Does everyone believe FSD will be viable in the next 3 years? 5 years? 10 years?

[Crissa]

One thing that never gets talked about here is the ethics problem.

No, that seems all that's talked about, and it's bunk. A driver never has that information or decision to make. So it's nonsense. A basic utility of 'reduce total deaths' will result in the car always attempting a straight line stop. There are no situations in which a driver has to decide between their own safety and a ped/cyclist unless the driver was already going too fast for the situation. And the robots won't do that.

-Crissa

Advertisement

 

[Dids]

No, that seems all that's talked about, and it's bunk. A driver never has that information or decision to make. So it's nonsense. A basic utility of 'reduce total deaths' will result in the car always attempting a straight line stop. There are no situations in which a driver has to decide between their own safety and a ped/cyclist unless the driver was already going too fast for the situation. And the robots won't do that.

-Crissa

True and then that driver loses his license.... FSD is all the same driver. How many accidents can it have till you pull it's license? And rational thinking doesn't apply... Airplanes are way safer than people think yet everyone is afraid requiring multiple pilots. No one requires 2 bus drivers.

 

[FutureBoy]

True and then that driver loses his license.... FSD is all the same driver. How many accidents can it have till you pull it's license? And rational thinking doesn't apply... Airplanes are way safer than people think yet everyone is afraid requiring multiple pilots. No one requires 2 bus drivers.

I think the 2 pilot requirement has more to do with the extreme severity of airplane accidents versus a bus accident. Busses rarely fall out of the sky (it does happen though). So busses tend to have accidents that span from fender-bender to mass casualty. Whereas airplanes tend to have only the close call or mass casualty. And mass casualty in an airplane tends to be of a larger size than on a bus.

So while the statistics indicate that airplanes are safer, the few times they do happen, the severity tends to be greater.

 

[Dids]

I think the 2 pilot requirement has more to do with the extreme severity of airplane accidents versus a bus accident. Busses rarely fall out of the sky (it does happen though). So busses tend to have accidents that span from fender-bender to mass casualty. Whereas airplanes tend to have only the close call or mass casualty. And mass casualty in an airplane tends to be of a larger size than on a bus.

So while the statistics indicate that airplanes are safer, the few times they do happen, the severity tends to be greater.

Agree with you too and you just named it. "Fall out of the sky". But it is irrational. In 2018 in the usa 43 people died on buse accidents and 1 person in an air accident. So clearly rationality has nothing to do with the amount of safety we demand from transportation and I think that irrational fear will carry over to machine control.

 

[ajdelange]

Sure, it's not perfect but I have taken 1500+mi trips with a whopping 1-2 interventions, and every update just gets better.

That's 3 nines. Not nearly enough. As Elon says its those additional required 9's that are hard to get. Now that the low hanging fruit has been picked the rate of progress is going to slow way down.

 

[MEDICALJMP]

That's 3 nines. Not nearly enough. As Elon says its those additional required 9's that are hard to get. Now that the low hanging fruit has been picked the rate of progress is going to slow way down.

I am hopeful about FSD, but I am aware of the realities. Fusion and truly autonomous driving have a lot of similarities. The 1st article I read on fusion energy was 1970. It was “only 10 years away!” It is now 2020 and every article I have read since then until now says fusion energy production is 10 years away.

 

[Dids]

I am hopeful about FSD, but I am aware of the realities. Fusion and truly autonomous driving have a lot of similarities. The 1st article I read on fusion energy was 1970. It was “only 10 years away!” It is now 2020 and every article I have read since then until now says fusion energy production is 10 years away.

There is a difference in that sustained fusion has to be perfect or it doesn't exist... it either is sustained or it isn't. FSD can be abysmal at safety but it already exists and it's really safe FSD that is the problem.
What I'm trying to say is define FSD. Is it 100% zero intervention? Is it geo fenced applications? There are already geo fenced FSD devices capable of no intervention.

 

[ajdelange]

One thing that never gets talked about here is the ethics problem.

This is actually a more interesting area than I thought it might be. Heres an article from the Stanford Business school that might be of some interest.

https://www.gsb.stanford.edu/insights/exploring-ethics-behind-self-driving-cars

Be sure to check the spider web diagram in this one

https://www.nature.com/articles/d41586-018-07135-0

 

[Dids]

I think Tesla FSD beta has the competition scared. The amount of stories in my news feed about how GM is better at cruise control and how FSD doesn't really work is way up!

 

[ajdelange]

Another interesting aspect of this is that CU just rated Tesla FSD a "distant second" to GM's current offering. Why? The main reasons were that Tesla isn't so good at verifying driver attention and that the Tesla system is so much more capable. They didn't word it that way though. They said they downgraded Tesla because it worked pretty much anywhere not solely in places where they deemed it "safe" to have autonomous. They suggested that limited access highways with a dividing island and absence of pedestrians were safe enough.

The human folly in all this is almost more interesting than the technical aspects.

 

[MEDICALJMP]

This is actually a more interesting area than I thought it might be. Heres an article from the Stanford Business school that might be of some interest.

https://www.gsb.stanford.edu/insights/exploring-ethics-behind-self-driving-cars

Be sure to check the spider web diagram in this one

https://www.nature.com/articles/d41586-018-07135-0

Interesting articles. You always add something worthwhile to whatever conversation we hit upon.

