Elon Musk: A future worth getting excited about

[kelvin 660]

An interesting interview with Elon Musk: Talks about Tesla growth, FSD, robots, and much more...
Will FSD be here within the year??

A future worth getting excited about

What's on Elon Musk's mind? In conversation with head of TED Chris Anderson, Musk details how the radical new innovations he's working on -- Tesla's intelligent humanoid robot Optimus, SpaceX's otherworldly Starship and Neuralink's brain-machine interfaces, among others -- could help maximize...

www.ted.com

 

[Wol747]

>>Will FSD be here within the year??<<

My answer would be yes. And no.
In the US I think they will fudge the numbers and make level 3 legal in places which will allow Musk to say he's "delivered" FSD even though level 3 is nowhere near what we generally understand to be FSD.
Elsewhere I don't see it happening for years, if ever.
I'm sick of seeing the same "it's xxx times safer than humans driving" line: the thousands of videos showing beta trips have almost none without interventions and disengagements but it's easy to morph them into the "it's xxx times safer than humans driving" because there have been few collisions.
There is a VAST difference between a competent autonomous car that is operated with a responsible driver and one that can be effectively left to its own devices.

 

[gangzoom]

Musk isn't stupid, nor does he need anymore $$$$$. His track record for delivering things everyone thinks is impossible is more than impressive, time lines lag but Tesla, Space X, Starlink are proof of substance.....

And it appears FSD in some parts of the US is getting there, this is before DO JO has come online, which I persume is needed for mass training in Europe? Either way I cannot wait to see FSD beta rolled out in Europe.

 

[cryo]

Something to take into considerations, what kills more people humans driving or the Tesla car self driving? Musk states when the car is far safer than a human operator you not only have self driving but it should be mandatory. I suspect Tesla cars are not far short of this point now if not already there.

 

[Irata]

Something to take into considerations, what kills more people humans driving or the Tesla car self driving? Musk states when the car is far safer than a human operator you not only have self driving but it should be mandatory. I suspect Tesla cars are not far short of this point now if not already there.

If that's the only criteria, then I can make my own FSD car tomorrow.

It also needs to get from a to b, not damage itself or property, not scare the passengers, cope with any situation without deadlock, etc.

 

[GeorgeSymonds]

Something to take into considerations, what kills more people humans driving or the Tesla car self driving? Musk states when the car is far safer than a human operator you not only have self driving but it should be mandatory. I suspect Tesla cars are not far short of this point now if not already there.

Thats been debunked many times, the current data doesn't support the argument at all. Musk doesn't want to acknowledge the truth in this regard as it's not on message.

His argument is "cars on autopilot do more miles between accidents than cars not on autopilot"
The counter argument is "autopilot is most commonly used on the safest roads and the safest times in the safest weather and drops out if it can't cope, the other measure includes all roads, all conditions and at all times including snow, ice, complex junctions, busy streets. etc". I think UK motorways, which are probably when APO is used most, are statistically 4 or 5 times safer than average, so autopilot should be 4 or 5 times safer than "all road driving" to reach parity with humans on motorways.
The acid test is, if you believe autopilot is better than humans, why would you ever try and drive a single meter without using it?

>>Will FSD be here within the year??<<

My answer would be yes. And no.
In the US I think they will fudge the numbers and make level 3 legal in places which will allow Musk to say he's "delivered" FSD even though level 3 is nowhere near what we generally understand to be FSD.

The thing that makes me think they won't is the liability question. Level 3 moves responsibility away from the driver. The litigation that will follow in the event of an accident will be enormous. In Europe we're saying the insurers will be responsbile, maybe you could argue insurers pick up the tab anyway but they'll need to be happy its safer. And people will still sue Tesla for defective software. It's massive legal hurdle I recon and I wouldn't be surprised if Musk knows that which is why he's trying to set up his own insurance company.

