Musk tweets, traffic lights, stop signs and roundabouts are in testing

[berkeley_ecar]

3mm6md

Please don't take this the wrong way but this standard response by so many posters across multiple forums is getting awfully tiring.

I can't be the only forum participant who has NO CLUE as to what "3mm6md" means, so please do inform the rest of us. A quick search using Google provided no help. I'm guessing something along the lines of "3 months mentioned, 6 months delivered."

 

[alexdav]

I can't be the only forum participant who has NO CLUE as to what "3mm6md" means, so please do inform the rest of us. A quick search using Google provided no help. I'm guessing something along the lines of "3 months mentioned, 6 months delivered."

First time I've seen it, but I'm guessing it's "3 months maybe, 6 months definitely".

 

[Electroman]

Elon has been wrong - a bit consistently on EAP and FSD features and timelines. I don't fault him. When you push the boundaries you don't know what you don't know. So I and most Elonbots already know that.

But Elon has almost always achieved everything else he said he will do, a bit late mostly, but he did what he said.

 

[Kuhz]

First time I've seen it, but I'm guessing it's "3 months maybe, 6 months definitely".

This

Stole the abbreviation from @BerTX so he owns the rights

 

[Darren Donovan]

The difference between this tweet and gross exaggerations like you can summon your car from NY while in LA is that the technology to detect lights, stop signs, etc already exists with other self driving cars. So it’s not a stretch to believe it can come to our Tesla’s at some point. That fact that he’s openly talking about it again is encouraging to me. This is the next major leap for AP and I for one can’t wait to get it.

So whatever EM tweets we have to run it through our own "interpreter" or "translator" software and then have to temper it with reality. Whatever happened to just releasing factual, realistic information to the public. Tesla is a big and very important company now, not a startup in some garage. Act like it.

 

[electracity]

So whatever EM tweets we have to run it through our own "interpreter" or "translator" software and then have to temper it with reality. Whatever happened to just releasing factual, realistic information to the public. Tesla is a big and very important company now, not a startup in some garage. Act like it.

The irony is his developers probably all view these tweets as b.s. When you knee deep in vast complexity it is not motivational for the CEO to constantly indicate that you've just about got the system ready.

In the future AP will show remarkably complex behavior as Tesla approaches FSD. AP currently is very useful in some situations, but it is still trying to get the basics right.

 

[OPRCE]

Oh yea of little faith, why doest though doubt?

"Oh ye of little faith, why dost thou doubt?"

 

[ItsNotAboutTheMoney]

So whatever EM tweets we have to run it through our own "interpreter" or "translator" software and then have to temper it with reality. Whatever happened to just releasing factual, realistic information to the public. Tesla is a big and very important company now, not a startup in some garage. Act like it.

Just run it through you logical brain.
Don't let your subconscious linguistic interpreter fill in gaps.

He said soon, which is meaningless.
He said will be able to, which doesn't say how well.
But, it's in development.

 

[Matias]

Elon's tweets mean nothing anymore, other than he's having a sleepless night maybe. He tweeted so many things which never came, even though the tweets clearly said soon, or even specific timelines like 3 months, 6 months, end of year. My AP1 car is still not finding me anywhere on private property like Elon told us during launch, it doesn't produce 691hp like Elon tweeted (to be fair, he tweeted 700hp P85D, the website said 691hp), it doesn't have an SDK for me to write my own apps which was "just around the corner" in 2013, etc, etc. Evidently his work on a time machine didn't work either or he would have gone back to fulfill his "coast to coast FSD demo by end of 2017"..

Good thing with a time machine is, that you’re not in a hurry to make it.

 

[Matias]

I think they are actually getting closer this time around..

This is most likely technically correct.

 

[ai4px]

So..... about 3 months ago Ford had a press release that they have managed to get two cars to communicate with each other. Isn't that quaint. AP2.5 Teslas will soon be on to recognizing traffic lights. They already change lanes around slow drivers and take exits from the interstate.

 

[Cheburashka]

AP2.5 Teslas will soon be on to recognizing traffic lights.

Yep, right after your car find you anywhere on your property and after the coast-to-coast FSD drive.

 

[whitex]

I think they are actually getting closer this time around. When they have the initial platform running, everything becomes easier afterwards.

Of course they are getting closer, but it doesn't mean that they are close at all. If you start walking from Seattle (USA) to London (UK), every step you take you get closer, but just because you took 100 steps today vs. yesterday's 10, doesn't mean you're close. There is an ocean to cross along the way you for which you don't have a plan how to walk across too, but you're not at the beach yet so why worry about it.

The approach of "let just get closer" works for some things, no doubt, but self driving is one of those "all or nothing" goals. Imagine how dangerous a self driving car would be if it was so good, it would require human intervention once a year to prevent it from a bad collision. Drivers would start to rely on it and will not be ready to prevent their own demise when than once a year time comes. Even you if you get lucky and don't die the first time, over the years it's almost guaranteed to kill you.

