OFFICIAL BUTTON WATCH

[Knightshade]

Did anyone without a channel get an invite?

Many, yes.

Did Chuck and aidrver take a driving test to show they were "expert drivers" before they were accepted? Were they already known as expert drivers for some reason?
What was the application process to be a beta tester?

You realize Tesla has all the data needed to make judgements about how safely you drive the car right? (assuming you didn't opt out of data collection)

It's one of the reasons Tesla insurance it going to be a pretty important and profitable product in the future.

The reason this came up is I used it as an example of Elon being deceptive. He said it was going to "expert drivers" as the only criteria he listed. The data does not support this, so it does support Elon being less than honest with us. I agree they don't have to tell us how they chose, but Elon decided to do it. It's just evidence that Elon is not 100% forthcoming in his tweets. Yet, such a simple fact has to be defended heavily by the downvote cohort.

I race cars, am a pilot, and I design safety critical systems for a living. I didn't get an invite. But I don't have youtube channel... (and no, I don't care I didn't get one, and I would not be a good tester)

FWIW one of the 2 folks I personally know who got it was an actual race car driver, and has no you tube channel.
(the other runs a regional Tesla club, and while he technically has a youtube account at the time he got the beta I think he had like 3 total videos posted in years, and no significant # of followers).

Maybe Elon just doesn't like you?

 

[Terminator857]

It's not that simple. We're discussing what he means by "Releasing Beta V9" not when it will occur. Elon has many, many misleading, deceptive, and crazy communications unrelated to time (going private, covid denials, pedo tweets, red pill tweets) that require a magic decoder ring that only exists in the future to understand what he "meant".

I'm still miffed my red interior hasn't arrived.

https://twitter.com/x/status/1002983121896103937

 

[powertoold]

Less than 24 hours to go, I can't believe it's actually happening! It's been about 5 months since the last update, and they've fixed most of the known issues. I have high expectations for V9.

 

[mikes_fsd]

I have high expectations for V9.

Me too, but just in case it is worth repeating -- for the FSD Beta participants, be alert "there will be unknown issues, so please be paranoid."

https://twitter.com/x/status/1413306409693892613

I am glad he call it out again (I know he did a similar thing last October with the initial FSD Beta rollout).

 

[Ptheven]

"Safety is always top priority at Tesla."

demonstrably false. They'd never use red rear turn signals if that were the case

 

[gearchruncher]

"There are unknown issues, so please be paranoid"
"Safety is always top priority at Tesla"

Dude's reality distortion field is strong. These things don't go together. If safety was literally the top priority, this is not the way they would test new software releases. Testing software on public streets with uniformed participants is not how you maximize safety, and even raises ethical questions. Telling testers to be paranoid in a tweet is not how you maximize safety. In safety critical systems, where safety is actually a top priority, there is no such thing as a public beta, nor do you remove radar and release software with reduced capabilities on the world (even temporarily).

I'm not saying I am against this process as a way to release L2 systems, but claiming safety is Teslas TOP priority is demonstrably false and is marketing puffery. Like any business they have many competing priorities, and in this case the economic and R&D needs clearly are in front of maximizing safety.

Plus, anybody actually trained in safety knows there is no such thing as "safety." All systems have risk. If 100% safety is your top priority, you never make a car at all. There's only acceptable safety balanced against the functionality and value of your product, and anyone that claims safety is the top priority of their system clearly has failed to actually document their priorities. Cars are dangerous. But useful. That's always a hard tradeoff. But that message doesn't sell and isn't sexy, so Elon gets to say "top priority!" and lap up that fanboy praise as if tweeting something makes it true.

and they've fixed most of the known issues

Yet another genius marketing message from Musk.
It's called FSD constantly. But it's actually L2 city streets autosteer.
If it was FSD, it would be a "known issue" that it can't handle an intersection if the power is out, or the lights are working but a cop is directing traffic.
If it was FSD, it would be an issue that it breaks traffic laws
If it was FSD, it would be an issue that it requires a human to put torque on the steering wheel constantly.

However, as an L2 city streets beta, these are not "issues" - these are just not part of the design, and can be dismissed. They're just surprisingly not part of the design if you think what you are looking at is a full self driving beta. So it's the ultimate weasel word description- anything bad it does you can just explain away as not part of city streets beta, instead of something it's failing at, even though you're the one calling it FSD constantly.

I look forward to Tesla's progress as much as anyone, but let's not fall for his hype that this is somehow anything close to FSD with all known issues fixed.

 

[ahoen117]

Dude's reality distortion field is strong. These things don't go together. If safety was literally the top priority, this is not the way they would test new software releases. Testing software on public streets is not how you maximize safety. Telling testers to be paranoid in a tweet is not how you maximize safety. In safety critical systems, where safety is actually a top priority, there is no such thing as a public beta.

