Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

[cab]

I just canceled by MYP order to wait for HW4...and specifically RADAR. However, I think there is a very real possibility radar will be eliminated AGAIN in the future due to sensor fusion challenges or relegated to use on edge cases only. Minimally, I'm guessing radar will be in place for quite a while (gathering data) before they actually start using it.

 

[navguy12]

I have way way way less phantom braking vs AP from when I first got the car in 2018. I am on fsd beta so sensors are turned off. So you win some you lose some. Point is, no one talks about it anymore.

When I tried to drive with FSD beta with my May 2018 build TM3 on two lane highway (Ontario) 7 four weeks ago, after 10 minutes I turned it off as it was too scared of its own shadow. Jittery especially with long shadows perpendicular to the road.

 

[jerry33]

Same here. Only downside to driving this car long distances is the phantom braking issues, often weather related. Sure wish it had dumb cruise control.

I have about 35,000 miles of road trips and I find it far better than dumb cruise control. It’s been a long time since I had phantom braking on the highway. YMMV
City driving is another story.

 

[gsmith123]

I would have thought that to be safer than humans (kind of a low bar) the car would have to see better than humans.

Counterpoint: even if the perception is "only" as good as humans, the car has substantially faster reaction times and is always paying attention.

 

[Knightshade]

That is the rumor. I don’t think Tesla has confirmed it.

https://twitter.com/x/status/1633136225954455553

That's the new radar in the Tesla parts catalog showing where it mounts.

 

[cab]

The old radars were kind of awful, as in they have very poor lateral resolution. They can't detect if the thing they see is in your lane, above you, beside you, etc.

It's an admission the old radar was insufficient, but not necessarily an admission that deciding to remove the crappy radar was a bad idea.

(All of that said: we still don't have definitive confirmation HW4 cars have radar!)

True on radar confirmation even though Tesla's own API in reporting the RA2 code (v2 radar)....we just need some willing Model S/X owner to be the guinea pig and pull off their bumper. I'm guessing we will see that in the next 30-45 days though.

 

[navguy12]

Don't think parking sensors need to worry about that. And as for fog and mist, camera will see as far as a human and human drives just fine. Will just need to reduce speed.

Nope, my eyes are much better in snow removal situations where brine spray is involved. Vision only system is useless, doesn’t work in that scenario.

 

[Grifon]

Isn't Tesla adding radar back in with HW4? If so that would seem to be an admission that removing it was a mistake.

The radar being added back in is a HD radar, per Elon it does provide value beyond just the visual sensors.

 

[navguy12]

I still keep my stance that people put too much emphasis that a robotaxi needs to drive like a human. It's about being as safe or safer than a human, not to drive like a human (because humans are not safe drivers, they never follow traffic laws and treat everything like it's a suggestion). In a world with only robotaxies, people will be okay with it going 20 on a 50 if there's fog. A human would try to go as fast as he finds tolerable in similar situation as humans always overestimate their abilities.

We know that that robotaxi services like cruise and waymo has speed limiter under the speed limit and takes as many easy routes as possible. Only Tesla folks want a Tesla to be able to complete unprotected left turns over 3 lanes with traffic running at 60mphs.

My prediction is, Tesla will remain as L2 if you want it to drive more human like. Driving speed and minor geofencing will set in for any L4 services, nerfing the ability for Tesla to do really difficult situation. Then maybe after a decade that can relax.

Perhaps true. But in my current real world situation I have to make a routine commute, once a week, between Guelph ON and Ottawa ON at the prevailing traffic speed. And on the 407 ETR that is 120 kph in brine spray.

 

[gsmith123]

we just need some willing Model S/X owner to be the guinea pig and pull off their bumper. I'm guessing we will see that in the next 30-45 days though.

Come Model S/X buyers, why haven't you torn your car apart yet!

I think it's more likely than not they have radar, but it's hardly certain until we see it. The radar itself may not be done with design and testing, the supply chain may not be ready (or supplying them at enough volume), maybe they're going to stick it in a few cars to collect real world data about different radar designs (this seems the most unlikely), etc.

They've got a lot of freedom because the radar isn't required for anything and they've been careful to make zero public commitment.

 

[navguy12]

I have about 35,000 miles of road trips and I find it far better than dumb cruise control. It’s been a long time since I had phantom braking on the highway. YMMV
City driving is another story.

What percentage would your road trips be two lane highways? I agree, on limited access freeways in good weather, the phantom braking episodes are relegated to ascending ground approaching a bridge where the optical illusion is said bridge is a static barrier.