 

[emmz0r]

Of course they are getting closer, but it doesn't mean that they are close at all. If you start walking from Seattle (USA) to London (UK), every step you take you get closer, but just because you took 100 steps today vs. yesterday's 10, doesn't mean you're close. There is an ocean to cross along the way you for which you don't have a plan how to walk across too, but you're not at the beach yet so why worry about it.

The approach of "let just get closer" works for some things, no doubt, but self driving is one of those "all or nothing" goals.

Indeed.
Either they are within reach, or the bet on pure deep learning is a HUGE miss and they have to rewrite everything.
I think they are within reach.

 

[whitex]

Indeed.
Either they are within reach, or the bet on pure deep learning is a HUGE miss and they have to rewrite everything.
I think they are within reach.

What does that mean "within reach"? Whose reach, by when?

 

[emmz0r]

What does that mean "within reach"? Whose reach, by when?

I think they are on the right track and they will gradually reach their goals. It won't be like 5-10 years and then BAM level 5. I actually believe Elon when he says it handles roundabouts and stuff ,and then a lot of progress is made. If this method proves to continue to be successful, it's "just" a matter of testing and tweaking. You will have some leaps of progress, then a quiet period of testing and tweaking etc. It will feel like you are getting something new all the time. It won't be that boring "in 5 years" crap.

 

[whitex]

I think they are on the right track and they will gradually reach their goals. It won't be like 5-10 years and then BAM level 5. I actually believe Elon when he says it handles roundabouts and stuff ,and then a lot of progress is made. If this method proves to continue to be successful, it's "just" a matter of testing and tweaking. You will have some leaps of progress, then a quiet period of testing and tweaking etc. It will feel like you are getting something new all the time. It won't be that boring "in 5 years" crap.

Incremental path to level 5 will be very dangerous (so I guess not boring)- the better the car gets, the more dangerous it is because people inherently disengage their attention (you can read on the internet studies by Google/Waymo, Toyota, and independent academics who concluded that 4 years ago). The side danger is that people dying in such almost-but-not-quiet-full-self driving vehicles can cause a public panic which leads to an overreaction by the politicians, which could hold the entire field back for years.

 

[emmz0r]

Incremental path to level 5 will be very dangerous (so I guess not boring)- the better the car gets, the more dangerous it is because people inherently disengage their attention (you can read on the internet studies by Google/Waymo, Toyota, and independent academics who concluded that 4 years ago). The side danger is that people dying in such almost-but-not-quiet-full-self driving vehicles can cause a public panic which leads to an overreaction by the politicians, which could hold the entire field back for years.

Isn't that exactly the situation we have today?
I think it has to be done like this, or level 5 will never be ready.

 

[Charliek]

While we are all " waiting" for these features, . I'll mention a few things. while I use NOA on all the main roads here in VA and autopilot on the side roads. it does a pretty decent job with light to moderate traffic. I also use it a lot in stop and start traffic just sititng in the right lane letting it do its thing . I personally like it for the less taxing drive it gives me, ( ps I do ignore some of the lane change recommendations as they are iffy, mostly too early) . On some recent trips to Maryland ( the land of "round abouts" here in the east!) I decided to experiment with a few of them letting the car in NOA and regular autopilot try to enter them (I'm on ver 48.4 I thnk , MS AP2 ) . as a systems engineer it is fun to watch a program hit a state it was unprepared for and see what it does . as a driver not so much . in this case the car slowed in the left lane tried to follow the car in front of it, didn't know what to do with the car coming around the round about ( which had the right of way). on the three time I tried it, it responded differently as the turn into the round about was essentially a right turn which put it to the stops of the turn algorithm, twice warning me with the red "take over immediately" ( I only do those sort of things when I am by myself, not the family) . once it shuddered trying to decide . until I took it out of AP . when in NOA it warns you that an unsupported maneuver is coming up and it pops out of NOA before the round about ( that is new). My guess is when we see this as a feature it will be very conservative, will enter the round about slowly and will have to pick up the exiting cars in the circle, and then entering cars in the round about from the other feeders. ( some have like seven entrance /exits) and since NOBODY does it right it will do alot of stopping and waiting locking onto in circle traffic and watch entering cars and then hitting the brakes. lets put it this way, I wont be napping when a round about comes up . would love to try this on some of the round abouts in DC.

 

[whitex]

Isn't that exactly the situation we have today?
I think it has to be done like this, or level 5 will never be ready.

That is the situation with Tesla, not true for everyone. Waymo has made a decision to proceed directly to Level 5, not L2, then L3, then L4. They are the technology leader right now. Just because you are aiming at Level 5 and nothing short of that, doesn't mean you can't have trained safety drivers and field tests. It does mean however that you are not selling Level 3 or Level 4 cars to the public for untrained drivers to endanger themselves and others.