Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

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

This site may earn commission on affiliate links.
Regarding FSD readiness for Robotaxis.
I think Tesla won’t need to really solve FSD to rollout Tesla Network.

What they can do is as said here already:
  • Rider responsible to monitor/drive the car during a trip.
  • After a trip Robotaxis drive itself carefully to pickup another passenger, when nobody is in the car.
For during trip with supervision they can already do it, no legislation needed.

For empty car, it’s an much easier FSD problem to solve then there are passengers sleeping in the car, because:
  • The car can drive overly cautiously, allow itself to be bullied all the time without angering a passenger.
  • With self sacrifice as a viable option, they don’t need to solve the “trolley car” problem, which I believe doesn’t have a solution.
  • Also by driving really cautiously, I think the car could avoid putting itself in an “trolley car” situation in the first place, just don’t go to any place you’re not very confident about.
For anyone who doesn’t think “trolley car problem” is a real problem, image the car is on a train track, but both front and back are deliberately blocked by some bad people who won’t budge, and a train is coming.
Do you think a FSD car should run over those people to protect the passenger?
To solve this, a car might need to be self conscious and maybe have some legal rights and responsibilities.:D
 
For anyone who doesn’t think “trolley car problem” is a real problem, image the car is on a train track, but both front and back are deliberately blocked by some bad people who won’t budge, and a train is coming.
Do you think a FSD car should run over those people to protect the passenger?
To solve this, a car might need to be self conscious and maybe have some legal rights and responsibilities.:D

Or just drive on the rail tracks and out-accelerate the train? :D

I think the better, more realistic example is: if a pedestrian unexpectedly and illegally steps in front of the car within stopping distance, should the FSD car swerve in front of a big truck approaching fast in the opposite lane, or hit a pedestrian at high speed and avoid collision with the truck?

Most human drivers instinctively swerve, fishtail, and get killed by the truck in a high speed side collision. Many human drivers will react similarly and possibly kill themselves trying to avoid hitting a cat or a dog ...

Here's a "lucky pedestrians" compilation:


I think that legally the best policy is for the car to always maximize occupant survival: if collision is unavoidable then pick the lowest energy collision instead of swerving into the opposite lane and collide with a truck, etc. Maybe always avoid hitting unprotected pedestrians if a lower energy collision with a hard object is possible. I.e. when passenger injury is low probability, protect pedestrians. In all other cases be a selfish bastard.

(Also rule #0: never, ever hit cats.)

If the occupants didn't want to be maximally protected by the car and its resident AI then they should not have picked a Tesla FSD car, which acronym also stands for "Fully Safe Driving". :cool:
 
Last edited:
Ok, never mind.

Also today at 2:31 pm, somebody bought 500,000 shares of TSLA ($130M worth). I have never seen that kind of single minute volume, even during the 'Boomberg bair-raid' last Fall.

Somebody knows something** (Not an Advice).

DateVWAPVolumeTrade$
2019-04-23 14-30$263.164,172$1,097,903.52
2019-04-23 14-31$263.57503,635$132,743,076.95
2019-04-23 14-32$263.7013,924$3,671,697.53

I also noticed that neither the SP or the volume changed much after that large sale, so the "somebody that knows something" is definately not me**.

The trade didn't affect the Order Book appreciably, so a prearranged sale then? Options executed? Ideas and/or explanations greatfully received, TIA.

Cheers!
 
Full disclosure and not-an-advice: Sold all of my $TSLA stock today as I expect the SP to get hammered after the ER. Bought $ARKK with the profits as Cathie seems far more nimble than I will ever be.

I will get back in soon as ~ Q2 is over.

IMHO, the only thing that will matter to the SP (I.e. something that Wall St cannot refute) until EOY (*) is cash balance and delivery #s... the rest is just noise to WS

(*) - EOY or when FSD is feature complete, whichever comes last :p
 
Also today at 2:31 pm, somebody bought 500,000 shares of TSLA ($130M worth). I have never seen that kind of single minute volume, even during the 'Boomberg bair-raid' last Fall.

Somebody knows something** (Not an Advice).