 

[AndrewZ]

I still keep my stance that people put too much emphasis that a robotaxi needs to drive like a human. It's about being as safe or safer than a human, not to drive like a human (because humans are not safe drivers, they never follow traffic laws and treat everything like it's a suggestion). In a world with only robotaxies, people will be okay with it going 20 on a 50 if there's fog. A human would try to go as fast as he finds tolerable in similar situation as humans always overestimate their abilities.

We know that that robotaxi services like cruise and waymo has speed limiter under the speed limit and takes as many easy routes as possible. Only Tesla folks want a Tesla to be able to complete unprotected left turns over 3 lanes with traffic running at 60mphs.

My prediction is, Tesla will remain as L2 if you want it to drive more human like. Driving speed and minor geofencing will set in for any L4 services, nerfing the ability for Tesla to do really difficult situation. Then maybe after a decade that can relax.

Haha this is an interesting angle to contrast Elon’s statements about robotaxis increasing vehicle utilization: if the cars are traveling half as fast as humans, they’ll be utilized twice as much by default!

Robotaxis don’t need to be as safe as humans, Elon says that is easy. Robotaxis need to be 10x, maybe 100x, maybe even 1000x safer than humans and that is what Elon says is hard.

 

[Nocturnal]

Haha this is an interesting angle to contrast Elon’s statements about robotaxis increasing vehicle utilization: if the cars are traveling half as fast as humans, they’ll be utilized twice as much by default!

Robotaxis don’t need to be as safe as humans, Elon says that is easy. Robotaxis need to be 10x, maybe 100x, maybe even 1000x safer than humans and that is what Elon says is hard.

If it's easy then why isn't it done yet?

 

[dhanson865]

I'm pretty sure V2 minis they are currently launching do not have the large (25 square meter) cell phone antenna.

"SpaceX will be ready to launch Starship in a few weeks, then launch timing depends on FAA license approval. Assuming that takes a few weeks, first launch attempt will be near end of third week of April, aka …" - Elon Musk

I fully believe there will be full sized V2 sats in orbit this spring (early to mid 2023) and significant coverage by Fall 2023.

So that in the US Tmobile phones won't ever go to 0 bars. The only question is when will Tesla switch away from AT&T as the cell provider in the cars? Can the cars with LTE modems just accept SpaceX/Tmobile data with a software update?

 

[Singuy]

I would have thought that to be safer than humans (kind of a low bar) the car would have to see better than humans. Now I don’t know if it’s better to have cameras that see further into the IR and UV range or to use an ancillary device such as radar. I do know that people won’t adopt it initially if it goes too slow.

See this is where I think people gets it wrong. The safest super human ability from a robotic taxi is it's ability to remove temptation. It wouldn't drink, watch a movie, or read a text. This is obvious. What is not obvious is that a robotic taxi can drive 15 mph under the speed limit for 10hrs and a human couldn't do it for 10 seconds.

Go to an open empty back road. Try driving that entire road 15mph under the speed limit for a hour. See if you will go crazy.

 

[jerry33]

What percentage would your road trips be two lane highways? I agree, on limited access freeways in good weather, the phantom braking episodes are relegated to ascending ground approaching a bridge where the optical illusion is said bridge is a static barrier.

Around 20%.

 

[jerry33]

See this is where I think people gets it wrong. The safest super human ability from a robotic taxi is it's ability to remove temptation. It wouldn't drink, watch a movie, or read a text. This is obvious. What is not obvious is that a robotic taxi can drive 15 mph under the speed limit for 10hrs and a human couldn't do it for 10 seconds.

Go to an open empty back road. Try driving that entire road 15mph under the speed limit for a hour. See if you will go crazy.

If a 12 hour trip takes 15 hours it will be an issue for many. 100% agreed for city/suburban driving.

 

[AndrewZ]

If it's easy then why isn't it done yet?

FSD already is safer than humans according to Tesla’s stats

What doesn’t exist is a level of safety mitigating sufficient risk to allow assumption of liability for the dynamic driving task

 

[Nocturnal]

FSD already is safer than humans according to Tesla’s stats

View attachment 918511
What doesn’t exist is a level of safety mitigating sufficient risk to allow assumption of liability for the dynamic driving task

Yes but that's with a small set of very dedicated humans monitoring it. The software on it's own is nowhere close.

 

[Webeevdrivers]

Big deal until Tesla vision becomes just as same or better, just like no radar.

Sure. But that is probably years away. Robotaxis and full FSD was only supposed to be a few years away five years ago. Now it’s looking like we are a decade away or more.

It just seems like an easy option that they could add for not a ton of money. Some people like it and find it very useful. Both for parking and auto park reasons.

Jmho.