As far as we know, there hasn't been a single accident that involves FSD beta (Please correct me if I'm wrong). You can't make a claim that safety isn't a top priority without data to show what they're doing isn't safe...

 

[gearchruncher]

As far as we know, there hasn't been a single accident that involves FSD beta (Please correct me if I'm wrong). You can't make a claim that safety isn't a top priority without data to show what they're doing isn't safe...

Everything is safe until it isn't. Are you saying the 737-Max was safe for the first 2 years before it had the first crash? That a new drug is safe until the first person experiences massive side effects?

71 people under NDA have a city streets beta. This is a wholly insufficient data set to judge if this product is as safe or safer than humans, which are our current baseline level of acceptable safety. So lack of a crash isn't "proof" in such a small data set.

But we do know that without fast, constant human monitoring, FSD beta will kill you. Watch this video and tell me that FSD is safe without a very responsive driver behind the wheel. If your system is only safe because humans are expected to override it, then it's not a safety based system at all, and safety shouldn't be in your description. This is like saying "safety is our top priority" when designing the music system in the car.

But it doesn't matter. Elon didn't say "a" top priority. He said "The top priority at Tesla." It is trivial to prove this false. You do not release beta software to be used on public streets if safety is your singular top priority, and you do not tell your testers to be "paranoid" in a tweet. You just don't release it. Also, people have absolutely died using non-FSD autopilot in Teslas. He didn't say safety for FSD is the top priority. He said it was top for all of Tesla. Well, if you want to use any crash as a proof that Tesla doesn't have safety as the top priority, there you go, people have died in Teslas on AP. And not on AP. (did you know nobody has ever died in a Volvo XC90 in the UK in the last 20 years?)

Safety is a priority at Tesla. So is making money, marketing their products, selling useful cars to people, developing future technologies (which might increase safety!), and making regulators happy. None of those are "the" top priority. They're all hard tradeoffs. If there's only one to priority, an excel spreadsheet can run your business.

 

[gearchruncher]

Seriously? A pharmaceutical lawyer/professor upvoting the idea that if no injury has yet occurred with a product, it must be safe? Pharmaceuticals are the poster child for unexpected consequences, hard trade offs, ethical missteps, and pulled products.

Why don't we allow public "betas" of drugs if this is the logical way that a company with safety as the top priority behaves?

For all of those upvoting the idea that because there have been no crashes on FSD, it must be safe, how do you explain the crashes on normal AP? It seems you'd defend FSD as safe even if crashes had occurred, just like you defend crashes on AP as the driver's fault, not the system. As an L2 system, you constantly point out that the safety comes from the human overseer, not the actual system, which has a disclaimer that the human is responsible. So it's irrelevant. The way Tesla makes city streets beta safe is by picking expert drivers, not by making the system itself inherently safe, even Elon said that.

 

[Ptheven]

Actually it occurs to me that saying safety is their top priority could be true, it just means that their competence level is frighteningly low.

 

[McFlurri]

"Safety is always top priority at Tesla."

demonstrably false. They'd never use red rear turn signals if that were the case

Never owned a car that had anything BUT rear red turn signals... so either I'm lucky to be alive, or there is not a safety risk...

 

[ahoen117]

Everything is safe until it isn't. Are you saying the 737-Max was safe for the first 2 years before it had the first crash?

71 people under NDA have a city streets beta. This is a wholly insufficient data set to judge if this product is as safe or safer than humans, which are our current baseline level of acceptable safety. So lack of a crash isn't "proof" in such a small data set.

But we do know that without fast, constant human monitoring, FSD beta will kill you. Watch this video and tell me that FSD is safe without a very responsive driver behind the wheel. If your system is only safe because humans are expected to override it, then it's not a safety based system at all, and safety shouldn't be in your description. This is like saying "safety is our top priority" when designing the music system in the car.

But it doesn't matter. Elon didn't say "a" top priority. He said "The top priority at Tesla." It is trivial to prove this false. You do not release beta software to be used on public streets if safety is your singular top priority, and you do not tell your testers to be "paranoid" in a tweet. You just don't release it.

Safety is a priority at Tesla. So is making money, marketing their products, selling useful cars to people, developing future technologies (which might increase safety!), and making regulators happy. None of those are "the" top priority. They're all hard tradeoffs. If there's only one to priority, an excel spreadsheet can run your business.

I can't speak to the 737 max because I honestly don't know enough about the situation other than what I heard on the news and a quick skim through the Wikipedia article.