 

[rotor2k]

I think UK motorways, which are probably when APO is used most, are statistically 4 or 5 times safer than average, so autopilot should be 4 or 5 times safer than "all road driving" to reach parity with humans on motorways.

Although I don't disagree, intuitively it feels like your chances of dying go down as the speed limit goes down. So even if street level FSD never works, but motorway FSD works flawlessly, wouldn't you say that's a goal worth pursuing?

 

[browellm]

Although I don't disagree, intuitively it feels like your chances of dying go down as the speed limit goes down. So even if street level FSD never works, but motorway FSD works flawlessly, wouldn't you say that's a goal worth pursuing?

Is the metric in question deaths from accidents, or just accidents? Quite a distinction.

 

[GeorgeSymonds]

Although I don't disagree, intuitively it feels like your chances of dying go down as the speed limit goes down. So even if street level FSD never works, but motorway FSD works flawlessly, wouldn't you say that's a goal worth pursuing?

I've no problems with the goal, only that Musk claims its safer and punts out misleading quarterly safety stats with a narative that fuels that.

This article kind of explains it better than I can.

Tesla safety report and accident statistics explained

Tesla safety and accident statistics explained, and the need for caution

tesla-info.com

It was also interesting that no mention was made to why they dropped passive safety system metrics although it happened around the time they were making big changes to the software and removed the radar, something passive systems like emergency braking depended upon.

 

[Artiste]

I don’t think FSD will be here by the end of this year or for many years to come, especially in this country. Musk has been spouting ridiculously inaccurate timelines for years, and I prefer to judge him on his past performance rather than on his latest almost certainly over optimistic predictions. We have been told in the past that the latest hardware or the re-written software is going to deliver FSD. It hasn’t. Now it’s all about MO JO which apparently is going to solve everything. More like mumbo jumbo to me.

Yesterday I drove for miles on single track roads in rural Wales. When you meet an oncoming car one of you has to reverse, sometimes for quite some distance, to the last small passing place so you can squeeze past each other. If two FSD cars meet in these circumstances will they really be able to decide which one has to reverse and will they be able to execute such a manoeuvre? This scenario might never matter to most people but it’s a regular feature of my driving.

I can see a limited form of FSD working on motorways and duel carriageways but are we going to see, as Musk predicts, cars without steering wheels? No chance. Luckily FSD hods little interest for me as I actually thoroughly enjoy driving.

 

[rotor2k]

Yesterday I drove for miles on single track roads in rural Wales. When you meet an oncoming car one of you has to reverse, sometimes for quite some distance, to the last small passing place so you can squeeze past each other. If two FSD cars meet in these circumstances will they really be able to decide which one has to reverse and will they be able to execute such a manoeuvre? This scenario might never matter to most people but it’s a regular feature of my driving.

Just mark those roads as "too narrow/unfit for FSD". A car without a steering wheel would simply decline to drive you somewhere that could *only* be reached by one of these roads, just as today there are roads/paths that are unfit for a car but are perfectly passable for a tractor. Obviously we want 100% of destinations on earth to be reachable, but why make that a hard requirement on day 1?

 

[Wol747]

Thats been debunked many times, the current data doesn't support the argument at all. Musk doesn't want to acknowledge the truth in this regard as it's not on message.

His argument is "cars on autopilot do more miles between accidents than cars not on autopilot"
The counter argument is "autopilot is most commonly used on the safest roads and the safest times in the safest weather and drops out if it can't cope, the other measure includes all roads, all conditions and at all times including snow, ice, complex junctions, busy streets. etc". I think UK motorways, which are probably when APO is used most, are statistically 4 or 5 times safer than average, so autopilot should be 4 or 5 times safer than "all road driving" to reach parity with humans on motorways.
The acid test is, if you believe autopilot is better than humans, why would you ever try and drive a single meter without using it?