DateVWAPVolumeTrade$
2019-04-23 14-30$263.164,172$1,097,903.52
2019-04-23 14-31$263.57503,635$132,743,076.95
2019-04-23 14-32$263.7013,924$3,671,697.53

I also noticed that neither the SP or the volume changed much after that large sale, so the "somebody that knows something" is definately not me**.

The trade didn't affect the Order Book appreciably, so a prearranged sale then? Options executed? Ideas and/or explanations greatfully received, TIA.

Cheers!

Noticed it too, has the signature of an "arranged trade": where seller offers the shares in the open market as a 500,000 shares sell limit order at a prearranged moment and buyer who coordinates with the seller immediately consumes that limit order. It's usually done in quiet periods of the trading day.

This is often preferred to OTC trades which are more difficult to price.

The advantage is price discovery without slippage, plus a marketplace transaction that removes trust issues.
 
Let's try a little first-principles thinking.

For starters:
  • What thrilled me about watching the FSD webcast yesterday was the discovery that there are teams within Tesla that are obsessed with learning and good uninterrupted communications -- hooray!
  • Elon is deeply involved with these teams -- excellent!
  • The essence of Neural Nets is learning, based on massive ingests of data which is then fine-tuned and tweaked.
  • These teams have brilliantly seeded nearly the entire multi-hundred-thousand-car Tesla fleet with the ability to tell the mothership what's going on out on the road -- massive competitive advantage.
  • Tesla's FSD/AI teams obsess over the fleet sending back data on driver interventions, which arguably are failures of the car to do something right on its own while in autonomous mode
  • By having teams of annotators who go in, study the videos of what the car was doing when a driver intervened, and essentially teach the neural net how to do it right on its own, the NN learns, you avoid those types of interventions in the future
  • The result is a smarter, SAFER self-driving car, arguably leaps and bounds above any competitor for years to come
Okay, now imagine if Tesla approached its currently-so-so Company/Customer Communications (CCC) with the same level of obsession and passion. Imagine if Tesla realized that:
  • Every http request on the Tesla website is a customer intervention.
  • Every tap of a button in the Tesla app is a customer intervention.
  • Every visit to a store is a customer intervention.
  • Every order by a customer is a customer intervention.
  • Every email, phone call, text, etc. from a customer to an order specialist is a customer intervention.
  • Every call to the trucker asking when the car will be delivered is a customer intervention.
  • Every call to Tesla Roadside Service is a customer intervention.
  • Every repair situation is a customer intervention.
  • Every time a customer calls a service center to find out the status of their car: it's a customer intervention.
  • Every inquiry from a customer, from email, phone, text, in-person, whatever, all day long, 24/7/365: it's a customer intervention.
  • Every time a customer hangs up in frustration after being on hold for 30, 60, 90 minutes and not getting anywhere, it's a customer intervention.
  • Every time you try the VIP/executive "escalation" feature to get a response from Tesla, it's a customer intervention.
  • Every customer intervention is just as important, just as valuable (arguably moreso) than a driver intervention.
Now think about Tesla's corporate culture as just another neural net. One that right now is very dumb and running on the equivalent of a Pentium chip from 1994 with 16MB RAM. :)

If you consider all the times a customer attempts to "intervene" with Tesla, especially when Tesla's been nonresponsive/noncommunicative, you realize how effed up the company still is from the customer experience perspective. Sure, lots of you out there have great, satisfied experiences. I get it. Especially if you live in California. Lots of other places, not so much. Some are downright horrible experiences.

Compare the standardized homogeneity of the FSD universe in Tesla cars with the heterogeneity of Tesla's corporate culture in terms of CCC. Compare what Tesla can and will achieve with the exact same constantly-improving neural net downloaded into its entire fleet of cars, with the random/uneven experiences customers have with Tesla company interactions around the country/world. Everything in the latter case depends on the initiative/dedication/training/attitude/skills of the individual Tesla person you interact with. Get a good one, great. Get a so-so one, awful. (Imagine if some Tesla cars were just stupid and never "learned". Imagine if the fleet had tons of such cars in it.) Problem is, in the customer experience world, you deal with multiple Tesla employees. Customer issues are handed off to multiple people. Maybe the first one you deal with is great. Next one you're handed off to, not so great. One after that, great. Then you're assigned a total dud. And the dud is the most important one in the chain, and you never get an email reply or phone call reply from them. Ever. (That's my general experience.) Poorly managed expectations. Frustrated, exasperated customers. Fail.