I don't know how popular this opinion is, but a zero tolerance policy for safety hurts innovation. When you're using a beta system that has the potential to kill you, you have to pay attention. If you aren't paying attention and something happens, that is 100% be on you, not Tesla and the Autopilot team.

I think its fairly well understood by the (currently) small data set of people testing that FSD beta is experimental software. If "the button" ever does show up, no one is forcing any one to push it. Those who want the beta and are willing to assume the risk will use the beta. Those who aren't willing, shouldn't accept the terms.

I'm not saying you should ship a product as a public beta that is knowingly bad, but how are you supposed to find edge cases where the software doesn't work properly if you don't put it through real world testing to find the issues?

 

[Ptheven]

Never owned a car that had anything BUT rear red turn signals... so either I'm lucky to be alive, or there is not a safety risk...

It's well-established that amber rear turn signals are safer. It's not like saying red rear turn signals have a 100% chance that your family will be hunted down and turned into lamp shades, just that there's about a 5% higher probability that you'll be in an accident with a vehicle approaching from your rear.

 

[gearchruncher]

I don't know how popular this opinion is, but a zero tolerance policy for safety hurts innovation.

Yep, I said that already. It's literally how any actual safety professional thinks, because the only 100% safe system is the one that doesn't exist:

I'm not saying I am against this process as a way to release L2 systems, but claiming safety is Teslas TOP priority is demonstrably false and is marketing puffery. Like any business they have many competing priorities, and in this case the economic and R&D needs clearly are in front of maximizing safety.

Plus, anybody actually trained in safety knows there is no such thing as "safety." All systems have risk. If 100% safety is your top priority, you never make a car at all. There's only acceptable safety balanced against the functionality and value of your product, and anyone that claims safety is the top priority of their system clearly has failed to actually document their priorities. Cars are dangerous. But useful. That's always a hard tradeoff.

When you're using a beta system that has the potential to kill you, you have to pay attention. If you aren't paying attention and something happens, that is 100% be on you, not Tesla and the Autopilot team.

It seems like you are fully in agreement that Safety is NOT Tesla's top priority. They have handed that off to the users of the beta. They refuse to take responsibility for your safety. That means it's not their priority- they aren't even really considering it. Yet Elon says it's their top priority, and people say "I'm so glad Tesla is being super safe!" Then, when that's pushed on, the answer is "it's up to the driver, it's not Tesla's fault!"

I'm not saying you should ship a product as a public beta that is knowingly bad, but how are you supposed to find edge cases where the software doesn't work properly if you don't put it through real world testing to find the issues?

Via internal testing with paid professionals? Like lots of other safety critical companies have done for decades? How did we ever manage to ship an airplane or a CT scanner? How did Tesla manage to ship airbags or ABS systems which are actually safety critical?

Look, it's clear that everyone here knows safety is not Tesla's top priority. It can't be. It doesn't make them wrong- it's the only way to move forward, and it's the way all companies work. But it makes Elon saying "it's our top priority" look silly, particularly when he just said "be paranoid", and it makes people hyping this statement look at bit based in just supporting Elon in every.single.thing.he.says. It's all marketing puffery, and yet another demonstration that not everything Elon says means he's particularly knowledgeable in that area.

 

[Knightshade]

Via internal testing with paid professionals? Like lots of other safety critical companies have done for decades?

Do you legitimately think they don't already do that before public betas?

Like they just compile some code and throw it over the public fence entirely untested?

How did we ever manage to ship an airplane or a CT scanner? How did Tesla manage to ship airbags or ABS systems which are actually safety critical?

The problem is a CT scanner doesn't have 500,000 possible corner cases that would cause a failure that we don't know about in advance.

Self driving via NN perception does.

So literally the only way TO test it properly is keep having increasingly large #s of people, in increasingly varied locations and situations, testing it in the real world.

If you only did a bunch of internal, tiny, narrow, testing, you'd end up with a system that works GREAT for that small group of people- but then still has all those myriad unknown weird corner case risks for everyone else, but with a false confidence it's "safe"

So Teslas methods here- with very heavy language clearly telling you what's up... and even removing folks who via the internal camera appear to using it unsafely, is the most effective method of testing available. for this application.

 

[Terminator857]

Tesla FSD Beta will improve safety. A year or so down the road Tesla will release statistics that say that. Elon telling people to be paranoid improves safety. Your not safe when you are playing on your phone, oblivious to your surroundings, or confident nothing is going to happen while racing down the streets.

The closer we are to true FSD the safer we are. Tesla's plan of releasing FSD beta to the masses gets us closer to that goal and increased safety. Only Elon is willing to take that risk and is why Tesla will win the FSD race many years down the road.

As far as we know, there hasn't been a single accident that involves FSD beta ...

