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.
OT
OK, that's an interesting argument.

I suppose you might be totally right.

I won't take Uber at all for numerous reasons including their poor-quality, uninsured, underpaid drivers, but I realize many people do.

Uber's availability is startlingly limited, too, by driver availability -- you can't get them at random times, you can't get them in random places, they'll refuse to take you to random places.
So even if Tesla Network was geofenced away from tough-to-handle areas *and* unable to operate in bad weather, it would probably be no worse than Uber.

So then I guess we get back to regulation. When an Uber crashes, everyone blames the driver and the Uber company usually skates. (Though occasionally Uber gets banned from town for not following regulations.) If something goes wrong with a Tesla robotaxi, Tesla gets sued for millions, the media has a frenzy, and regulators go wild with bans on "killer robots".

If Tesla can get sufficient insurance, I guess, go for it.
I had to make an insurance claim against an Uber driver. Drivers insurance, after quite a bit of dragging their feet, finally said they won’t cover it. Went direct to Uber corporate. Once I finally got in touch with the right person at their insurance company, they ended up paying out quite a fair amount and didn’t squabble one bit with my demands. Just an anecdote but they seemed well insured.

Edit: Just went back through my email because I couldn’t for the life of me remember the insurance company. After finding it, I see now that it was Lyft and not Uber. My apologies.
 
Last edited:
He works on rocket design. He's been leading in the car and software industry for 15+ years. I think he knows the problem well. I've never heard Elon so confident about anything.

Elon's confidence is a contrary indicator.

Alien Dreadnought. "Six months definitely." "Funding secured."

If Elon starts making carefully hedged statements... that's a positive indicator, IMO. Means he's actually dug into the details.
 
I would argue that it would actually be easier to train a neural net for FSD in such scenario. The concept learned from such data would most likely be summarized as
1) You can drive on all driable area
2) Just don't come in physical contact with another object except the road.

I would argue that a NN can learn and apply these two rules for any unrecognized / danger scenario in cases where driver are unable to perform evasive maneuver.
If you drive with these - either you will never cross a "junction" or get into an accident.

You need to learn how to "earn" the right of way.
 
  • Informative
Reactions: neroden
The presentation has convinced me that Tesla is waaay ahead of all other companies attempting automated driving, including waymo. Whenever (probably much later than Elon hopes) the "feature completed FSD" Elon talked about gets released to the fleet, I think it will pass as a basic, rudimentary L3 system, and it will be the first such in the hands of the public.

That L3 FSD will require a fair amount of human intervention at start, but they will use their iterative learning engine to collect the edge cases, re-train NN, run in shadow mode, then release and start over. After some N rounds of these iterations the system may improve to the point of becoming an L4 system. Mind you, this may require a large value of N and consequently much longer time-frame than a car lease... Regulatory approval for driverless operation can only come after L4 is reached and safety of that is proven to the regulators.
So I do not see the robotaxi revenue materializing anytime soon enough to excite investors today.

As for reaching L5, that is another ballgame entirely. That would need to handle not only the crazy road in the Himalaya under the waterfall, but also cases like cargo falling off a truck ahead on the road, a semi-truck jackknifing, a bridge collapsing, mud-slide, flood over the road etc. Plus any of those in extreme weather condition, such as thick fog or blizzard.

In fact, I do not even believe that the current hardware is suitable for L5. An example for inadequacy: there are weather conditions in my neck of the woods in winter slushy-sticky conditions, when the cameras get covered in muck after a few minutes of driving making the super intelligent NN useless because it is blind. Only the forward facing cameras are cleaned by the wipers, the other 5 turns useless and the system (as witnessed this past winter) simply states that autopilot is not available. An L5 hardware needs wipers or water spray cleaning mechanism for all of its cameras.
 
Last edited:
Motor Trend has an article up about the Model S refresh.