Perhaps we need to revisit Asimov’s Three Laws for Robots:

• A robot may not injure a human being or, through inaction, allow a human being to come to harm
• A robot must obey the orders given it by human beings except where such orders would conflict with the First Law
• A robot must protect its own existence as long as such protection does not conflict with the First or Second Law

A self-driving vehicle is, for all intents and purposes, a wheeled robot. In all the discussions I have read about the Trolley Dilemma nobody has ventured towards that area. They have treated it as either or. Yes, the computational requirement to differentiate a human from a mailbox would be more than what is likely needed for Full Self-Driving now. Not only do we have various shades of human beings we have various non-traditional bipedal humans. We have one legged humans, and two legged humans; we have humans who walk and humans in wheelchairs; we have humans who ride bicycles, we have humans on scooters and motorcycles. It will be a challenge.

 

[Geo]

I am hopeful about FSD, but I am aware of the realities. Fusion and truly autonomous driving have a lot of similarities. The 1st article I read on fusion energy was 1970. It was “only 10 years away!” It is now 2020 and every article I have read since then until now says fusion energy production is 10 years away.

While the race for Lv5 Autonomy sold to consumers in their own cars continues,
the actual race to make Lv5 Autonomy commercially running in public streets and open to the general public, has actually been won.

Waymo, who have been making a Robotaxi service available to passengers without any human supervisor in the car for years now, but it was only open to individuals who enrolled in Waymo's program.
Waymo, 2 weeks ago, finally declared it complete.
It is selling rides to the general public.
This is currently happening in Phoenix Arizona, (50 sq miles, expanding to 100 sq miles) but will be rolling out to all major cities.
It is a reality today

Note: Tesla is still at Lv 2 by comparison, and when it gets to Lv5, will have one huge threshold to cross, It will have to take legal responsibility for the owners of cars using Lv5 Autonomy outside its own Robotaxi service !

It is a moot point, and no longer a question when the customers cars are being operated by Tesla in the Robotaxi service, the responsibility is Tesla's without question either way you look at it.
So Tesla will happily take responsibility for the vehicles when in use in Tesla's Robotaxi service.

So no matter how soon Tesla gets to Lv 5 Autonomy, we buyers of Tesla's and FSD have a long long way to go before we can enjoy being alone in the car (or sending it out on an errand on its own) and be labelled not in charge of the car.
And all that implies. . . . . . . .

Unless for our own trips we place the car in Tesla's robotaxi service and pay a taxi fare to ride in our own car !

It will be a legal responsibility question more than a technical threshold about when the car reaches Lv 5 Autonomy, that determines when the FSD that customers have been promised and been paying for since about 2014, is truely available for Tesla owners.

Post Edit.

To be clear, Lv5 Autonomy, Full Self Driving for a Tesla car owned by the general public will only be deemed as such when the following happens.

A/ The documentation in the Tesla Handbook removes all reference to the driver being vigilant
when operating FSD.

B/ The system will not prompt the driver periodically to check if he is alert or monitoring the car.

C/ Tesla will have legal responsibility if anything goes wrong, or the car gets into an accident,
whether the car is at fault or some other car is.
Cos the occupant is sleeping, or not present, or whatever !

That is a long way off, I would suggest.

 

[MEDICALJMP]

George, what you are talking about is in a Geo-fenced area with clearly defined boundaries, markers and pre-mapped streets. That is hardly level five. Tesla is trying to come up with something much better. They want you to be able to navigate where you want and not have to have pre-designated route. What Waymo has is good only one city and only for particular areas. Try getting in their taxi and have it navigate from Phoenix to Tucson. It won’t do it. Want to buy that vehicle? You can’t and likely you could not afford to.

 

[Crissa]

What Waymo has is good only one city and only for particular areas. ...

That's not true. Waymo's system is approved for just one area. That's not the same: They know it works 99.99999% for those areas. Their system does all the object identification locally like Tesla, but constantly updates to and from a central version of the AI. It does this so it can learn and remember hazards.

The memory and refresh of Tesla's FSD is much slower.

-Crissa

 

[Geo]

George, what you are talking about is in a Geo-fenced area with clearly defined boundaries, markers and pre-mapped streets. That is hardly level five. Tesla is trying to come up with something much better. They want you to be able to navigate where you want and not have to have pre-designated route. What Waymo has is good only one city and only for particular areas. Try getting in their taxi and have itnavigate from Phoenix to Tucson. It won’t do it. Want to buy that vehicle? You can’t and likely you could not afford to.

Yes you are right for the most part.
Tesla's system is trying to be a much more general system, meaning it will work in more places.

However, both Tesla's and Waymo's and others systems, simply disengage when for whatever reason the going gets tough. Weather conditions, traffic conditions, geo fence limit.
Tesla's disengage more often, cos it operates more widely and has less technological resources it relies on, like Lidar, like Hi Definition Maps, etc.

And Yes Waymo's is limited to a geofenced area, But it disengages far less.
And that area is now expanding.

So yes you cant at the moment drive from Phoenix to Tucson, but either way, like I said the race is still on for wide spread implementation of Lv 5 Autonomy, but the first commercially available open to the public Robotaxi service has arrived just 2 weeks ago !
And it will begin rolling out to all major cities.

So if in the next 24 months say Waymo rolls out to 20 cities and Tesla still don't have any robotaxi service, it won't be any consolation to Tesla that their system may be better in some way some day ! But if Tesla gets it together, then yes they will leap frog Waymo and say they are close to world wide. But that is a very big if.

So for the moment the score is Waymo win, Tesla lose. But there is time, opportunity still remaining.