Several videos have shown running over curbs (bent rims?). If you say bad things about Tesla your FSD beta may get removed or may not get invited to the next big thing, so there is a financial incentive to under report, so as to not lose youtube revenue.

 

[gearchruncher]

but then still has all those myriad unknown weird corner case risks for everyone else, but with a false confidence it's "safe"

but how are you supposed to find edge cases where the software doesn't work properly if you don't put it through real world testing to find the issues?

Ahh, the magic of the "edge cases" excuse! When Tesla, the company who has safety as the TOP priority, preciously released software that would kill you making a left turn across a few lanes of traffic. Who could have ever seen someone trying to do that!

Tesla is so far away from handling normal cases safely that the edge cases are just a straw man to justify their minimal internal testing. But of course, that minimal internal testing is very reasonable when the definition of your product is not around safety, and requires constant, vigilant monitoring of the system by an operator. It's just that a hyper monitored system isn't described as "full self driving" by any normal human, and these processes don't scale to L4 development.

Like I said, I agree that this is likely a reasonable process for Tesla and is good for their overall business. It's just not what you would do if safety was your top priority like Elon claims.

 

[gearchruncher]

Tesla FSD Beta will improve safety. A year or so down the road Tesla will release statistics that say that. Elon telling people to be paranoid improves safety. Your not safe when you are playing on your phone, oblivious to your surroundings, or confident nothing is going to happen while racing down the streets.

Tesla's already said AP increases safety. The issue is that the statistics they released fail to prove that in any way and compare apples to oranges. So that's not exactly a confidence inspiring path, when Tesla has to obscure numbers in a much easier operational domain in order to make it look safer. But it's also expected, because AP is not about safety, it's about convenience, like many other driver assist functions (cruise control, auto park, etc)

If you believe distraction is the primary cause of accidents, then Tesla should be mostly focused on the driver monitoring functions. A company with safety as the top priority would be figuring out how to minimize accidents in the quickest, most efficient way, which would be through monitoring. Not through an L2 system which requires even more vigilance than self driving, and which turns off when you get distracted, preventing it from helping in distracted driving cases....

How the heck does a manually engaged, constant monitoring required, "be paranoid" system improve safety other than the fact that it hopefully causes drivers to pay more attention because it's so dangerous? You've just recreated the Tullock Spike- the idea that the safest car in the world has a spike coming out of the steering column so you'll be hyper vigilant.

 

[ahoen117]

Look, it's clear that everyone here knows safety is not Tesla's top priority. It can't be. It doesn't make them wrong- it's the only way to move forward, and it's the way all companies work. But it makes Elon saying "it's our top priority" look silly, particularly when he just said "be paranoid", and it makes people hyping this statement look at bit based in just supporting Elon in every.single.thing.he.says. It's all marketing puffery, and yet another demonstration that not everything Elon says means he's particularly knowledgeable in that area.

I don't work for Tesla so I couldn't tell you what their internal priorities are. Looking at it from the outside you could make arguments for or against safety being their top priority. Its all about perspective and what specifically you choose to look at.

I don't really have an issue with that tweet. The "be paranoid" statement is a just saying to pay attention in case there are issues. It's not a very professional way to say "pay attention to your surroundings" but that doesn't make it not a safety conscious statement.

It sounds like where we disagree is that I'm taking it as the company is being as safety conscious as possible while testing a experimental software. This is compared to the alternative in saying the experimental software itself isn't safe. And to be clear, I agree with that. In FSD beta 8.2 videos I have seen some scary situations, particularly in left turns across traffic, which could be very bad if the driver isn't paying attention.

Several videos have shown running over curbs (bent rims?). If you say bad things about Tesla your FSD beta may get removed or may not get invited to the next big thing, so there is an incentive to under report.

I'll give you that one, I have seen a few instances of curbed wheels in videos. No one wants that to happen, but that's not what I'd consider an accident in the context of what I was talking about. I was meaning more like a single or multi car accident where insurance was involved.

 

[helvio]

@gearchruncher - If Tesla's top priority isn't safety, what is it? Last I checked, you're not the chairman making the decisions, so I really question your assertiveness.

I've never seen cars as safe as Teslas. Don't trust me, go to NHTSA and look it up: Ratings | NHTSA
Search for TESLA and go through all of the 10 pages. It would make an OCD person climax. 99 results, ALL of them being 5 stars - with the exception of the not rated ones.

Now we get presented to FSD. An optional package. OPTIONAL. And no less than the CEO says VOLUNTARILY and PUBLICLY, it's a BETA, treat it as such!
Sorry, I don't have a problem with that. It does not negate that safety is the top priority at Tesla. As you said, there's no 100% safety. It's a race of the 9s. Tell me WHAT about Tesla has more 9s than Safety, please?