Imagine in FSD division of the company, they discovered the driver intervention-count stayed high, perpetually. Elon would freak out, wonder WTF they were doing wrong, maybe heads would roll. But imagine what the customer intervention count is with Tesla every day. Who's minding that count, Elon? Who's obsessing about getting it to zero? Or think of it this way, where is the "March of Nines" for customer experience? How do we get to 99.9999999% blissful customer experiences?

So, let's fix this puppy. Imagine applying some first-principle thinking Elon-style to CCC.
  • Imagine every single Tesla owner, every one, including every would-be/wannabe future owner, those who are thinking about buying: they're all part of *the Tesla market*.
  • Think of the Tesla Market as just another fleet.
  • Tesla, you own this fleet too. But you're not doing much with it right now. Not efficiently anyway. Show it some love.
  • Imagine every time there is a customer intervention, Tesla studies it just as obsessively as FSD driver interventions.
  • Imagine every time there is a customer intervention, someone on this CCC "annotation team" studies why the customer intervened, and figures out how the company can avoid doing whatever it did that let to poorly managed expectations, unresponsiveness, 60+ minute on-hold experiences on the phone, etc., in the future, and feeds that back into the company DNA -- which one should think of as just another neural net, that is desperately eager to learn but is being starved of data.
  • Imagine how many customer interventions happen every day.
  • Imagine all the "annotations" Tesla would be faced with if they obsessed over them. Every day. It'd be a lot.
  • But over time the onslaught would drop fast--why, I would bet exponentially.
  • Imagine how fast they could cut down on the failures that triggered a customer intervention in the first place.
  • The Tesla corporate culture "neural net" needs training now how to improve, by new management practices, standards of accountability, and new technologies and business practices installed wherever they're needed most, to streamline/improve communications and responsiveness and overall customer experience.
  • The lack of accountability and the mismanaged expectations are what are triggering so much frustration on the customer side. Consider customer frustration or disappointment no different than a car operating under FSD having an accident.

There is a way to FIX the CCC problem at Tesla. So that the owner experience is as good, as brilliant, as game-changing, and as far out in the lead as the technology, the cars, and the software. THAT is what Tesla SHOULD BE. Tesla as a brand should mean EVERYTHING--products, ordering, service, app, the whole lifecycle of a customer, is way beyond every other automaker. You would have loyalty like you would not believe. Tesla's CCC experience is holding the company back from achieving this.

C'mon Tesla. You have to fix this. You have no choice. If you're betting the farm on a robotaxi future, face the reality: you have no robotaxi future if the customer experience of the robotaxi service sucks. And if you do nothing in the next 12-18 months, and launch the Tesla Network with today's CCC, the robotaxi service will suck, from a car-owner perspective and a passenger pespective. And you lose. We all lose.

Like Elon said yesterday about LIDAR: "Lame. Lame."

Don't be lame, Tesla.

This guy lists.
 
Or this:

giphy.gif

Well now that FCA is onboard with Tesla, we can get as many busted-azz old Dodge's as required.

I wonder if the SpaceX ed. Roadster will land better than the Gen. Lee? :rolleyes:

Cheers!
 
Firstly, FSD is clearly not feature complete until it can replicate simple traffic maneuvers that Russian soldiers performed in WWII already:

giphy.webp


As a second milestone I think FSD must be able to handle light traffic in Mumbai:

giphy.gif


Finally, any FSD system that wants to earn the prestigious "TMC Seal Of Approval" needs to be able to handle the following standard traffic scenario, where an ordinary pedestrian removes a stray cat from the road:

giphy.gif

If you doubters think the last one is fake, here's proof and full length footage showing that it's all real, with happy cat ending!

 
Feeling a lot better about the S/X refresh than the FSD demo. That's how you do it : take your strength and experience, improve on it beyond what a competitor can do and drop it on the market when you can actually deliver the goods. Well done. The usual outcry from customers who bought recently drive home how much of a value upgrade this refresh provides.
 
