Elon Musk $41B offer for Twitter
Discussion
soupdragon1 said:
Say you had 2 employers, and employer A had 5 safety incidents in a year, and employer B had 20. People automatically assume that the employer A is the safest, as it only had 5 safety related incidents. But what if employer A had 1 x fatality, 3 x broken limbs and 1 x minor injury, vs employer B who had 15 paper cuts, 3 x fingers trapped in the photocopier and 2 x falls on a slippy floor, all minor injuries. Suddenly employer B is safer.
^ThisContext is everything and of course we need to be extremely cautious about a provider of the technology saying that its perfect. While Tesla gets the press with FSD, lets not forget that this should also include Ford, GM, Mercedes etc. We should be extremely cautious to take their numbers at face value - how do we define what safety is and who is the arbiter of it? Dont know and we have a long way to go from here. One thing is clear though, we shouldnt let the manufacturers dictate what they consider to be safe because of course, there is no track record here at all.... [cough]Ford Pinto, VW dieselgate[cough]
And lets not forget that the vast amount of miles done on systems like FSD is on freeways / interstates, which are technically the safest. We need data and context to be broken down and details provided. Maybe something like FSD is super safe on these pin-straight roads we have in the US? I dont know. But I have to think that doing the school run in the morning is one of the riskiest drives you can actually do! Again, dont know, but I have seen some very scary driving from otherwise sane adults at this time.
I do think that Tesla is missing a trick though. The likes of Ford and GM have focused on fully mapped out roads and geo-fence their operation. Tesla is trying to crack a very hard issue and arent taking it in stages. There is far too much evidence where FSD screws up as well as drivers who screw up (due to the misunderstanding that Elon said it was perfect, so it must be). Maybe they should have taken a phased approach and built out something from there. I cant help thinking that they would get a lot broader support if they didint make these massive claims, which arent true.
EddieSteadyGo said:
soupdragon1 said:
I put a quick excel sheet together to help explain my point.
...
I think there is an error in your spreadsheet....
You have the "deaths avoidable" in "regular cars" being at a lower rate than the robotaxi car rate. Hence why it gives more deaths with robocars when it should be less.
I'm showing that if there were 0.1% deaths, FSD being twice as good as the average would equal a death rate of 0.05%. Sounds good on paper.
When you remove unavoidable deaths (drunk truck driver, hitting spilled slurry on a bend, deer jumps on your bonnet, car has mechanical failure etc) then you see that the human driver is still safer at 0.04 vs 0.05 FSD.
It's necessary to remove those 'unavoidable' stats as it's wrong to assume that driving a car with FSD means you won't find slurry on a bend etc
It's all about finding a method to make the comparison "like for like'
Eg, how many lives are saved with basic AEB?
Tesla previously tried to show how safe they were by comparing their car to the average car (which includes 30 year old bangers for example) but a like for like would be to compare itself to other modern cars of similar age and spec. But of course, they wouldn't do that, as the stats aren't as convincing.
Like for like analysis is more meaningful than cherry picked stats to provide a misleading narrative. That's Teslas MO as we are well used to by now.
Pick any modern car made in the last 3 years. Any car on the planet. There is a strong likelihood that it will be safer than the average car on our roads. Tesla continually compares itself to 'average' but average is a low bar in what is an ever improving arena.
It's all about getting underneath the headline data and doing proper scrutiny/due diligence.
soupdragon1 said:
...I'm showing that if there were 0.1% deaths, FSD being twice as good as the average would equal a death rate of 0.05%. Sounds good on paper.
When you remove unavoidable deaths (drunk truck driver, hitting spilled slurry on a bend, deer jumps on your bonnet, car has mechanical failure etc) then you see that the human driver is still safer at 0.04 vs 0.05 FSD.
...
Honestly your logic doesn't follow. When you remove unavoidable deaths (drunk truck driver, hitting spilled slurry on a bend, deer jumps on your bonnet, car has mechanical failure etc) then you see that the human driver is still safer at 0.04 vs 0.05 FSD.
...
I understand your initial point that there is a certain rate of death from driving. And you could split that rate between deaths which are avoidable and those which are out of the control of the driver and hence unavoidable.
But to calculate the robotaxi risk, you are taking the overall rate of death (including unavoidable), dividing it by two, and adding back in the unavoidable rate of death. That's not the right method.
EddieSteadyGo said:
Honestly your logic doesn't follow.
It does.If the claim from Musk is that FSD will be twice as better than the average - which includes unavoidable deaths - then it is less safe than most drivers.
The goal should be to get FSD twice as safe as a driver *after* unavoidable deaths have been removed from the statistics.
gregs656 said:
It does.
If the claim from Musk is that FSD will be twice as better than the average - which includes unavoidable deaths - then it is less safe than most drivers.