EXCLUSIVE: 2019 Tesla Model S Review: From SF to LA on One Charge? - MotorTrend

We test it: Updated Model S can go from Bay Area to L.A. on a single charge ... As we pull into the [Hawthorne] Supercharger stall, our elapsed time from the Bay Area stood at 6 hours, 11 minutes, 359 miles. With 83 kWh used, we had 11 percent of the battery remaining—which equates to 41 more miles at the rate I was going. Right at 400 miles if you add it up

Tesla just set the bar so high that other EV manufacturers are curled up in a shower crying.
 
The third most likely thing we'll see is a drive unit swap over to the Model 3 drive unit, for one of the motors (aka longer range, faster charging (mph/kph), greater longevity (fewer cell cycles per mi/km), greater sustained track performance, etc, without changing anything "structural" in the vehicle, and drive units are far cheaper than battery packs and the Model 3 drive unit's production process is probably the smoothest running, most automated thing that Tesla does.

Kudos to @KarenRei for calling this refresh!
 
Haven't been following the thread since the S/X news broke so if someone has already suggested this then oh well.

Does anyone get the feeling we're going to see 3 updates to the S/X line between now and Q2 2020? So this one, with the drivetrain update, suspension, slight range increase. Another one in Q3 with the interior refresh. And the 3rd one in Q2 2020 with the new battery pack(since Elon just mentioned it and we know they want to incorporate Maxwell tech) that will give a much bigger range increase.
 
Yeah, I don't put too much stock in this report. 3000 cycles would take over 2 years of continuous charge/discharge cycling at 0.25 C which is the avg rate at which a Model 3 LR discharges.

Note: 4 hrs to charge @ 0.25C, 4 hrs to recharge gives 3 cycles per day. 3000 cycles takes 1000 days. That's 2.75 years for a realistic full test to 3000 cycles as they claim.

But the very first Model 3s are now just 1.75 years old. This report is fudging something.

I'd be more swayed if this report included a graph of Cycles vs Capacity, but all they offer is financial cost comparisons to diesel. Or if the author provided a link to the actual study instead of just "The reports are available at WirtschaftsWoche." Unconvincing.

So I'll stick with Dr. Jeff Dahn's results from Dalhousie University: 1500 cycles to 80% capacity was acheived with Tesla's new battery chemistry within a year after his testing contract begain in 2015. I'm confident current Model 3s and S/X is using that chemistry, until evidence indicates otherwise.

BTW, 1500 cycles is in line with the real-world results being achieved by Tesloop, the California/Nevada ride operator with its high duty cycle fleet:

"Between 194,000 and 324,000 miles Tesloop experienced battery degradation of ~22% (see below for details)"

Note thats approx. 1200 cycles to 22% degradation, the typical end-of-life capacity, but Tesloop was charging to 100% instead of the recommended 90%. Still not 3000 cycles:

Tesloop’s Tesla Model S Surpasses 400,000 miles (643,737 kM)

Unfortunately, the article you link to provides no evidence, it just cites an unreferenced report to support its other claims. Rating: Dubious.

Cheers!
They likely conducted experiment with controlled elevated temperature to simulate accelerated degradation and extrapolate back to normal operating conditions.
 
  • Funny
Reactions: Artful Dodger
Haven't been following the thread since the S/X news broke so if someone has already suggested this then oh well.

Does anyone get the feeling we're going to see 3 updates to the S/X line between now and Q2 2020? So this one, with the drivetrain update, suspension, slight range increase. Another one in Q3 with the interior refresh. And the 3rd one in Q2 2020 with the new battery pack(since Elon just mentioned it and we know they want to incorporate Maxwell tech) that will give a much bigger range increase.
Absolutely. This is done so that people don't feel like they got screwed. Slowly introduce new features. Think they did the same thing with the P85D to P90D (Nosecone) to P90D (Refresh front fascia) to P100D. It's a smart way to do it. I'm not buying one until the interior refresh though. I like my P85D+ unicorn too much
 
OK, now this is a really interesting argument to me. Because I think a good level 3 system is what's possible in the near future.