The thing that makes me think they won't is the liability question. Level 3 moves responsibility away from the driver. The litigation that will follow in the event of an accident will be enormous. In Europe we're saying the insurers will be responsbile, maybe you could argue insurers pick up the tab anyway but they'll need to be happy its safer. And people will still sue Tesla for defective software. It's massive legal hurdle I recon and I wouldn't be surprised if Musk knows that which is why he's trying to set up his own insurance company.

Agree entirely - except that level 4 doesn't remove driver liability.
However it would be difficult to argue that levels 4 or 5 would not place legal responsibility on the manufacturer: it would otherwise be like making the passengers in a train responsible for the driver causing a crash.
I would argue that FSD beta isn't proving anything except the ability of the car: the only REAL data would be if cars were driving without the driver having the ability to take over - and that's not beta at all.

 

[Wol747]

>>I can see a limited form of FSD working on motorways and duel carriageways but are we going to see, as Musk predicts, cars without steering wheels? No chance. Luckily FSD hods little interest for me as I actually thoroughly enjoy driving.<<

"duel carriageways" - perhaps nearer the truth than not merely a misspelling? (Considering the road rage that seems to be the norm...)

 

[Wol747]

Just mark those roads as "too narrow/unfit for FSD". A car without a steering wheel would simply decline to drive you somewhere that could *only* be reached by one of these roads, just as today there are roads/paths that are unfit for a car but are perfectly passable for a tractor. Obviously we want 100% of destinations on earth to be reachable, but why make that a hard requirement on day 1?

That meets level 4 but is far short of "Robotaxi".
You might also have to omit various other scenarios such as Chuck Cook's unprotected (USA) left turns and many others (at the present state of development, anyway).

 

[gangzoom]

I can see a limited form of FSD working on motorways and duel carriageways but are we going to see, as Musk predicts, cars without steering wheels? No chance.

Yesterday there was a brief sudden 40mph zone on the M4 because a car tire was littered across the carriageway.

To 'solve' even motorways the software has to as Elon now says 'sole real world AI'. What they are talking about now with surround video analysis, object memory, and even local area memory will deliver true FSD. But it doesn't sound like the AP3 CPU will have enough power to run it all!!

Incidently this weekend we did 300 miles+ of M way driving with AP, and experienced only 1 episode of phantom braking which was actually apporiate - HGV moving around out of its lane!

I did take over for the complex junctions, but boring M way driving is very very good now even without the FSD beta unfiled stack.

 

[gangzoom]

Just mark those roads as "too narrow/unfit for FSD"

Waymo does this even for unprotected left turns right now, which is OK, but makes the journey unnecessary long.

 

[Sparkeur]

As an early adopter of FSD, who has phantom-braked and lurched wildly around European highways for the past two years, I live in hope.

 

[pgkevet]

So-called corner cases happen all the time. Wobbly loads, shed tyres, accidents, police attendance, road works, fog, low sun, mud splashing on the cameras and more. These cars are at best a stepping stone to fsd. They're going to need way more powerful computers and camera washers at a minimum before any semblance of true autonomy. Elon is just raising his share price with hype and unfulfillable promises again.

 

[gangzoom]

As an early adopter of FSD, who has phantom-braked and lurched wildly around European highways for the past two years, I live in hope.

Have you used AP in Mways recently? Phantom braking is pretty much gone, and there is zero lurching in lane of any kind.

 

[Sparkeur]

Have you used AP in Mways recently? Phantom braking is pretty much gone, and there is zero lurching in lane of any kind.

Just driven 600 miles from Bordeaux to southern Spain on toll roads. Unfortunately, my experience does not match yours.

Still brakes for motorway bridges from time-to-time, two braking events where I genuinely couldn't see any reason. Still gets confused with some exits off the motorway leading to a 'lurch' - for some reason, it sometimes also signals the opposite direction even though we aren't taking the slip road. This really confuses traffic behind me. Still highly sensitive to trucks close to white lines in their lane. At night, gives red light traffic signal warnings from container trucks with high red tail lights.