The goal should be to get FSD twice as safe as a driver *after* unavoidable deaths have been removed from the statistics.
Nope. If you are going to include "unavoidable deaths" in the average, and we are using the assumption that FSD reduces the average deaths by half, and the unavoidable deaths are truly unavoidable, then by default FSD death rate must be better than the human driver death rate.If the claim from Musk is that FSD will be twice as better than the average - which includes unavoidable deaths - then it is less safe than most drivers.
The goal should be to get FSD twice as safe as a driver *after* unavoidable deaths have been removed from the statistics.
However, I not arguing with the underlying point that comparing avoidable deaths is what matters most. And also measuring this is going to be very difficult.
Edited by EddieSteadyGo on Friday 26th April 18:40
gregs656 said:
EddieSteadyGo said:
Honestly your logic doesn't follow.
It does.If the claim from Musk is that FSD will be twice as better than the average - which includes unavoidable deaths - then it is less safe than most drivers.
The goal should be to get FSD twice as safe as a driver *after* unavoidable deaths have been removed from the statistics.
I'll repeat from my earlier post, it can be easy to misinterpret this and its hard for me to articulate the point to people who aren't big into data analysis.
I'll try one more time.
Point 1. FSD can't influence the other cars or drivers on the road. Regardless of the quality of FSD, these accidents will take place anyway,
Point 2. The overall average amount of accidents includes all of those accidents from point 1, despite FSD having no opportunity to reduce them.
I hope this base position is clear. The average deaths in my example is 0.1% deaths per driver. If FSD is twice as good as this statistic, it achieves 0.05% deaths per driver. Hopefully that is clear too.
Finally, when we look at where the initial 0.1% average comes from, 0.06% is due to unavoidable accidents, and 0.04% from driver judgement (avoidable)
FSD has an avoidable death rate of 0.05% and driver judgement has an avoidable death rate of 0.04%. So FSD is worse in its sphere of influence.
Hopefully that step through helps articulate the point. Looking at it another way, the reason why some people may think being safer than the overall average is a good thing is simply because it sounds logical. But its only logical if we assume that all deaths are from driver error. 1 lorry driver falling asleep at the wheel could kill 10 people. It wasn't the fault of those 10 people, so its wrong to assume that FSD being twice as safe could have saved 5 of them.
All I tried to do was show an example of breaking out data to make it more comparable to real world situations.
soupdragon1 said:
Yes, you are getting exactly what I'm saying.
I'll repeat from my earlier post, it can be easy to misinterpret this and its hard for me to articulate the point to people who aren't big into data analysis.
I'll try one more time.
Point 1. FSD can't influence the other cars or drivers on the road. Regardless of the quality of FSD, these accidents will take place anyway,
Point 2. The overall average amount of accidents includes all of those accidents from point 1, despite FSD having no opportunity to reduce them.
I hope this base position is clear. The average deaths in my example is 0.1% deaths per driver. If FSD is twice as good as this statistic, it achieves 0.05% deaths per driver. Hopefully that is clear too.
Finally, when we look at where the initial 0.1% average comes from, 0.06% is due to unavoidable accidents, and 0.04% from driver judgement (avoidable)
FSD has an avoidable death rate of 0.05% and driver judgement has an avoidable death rate of 0.04%. So FSD is worse in its sphere of influence.
Hopefully that step through helps articulate the point. Looking at it another way, the reason why some people may think being safer than the overall average is a good thing is simply because it sounds logical. But its only logical if we assume that all deaths are from driver error. 1 lorry driver falling asleep at the wheel could kill 10 people. It wasn't the fault of those 10 people, so its wrong to assume that FSD being twice as safe could have saved 5 of them.
All I tried to do was show an example of breaking out data to make it more comparable to real world situations.
I'm used to analysing lots of data so don't worry about that I'll repeat from my earlier post, it can be easy to misinterpret this and its hard for me to articulate the point to people who aren't big into data analysis.
I'll try one more time.
Point 1. FSD can't influence the other cars or drivers on the road. Regardless of the quality of FSD, these accidents will take place anyway,
Point 2. The overall average amount of accidents includes all of those accidents from point 1, despite FSD having no opportunity to reduce them.
I hope this base position is clear. The average deaths in my example is 0.1% deaths per driver. If FSD is twice as good as this statistic, it achieves 0.05% deaths per driver. Hopefully that is clear too.
Finally, when we look at where the initial 0.1% average comes from, 0.06% is due to unavoidable accidents, and 0.04% from driver judgement (avoidable)
FSD has an avoidable death rate of 0.05% and driver judgement has an avoidable death rate of 0.04%. So FSD is worse in its sphere of influence.
Hopefully that step through helps articulate the point. Looking at it another way, the reason why some people may think being safer than the overall average is a good thing is simply because it sounds logical. But its only logical if we assume that all deaths are from driver error. 1 lorry driver falling asleep at the wheel could kill 10 people. It wasn't the fault of those 10 people, so its wrong to assume that FSD being twice as safe could have saved 5 of them.
All I tried to do was show an example of breaking out data to make it more comparable to real world situations.
Point 1 and 2 are fine.
The issue is your assumption when you start with a total death rate of 0.1% and it halves with FSD to 0.05%. That halving *includes* unavoidable deaths because it was the total death rate (as per gregs656 who also restated your assumption). If the total death rate halved, it will still contain the 60% component which were unavoidable. So the FSD component would actually be massively better than the human driver.
In your spreadsheet, you take the total death rate, halve it, and then add back in unavoidable deaths. That isn't valid.
You either start by measuring a reduction in total death rates (in which case you don't need to add back in unavoidable deaths) or you start by measuring the avoidable driver component. But you can't start with one and switch to the other.
Anyway, sounds like we need to agree to disagree, as I suspect it's a bit boring for everyone to read our back and forth
Edited by EddieSteadyGo on Friday 26th April 21:32
dobbo_ said:
So which will be first?
FSD
Robotaxi
Roadster
All of which have been due for years
FSD in good weather comes first imo. FSD
Robotaxi
Roadster
All of which have been due for years
Then some kind of limited Robotaxi, probably in somewhere like Florida, where he can get some political cover from DeSantis, to help protect against the inevitably legal quagmire.
We already have had the vanity project (Cybertruck), and I don't think Musk can't afford another one in the short term, so the Roadster would likely come after that.
EddieSteadyGo said:
The issue is your assumption when you start with a total death rate of 0.1% and it halves with FSD to 0.05%. That halving *includes* unavoidable deaths because it was the total death rate (as per gregs656 who also restated your assumption). If the total death rate halved, it will still contain the 60% component which were unavoidable. So the FSD component would actually be massively better than the human driver.
I think this is where the misunderstanding is - I don't know exactly what Musk claims but I took it that he means something like 'When FSD it to blame, it will be twice as good as average'The 'average' includes unavoidable deaths.
So 'twice as good as average, when at fault' is not as good as avoidable driver deaths.
I think Musk is actually counting and discounting avoidable and unavoidable to make it seem like FSD will be 'twice as good in avoidable situations'.
gregs656 said:
I think this is where the misunderstanding is - I don't know exactly what Musk claims but I took it that he means something like 'When FSD it to blame, it will be twice as good as average'
The 'average' includes unavoidable deaths.
So 'twice as good as average, when at fault' is not as good as avoidable driver deaths.
I think Musk is actually counting and discounting avoidable and unavoidable to make it seem like FSD will be 'twice as good in avoidable situations'.
Honestly, soupdragon's original analysis isn't right.The 'average' includes unavoidable deaths.
So 'twice as good as average, when at fault' is not as good as avoidable driver deaths.
I think Musk is actually counting and discounting avoidable and unavoidable to make it seem like FSD will be 'twice as good in avoidable situations'.
You can't start with the total death rate, halve it, and then add back in unavoidable deaths.
You could start with the avoidable death rate e.g. 0.4%, halve that, e.g. 0.2%, and then add back in unavoidable deaths at 0.6%. But that would give a comparable death rate of 0.8% which is 20% better.
EddieSteadyGo said:
dobbo_ said:
So which will be first?
FSD
Robotaxi
Roadster
All of which have been due for years
FSD in good weather comes first imo. FSD
Robotaxi
Roadster
All of which have been due for years
Then some kind of limited Robotaxi, probably in somewhere like Florida, where he can get some political cover from DeSantis, to help protect against the inevitably legal quagmire.
We already have had the vanity project (Cybertruck), and I don't think Musk can't afford another one in the short term, so the Roadster would likely come after that.
We’ve updated our future vehicle lineup to accelerate the launch of new models.” Musk said we might see the vehicles in early 2025, if not later this year.
Anyone want to place a bet when this might happen? Not forgetting that there does seem to be evidence that a lot of engineers in both California and Austin have been fired. Does this affect the 'cheap' Tesla? I guess we will find out.
As for self-driving cars - there is a reason why almost all of the us companies are running services in Phoenix; its a super easy road layout (for the majority) and with big wide roads. Florida does have some simpler layouts, but they also have a number of issues - swamps, ocean and Florida-man....
EddieSteadyGo said:
Honestly, soupdragon's original analysis isn't right.
You can't start with the total death rate, halve it, and then add back in unavoidable deaths.
You could start with the avoidable death rate e.g. 0.4%, halve that, e.g. 0.2%, and then add back in unavoidable deaths at 0.6%. But that would give a comparable death rate of 0.8% which is 20% better.
I don’t have a dog in the fight - but if Musk has claimed FSD will be twice as better than ‘the average’ (which includes all deaths) then the analysis is correct. You can't start with the total death rate, halve it, and then add back in unavoidable deaths.
You could start with the avoidable death rate e.g. 0.4%, halve that, e.g. 0.2%, and then add back in unavoidable deaths at 0.6%. But that would give a comparable death rate of 0.8% which is 20% better.
I think you are in agreement with him.
Byker28i said:
So despite early protestations that a mechanic would be sent out - all cybertrucks have to go back to the dealers to have the dodgy pedals fixed...
That is so cool that they launched such a revolutionary technically advanced product and the only issue is a rivet on the pedal pad! Tesla is amazing.EddieSteadyGo said:
gregs656 said:
I think this is where the misunderstanding is - I don't know exactly what Musk claims but I took it that he means something like 'When FSD it to blame, it will be twice as good as average'
The 'average' includes unavoidable deaths.
So 'twice as good as average, when at fault' is not as good as avoidable driver deaths.
I think Musk is actually counting and discounting avoidable and unavoidable to make it seem like FSD will be 'twice as good in avoidable situations'.
Honestly, soupdragon's original analysis isn't right.The 'average' includes unavoidable deaths.
So 'twice as good as average, when at fault' is not as good as avoidable driver deaths.
I think Musk is actually counting and discounting avoidable and unavoidable to make it seem like FSD will be 'twice as good in avoidable situations'.
You can't start with the total death rate, halve it, and then add back in unavoidable deaths.
You could start with the avoidable death rate e.g. 0.4%, halve that, e.g. 0.2%, and then add back in unavoidable deaths at 0.6%. But that would give a comparable death rate of 0.8% which is 20% better.
As we've been discussing, FSD doesn't control the entire sphere, aka, it can't control the other nut jobs on the road. That's why I've described the bar that FSD needs to reach isn't to be twice as good as the overall average, it needs to be better than a good driver. Good driver and average driver are obviously different standards.
In my made up example, the average is 1.0 but a good driver is 0.4.
FSD twice as good as average is 0.5 but twice as good as a good driver is 0.2. very different numbers.
soupdragon1 said:
Yes, I agree with that. That's logical. What you are describing is FSD being twice as good as a regular driver in its sphere of influence.
As we've been discussing, FSD doesn't control the entire sphere, aka, it can't control the other nut jobs on the road. That's why I've described the bar that FSD needs to reach isn't to be twice as good as the overall average, it needs to be better than a good driver. Good driver and average driver are obviously different standards.
In my made up example, the average is 1.0 but a good driver is 0.4.
FSD twice as good as average is 0.5 but twice as good as a good driver is 0.2. very different numbers.
Yes, but just in terms of your original conclusion, you started with the total death rate of 0.1% (which included unavoidable deaths) and you halved that total figure. And then you added unavoidable deaths back in again. So your spreadsheet for the FSD scenario included two lots of unavoidable deaths. That's why it showed total FSD deaths increasing. Anyway, I think you get my point as to why the original spreadsheet wasn't right, and if not, we will have to agree to disagree As we've been discussing, FSD doesn't control the entire sphere, aka, it can't control the other nut jobs on the road. That's why I've described the bar that FSD needs to reach isn't to be twice as good as the overall average, it needs to be better than a good driver. Good driver and average driver are obviously different standards.
In my made up example, the average is 1.0 but a good driver is 0.4.
FSD twice as good as average is 0.5 but twice as good as a good driver is 0.2. very different numbers.
[quote=off_again
Dont forget that Musk promised the 'cheap' Tesla.
We’ve updated our future vehicle lineup to accelerate the launch of new models.” Musk said we might see the vehicles in early 2025, if not later this year.
Anyone want to place a bet when this might happen?
[/quote]
Well, given every single other thing Musk has ever promised a timescale against (FSD, Tesla Semi, Cybertruck, Space-X missions to Mars) has been at least 6x longer than promised, if ever, my guess is 2040?
Dont forget that Musk promised the 'cheap' Tesla.
We’ve updated our future vehicle lineup to accelerate the launch of new models.” Musk said we might see the vehicles in early 2025, if not later this year.
Anyone want to place a bet when this might happen?
[/quote]
Well, given every single other thing Musk has ever promised a timescale against (FSD, Tesla Semi, Cybertruck, Space-X missions to Mars) has been at least 6x longer than promised, if ever, my guess is 2040?
dimots said:
Byker28i said:
So despite early protestations that a mechanic would be sent out - all cybertrucks have to go back to the dealers to have the dodgy pedals fixed...
That is so cool that they launched such a revolutionary technically advanced product and the only issue is a rivet on the pedal pad! Tesla is amazing.Gassing Station | News, Politics & Economics | Top of Page | What's New | My Stuff