FSD beta in its current form needs re-evaluation.

[Matias]

FSD beta in its current form needs re-evaluation.

Currently Tesla pushes FSD beta (actually more like alpha) versions to those users, who want to participate. Yesterday’s catastrophic release shows, that sometimes releases contain extremely dangerous errors. This can’t continue like this. Either Tesla redesigns its program, or it is forced to do so.

One minimum solution would be, that public beta testers have one version older software than internal beta.

 

[Spacecadetwzrd]

Opt out it’s that easy

 

[thewishmaster]

One minimum solution would be, that public beta testers have one version older software than internal beta.

What makes you think they don’t? It’s probably some integration step with the production firmware that failed. Internally their custom builds had to have worked just fine for them to not notice something so major

 

[Matias]

Opt out it’s that easy

Other road users can't.

 

[drtimhill]

One minimum solution would be, that public beta testers have one version older software than internal beta.

You realize that is already the case, right?

 

[Matias]

You realize that is already the case, right?

Apparently internal testing is not doing very good job then.

 

[Matias]

Other road users are not asked for agreement..

e.g. white car in 0:23 was not asked, whether it wanted to participate in this testing.

 

[drtimhill]

Apparently internal testing is not doing very good job then.

In this case yes, something slipped through. This usually happens for a "slipstream" fix when someone gets permission to make a very late change just before release without redoing the full test cycle. And in this case it was a very BAD choice. Hopefully someone will look very hard at how that slipped through. In my experience such late fixes are usually done under time pressure releases when management overrides engineering.

 

[Terminator857]

... Yesterday’s catastrophic release shows, that sometimes releases contain extremely dangerous errors. ...

You have an odd definition of catastrophic. Lets look at a regular release: from release notes it says: it can do the worse possible thing at the worst possible time. If you don't like it and are not willing to control its use, then you should opt out of the beta program.
Instead of catastrophic, I'd prefer to call it progress.

 

[McMoo]

You have an odd definition of catastrophic. Lets look at a regular release: from release notes it says: it can do the worse possible thing at the worst possible time. If you don't like it and are not willing to control its use, then you should opt out of the beta program.
Instead of catastrophic, I'd prefer to call it progress.

I don’t think anyone thought that could be full emergency braking on the highway. That’s not something you can interject quickly on and recover from by taking over. Turning into the wrong lane or other things like that are completely the drivers responsibility to notice and take over quickly before an accident.

 

[JHCCAZ]

I don’t think anyone thought that could be full emergency braking on the highway. That’s not something you can interject quickly on and recover from by taking over. Turning into the wrong lane or other things like that are completely the drivers responsibility to notice and take over quickly before an accident.

I'm not totally clear on this, but one of the things reported by the people having the hard braking events was that they could not override at all. If it's true that the car is doing something that is impossible to correct, that changes the concept of human-tester supervision.

I'd like to say that there should always be the capability to override in AEB event. But thinking about it further, it's not completely obvious how that should be implemented. I read reports that users were mashing the accelerator pedal all the way to the floor in an attempt to override the false braking (which had no effect). But as frustrating as that surely was, that's probably not a safe override input in a real AEB collision avoidance event.

If a driver is startled by an impending crash/AEB and reacts instinctively by trying to stomp the brake pedal himself, but gets confused in the stressful moment and stomps the Go pedal insread - quite possible and a known mistake scenario - it wouldn't be good to allow the pedal stomping to cancel a true AEB action.

I'd be interested in anyone's suggestion about how to enable Tesla drivers to correct for phantom AEB in a way that's less likely to accidentally cancel a real AEB. Something more helpful than "Tesla should never have a false AEB event". Of course it needs to become rare, but amost all L2 cars on the road today have occasional false alerts.

 

[McMoo]

Smashing the accelerator did override for me but the rate at which it slammed on the brakes probably brought us from 70 to 50 in the time it took to react.

 

[dogmodes]

I don’t think anyone thought that could be full emergency braking on the highway. That’s not something you can interject quickly on and recover from by taking over. Turning into the wrong lane or other things like that are completely the drivers responsibility to notice and take over quickly before an accident.

Also let’s not forget these events were happening with FSD disabled. They made fundamental changes to how the car operates without warning us. Then it seems they may have toggled off the safety features on cars without telling us as a fix. All of the constant Tesla apologizing gets old. They’re lucky so far we haven’t heard of anyone getting hurt.

All anyone is saying is they need to plan and communicate better and have stricter controls. But somehow people argue that.

 

[JHCCAZ]

Smashing the accelerator did override for me but the rate at which it slammed on the brakes probably brought us from 70 to 50 in the time it took to react.

That's good, but I think there were a few messages along the lines of this one that was just posted in another thread:

...Another key detail of yesterday's issue was that when the vehicle phantom emergency braked, depressing the accelerator had no change until the vehicle gave back control (for me it was around 10mph). So the brake followed by loss of acceleration control is really where the danger was. ...