I guess it raises the question: How comfortable are you taking a robotaxi which may, on rare occasions, randomly force you to take over midroute? (Or randomly pull over and stop dead, forcing you to call someone to pick you up?) Tesla definitely is not currently planning to implement remote drivers.



Geofenced, it would work. Musk keeps saying nothing will be geofenced, though. And is not interested in talking about remote drivers. :gnashes teeth:

Elon did briefly mention remote driving in response to a question. He referenced something like ET phoning home.
 
Good, enough with the S&X. Back to FSD.

Its not full FSD if it doesn't know how to handle this.

ps : Just to be clear, this is good, well behaved traffic.

View attachment 399990

J9ZDeE.gif
On the serious note, people seems to worry about path selection and proceeding part, which I think is misplaced attention.

The perception part is much harder and much more important. Flaws in that part is what got the passenger killed.

Once the car understand what's going on around it, making decisions to move around is not easy but much easier than the perception part, and we have lots of robotic research to tap into. Remember open AI won some title in gaming, beating the human champion, a bulk of the system can be borrowed for this very purpose.
 
BTW, 1500 cycles is in line with the real-world results being achieved by Tesloop, the California/Nevada ride operator with its high duty cycle fleet:

"Between 194,000 and 324,000 miles Tesloop experienced battery degradation of ~22% (see below for details)"
Note thats approx. 1200 cycles to 22% degradation, the typical end-of-life capacity, but Tesloop was charging to 100% instead of the recommended 90%. Still not 3000 cycles

EHawk is actually an anomaly, with some misdiagnosed problems which resulted in a pack replacement which wasn't needed, and the replacement was faulty.
August 2016
Before a firmware update that fixed this issue the vehicle’s range estimator became inaccurate. The estimator would decrease 10 miles even though the vehicle didn’t actually lose range. Upon inspection Tesla found there to be a battery chemistry issue that the software wasn’t calculating correctly prompting the service center to change the high voltage battery for safety and to study. 3 months later a firmware update was released, which had it been released 3 months earlier, Tesloop would not have had to change the battery.
Tesloop’s Tesla Model S Surpasses 400,000 miles (643,737 kM)

Better to look at all their X90's on original packs with 250-360K miles
Tesloop's High Mileage Teslas
 
  • Informative
Reactions: Yuri_G and AZRI11
So, to S/X. In addition to the customer-facing stuff announced in the update....

S & X are now using essentially the same drivetrain as Model 3. This means ECONOMIES OF SCALE. Production costs down, margins up.

---

On a personal note, I will also note that it is now possible to order more-or-less the same model I received back in 2013 for... roughly the same price. ($83,700 vs. ~$80,000, including dest & doc fee.)

But with Autopilot, dual motor AWD, slightly longer range, faster charging at Superchargers (though slower AC charging).
Different (substantially worse) infotainment software, a smaller frunk, a box in the purse slot, the glass roof, a slightly different front end, and slightly different seats, so there's that.

But still, that's an amazing value. Inflation applied to my original car price would have brought the identical car up to $87,295. Tesla really has been pushing the cost structure down.
 
Too bad SpaceX didn’t just test with a new capsule like Boeing did. SpaceX’s capsule has been to space and in the ocean. Hopefully they will just find an issue with the salt water. But in the long run, this will still be a good thing for SpaceX. Better to find out any weaknesses now.

One small problem with your statement. While SpaceX was planning on reusing the DM-1 for inflight abort, Boeing isn't/ didn't/ won't be doing an inflight abort test.

They did both use fresh capsules for pad abort tests.
 
Whew....at the hotel and finally caught up....you all are busy little beavers!

With all the L5 and Robo-taxi talk has anyone been able to prove (to the extent possible) that the Spontaneous fire in the Tesla in China was in fact staged?