's data. If they provided more detailed data, you (and other critics) still wouldn't believe it, and would keep coming up with arguments like this.

 

[Daniel in SD]

Again, you said you needed to see more data but you are raising the same type of red herring argument that could and would be raised by Tesla critics if Tesla presented more data, which again shows why its pointless for Tesla to do that.

Why is it a red herring?

Many reasons, the most obvious of which is that Autopilot is only used a relatively small fraction of the time in cars that have AP enabled. Also, insurance companies' Tesla rates are well known to vary wildly from company to company, and it is common for people who shop around to get lower rates on Teslas than much less expensive vehicles, which is consistent with reduced accident rates and greater safety.

Bottom line -- you don't believe Tesla's data. If they provided more detailed data, you (and other critics) still wouldn't believe it, and would keep coming up with arguments like this.

I'm just trying to find any evidence that Tesla's extraordinary claims are true. Forget about Autopilot, you really believe that Teslas get 4 times fewer accidents than the average vehicle?
It seems like you simply write off any skepticism of Tesla's data as bias. Are you claiming that liability insurance for Teslas should be a 4 times less than average? Is there a conspiracy by insurance companies? Is there something I'm not taking into account?

 

[MXLRplus]

I'm just trying to find any evidence that Tesla's extraordinary claims are true. Forget about Autopilot, you really believe that Teslas get 4 times fewer accidents than the average vehicle?
It seems like you simply write off any skepticism of Tesla's data as bias. Are you claiming that liability insurance for Teslas should be a 4 times less than average? Is there a conspiracy by insurance companies? Is there something I'm not taking into account?

My normal insurance company apparently doesn't think Teslas get in fewer accidents. Compared to our other cars similarly priced, even aluminum bodied EV and alum ICE luxury cars and glass-bodied hyper sports cars, the Tesla is 50% higher. Big thanks out to Elon for Tesla Insurance! It made our new Tesla no more expensive than other brands.

Now some are going to argue Repair Costs. Doesn't hold water when compared to a carbon fiber, aluminum, magnesium, composite hyper car that can cost $10,000 if you hit a curb with a front wheel. Carbon ceramic brakes bust off, as do air splitters, and that's >$5k for just those two parts. Minor accidents total them.

 

[Daniel in SD]

My normal insurance company apparently doesn't think Teslas get in fewer accidents. Compared to our other cars similarly priced, even aluminum bodied EV and alum ICE luxury cars and glass-bodied hyper sports cars, the Tesla is 50% higher. Big thanks out to Elon for Tesla Insurance! It made our new Tesla no more expensive than other brands.

Now some are going to argue Repair Costs. Doesn't hold water when compared to a carbon fiber, aluminum, magnesium, composite hyper car that can cost $10,000 if you hit a curb with a front wheel. Carbon ceramic brakes bust off, as do air splitters, and that's >$5k for just those two parts. Minor accidents total them.

That's why I'm talking about liability insurance rates which have nothing to do with repair costs.

 

[EinSV]

I'm just trying to find any evidence that Tesla's extraordinary claims are true. Forget about Autopilot, you really believe that Teslas get 4 times fewer accidents than the average vehicle?

I have no problem believing that. I regularly see how badly people drive and I also have had my own cars' active safety features warn me of upcoming hazards or help avoid accidents on many occasions. Four times fewer accidents is completely consistent with my experience.

As I said before, if people were really concerned about safety on the highways, they should be asking how to get automated driving technology developed as quickly as possible and into as many cars as possible as fast as possible. Human drivers make a ton of mistakes and the roads would be much safer with wider use of top notch automated driving technology like Tesla's.

It seems like you simply write off any skepticism of Tesla's data as bias?

I am saying that most Tesla "skeptics" are so hardened in their views that more data won't change their minds. It is like climate science "skeptics," most cannot be swayed by data. For example, for active safety features, Tesla has presented data that has a good internal control (Tesla cars w/o Autopilot) but most critics can't even seem to acknowledge that the reduced number of accidents compared to Teslas that don't have Autopilot hardware is strong evidence that Tesla's active safety features reduce accidents considerably. Instead, they simply dismiss it or latch onto whatever they can find to support their skepticism.

 

[diplomat33]

I just had a Tesla fanboy on Twitter tell me that Tesla will have L5 autonomy that works everywhere in the world in 2 years. LOL.

 

[Daniel in SD]

It is like climate science "skeptics," most cannot be swayed by data.

I sincerely doubt more people would be convinced if climate scientists didn't release any data. I would say Tesla has much more in common with climate change deniers who latch onto to a single data point that they claim proves what they'd like to believe.

Tesla has presented data that has a good internal control (Tesla cars w/o Autopilot)

Yes I think that is data is probably somewhat accurate. 1 "accident" per 2.1 million miles vs. 1 "accident" per 1.64 million miles. They don't appear to be correcting for demographics or anything else but it seems very likely that active safety features make cars safer. The Autopilot accident rate should definitely be corrected for when are where it's used which they don't seem to be doing.