In fact, in the beginning this is how it will work, it’s like zip car. You drive it - with AP that does 90% of the work. As the NN gets better, we get march of the 9s and most trips will finish without any intervention. Behold full FSD. It will happen in some areas earlier than others. Finally, in 2100 or there abouts in India.

I see your issue now - you took FSD too literally.

That's not gonna work. If the car can't get to you by itself without a driver, the whole scenario is broken. It needs to be reliable enough that the incidents of cars being stuck somewhere and Tesla needing to go "rescue" them are quite rare. So sure if you can drive and you ended up with one of those "stuck" scenarios you could solve it yourself but that isn't a "feature" that would be of particular value or interest. Self-driving should mostly just work, only then we have something to talk about.
 
  • Like
Reactions: neroden
Kudos to @KarenRei for calling this refresh!

And let's not forget this gem in the same comment:

"The least likely thing we'll see, IMHO, is a switch to 2170s any time soon. The last thing Tesla needs is to needlessly burn 10GWh/yr of their much-needed 2170 production (which Panasonic always seems to lag on making more of), free up their excellent 18650 cells for competitors to buy (or reengineer their other product lines to use them), add an entirely new pack production line (since the pack wouldn't be the same as either the 3's or the current S/X's), reengineer the vehicle (including its production lines) and requalify it (battery packs are structural elements, you can't just arbitrarily change them.. also lots of safety issues to requalify), and dozens of other things."​

For investors, people around here sure do a lot of wishful/magical thinking. :p
 
I am such an idiot. Deleveraged yesterday in advance of earnings. Yesterday morning, during that low. And now they do the refresh unveil before earnings? UGH. Market is going to jump today. This is an event I really wanted to be in for! :(

Being right about the motor changes and the lack of battery pack changes are great, but they don't make up for being unleveraged when the changes get unveiled....
 
Last edited:
That's not gonna work. If the car can't get to you by itself without a driver, the whole scenario is broken. It needs to be reliable enough that the incidents of cars being stuck somewhere and Tesla needing to go "rescue" them are quite rare.

Costs of a "rescue" can be kept pretty low if careful "remote driving" by trained Tesla personnel can get the car out of its blockage.

This would scale very well:
  • Just a few minutes of attention from a "rescue driver" can put the car into FSD mode again.
  • If the average intervention takes 5 minutes, then a single operator in an 8 hour workday can handle up to ~100 interventions.
  • At an hourly wage of $30 the labor cost is only $2.5 per intervention.
  • High cost manual interventions would be relatively rare: no LTE coverage, or car malfunction.
Such a method would allow even relatively high intervention rates in the initial roll-out phase, or when new locations are added to the Tesla Network's coverage map.
 
  • Helpful
Reactions: VValleyEV
Being right about the motor changes and the lack of battery pack changes are great, but they don't make up for being unleveraged when the changes get unveiled....

It's so hard to get the timing right as option contract buyers: even if they are completely right about everything a single day (or even hour) of timing noise can have a large negative effect.

As a consolation, if today's Q1 results are as awful as even bulls are expecting then you'll be right to have deleveraged ...
 
It's so hard to get the timing right as option contract buyers: even if they are completely right about everything a single day (or even hour) of timing noise can have a large negative effect.

As a consolation, if today's Q1 results are as awful as even bulls are expecting then you'll be right to have deleveraged ...

That might numb the wound somewhat. But if the market reaction to the S+X changes is like I'm expecting, we'd be lucky to get back down to the upper 250s (aka yesterday morning) after even a bad earnings report. S+X demand fear was one of the primary fears for Q2. I'm expecting at least $270s today.

I do expect a bad earnings report. A non-GAAP loss of ~$150M would require a lot of wildcard income sources hitting at once, and I just don't see it as likely (even though there's a lot of them this quarter). I don't expect a "catastrophic" earnings report - I'm not a margins pessimist like a lot of bulls seemingly have become - but -$150M is a high bar to get over. Even accounting for the fact that I think the expectation is a miss on that.
 
Last edited: