I’ll give it a shot.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable
- Thread starter AudubonB
- Start date
cliff harris
Member
I would caution against thinking this. It does seem to be the case that the existing hardware CAN do serious robotaxi-like performance for many drivers already. I suspect the improvement that is needed is software, not hardware. It might mean more training data and more time on a larger compute cluster to do training, but training is totally different to inference. I would not be surprised to find that we eventually get robotaxi performance from HW3,4 and 5, just with later hardware perhaps being more confident, and able to drive more assertively.
willow_hiller
Well-Known Member
If this LinkedIn post is accurate, it sounds like they're further along than we thought. They may have all the infrastructure in place for ride-hailing, and just haven't yet tackled stuck vehicles at scale.
To be fair, neither had Waymo or Cruise. They just put the hazard lights on and wait for a driver to come assist them.
philw1776
Member
You are not alone with driveway.
FWIW I get disengagements most drives. Car asks why and my thought is "WTF? I don't know". Latest release, FSD Bater.
As someone working in logging (with chainsaws) which is the most dangerous job in the USA, a ladder is still the only reason we had a serious injury (dislocated shoulder from failing to follow 3 points of contact rule). Wear your finest safety harness!
This is too vague: "serious robotaxi-like performance already". This handwaves away the reality of the vast distance between current solution and a solution that drives with > 100x reduction in critical disengagements needed for a robotaxi.
Will we have a the same model structure & size for the robotaxi-level model as today, just it was trained on with more data?
Will the sparse occurence of some hard-to-detect object not be easier to detect with cameras with 4X more resolution, leading to slightly less accidents?
If you took a set of 1,000,000 drivers and had them drive some days with normal glasses and some days with slightly out of tune glasses, do you think there wouldn't be any difference in incident rates?
Do you think that the trend in LLMs / transformers and neural scaling laws saying that accuracy improves with more compute, more data, and larger models does not apply to FSD?
I choose to simply think more conservatively - the HW4 compute and sensors will allow for a larger, more accurate model than HW3 when trained on the same amount of data.
At some point, HW4 solution will be more accurate than HW3 and continue to be so moving forward. This is likely to happen before either solution is at robotaxi level.
Meaning - HW4 will reach robotaxi level before HW3. There will be a period of time where HW4 cars have robotaxi-level reliability and HW3 do not.
I think that's a conservative viewpoint, one I think more people should take.
Furthermore, the same goes for HW5 (likely the hardware the actual robotaxi vehicles are getting).
If people are really betting on buying vehicles to use as robotaxis one day, the conservative play is to wait to buy a HW5 car.
I don't believe the semi is 4680 but is 2170.
Sure, just call the mayor of the town and ask for advice. Oops, he IS the mayor. Ok, then a local handyman. oops again, he IS the local handyman. As it happens, far, far back in this thread there is a decent guide to the city here:
Paxson, AK | Things to Do, Recreation, & Travel Information
The small community of Paxson, Alaska is located on the shores of Paxson Lake, near the intersection of the Denali Highway and the Richardson Highway. Find information on things to do, lodging, cabins, fishing, canoeing, campgrounds, and visitor services.
www.travelalaska.com
Just to be clear, because the tourist guide is not too explicit. Audie is pretty much alone in Paxson as far as any installation services or anything else is concerned.
weekend OT
I’m currently with the NE Tesla crew on top of Mt.Washington volunteering to drive runners down. The Wh/mi is amusing. I can see the Jalopnik article now. “Elon’s disaster Cybertruck can only go 70 miles on a charge”
To all my fans here, just because I think Elon’s currently a net negative to Tesla and the SP goes lower before going much higher doesn’t mean I’m not a Tesla fanatic.
I’m currently with the NE Tesla crew on top of Mt.Washington volunteering to drive runners down. The Wh/mi is amusing. I can see the Jalopnik article now. “Elon’s disaster Cybertruck can only go 70 miles on a charge”
To all my fans here, just because I think Elon’s currently a net negative to Tesla and the SP goes lower before going much higher doesn’t mean I’m not a Tesla fanatic.
Just a follow-up after a night's sleep:
Isn't it pretty worthwhile its in public release and its good enough that you can actually debate over whether its impressive or not in the first place?
This feature used to be considered vaporware only 2-3 months ago by the general public. Now, the rhetoric, or initial reaction, turned into "oh no I'm scared about the impressiveness of it" and "oh no I'm scared about the future due to this and what it means".
For me, I had a lot of disengagements while getting used to it. Far less once I got used to trusting it a bit more around traffic-related situations. It's definitely gotten better since April.
cliff harris
Member
Sure, its very reasonable to say the conservative opinion is to ignore FSD until robotaxi is 99.9999% reliable and is operating across an entire nation with no steering wheel or safety driver. Thats very fair, and very true.
But as an investor, waiting for that to be a the case is like watching the iphone announcement, but waiting until ten million are sold before deciding to buy AAPL. Very conservative, and sensible, but not a way to make serious returns.
When I first bought Tesla stock, the company was losing money, had very few superchargers, and only 2 models of car (S/X, roadster was trivial). They were definitely not a conservative investment. But I thought the product was awesome, and more importantly their rate of improvement and growth was amazing.
When I am in the mood to be a conservative investor, I buy global equity ETFs, government bonds, and solid dividend stocks like pharma companies. TSLA is not a stock for the conservative of mind.
I agree with this and do the same to a certain degree. Like everyone else, I hedge and go conservative (applying a FI/RE approach) with broad-based ETF's, bonds, and dividends when I'm trying to take what I need/want off the top. Some people call it "locking in the gains".
Tesla is very much a startup, or set of startups, and it hasn't and doesn't act like a "regular" $500B market cap company in the market. There aren't a lot of $500B companies in the world where you can legit make a case it'll turn $5T in a 5-10 year basis.
nor for the weak of heart
I'd like to think you are right, but we don't know what she means by, "I got to demo Tesla's ride-hailing platform". Her post shows the smartphone app, which we already knew at least had screenshots.
What I'm looking for is evidence that Tesla has started working on the back end fleet management software. You don't necessarily need it to start testing a robotaxi rollout with safety drivers. But as soon as you go driverless, you need a lot more infrastructure.
I think there is time to build it out, but it still sounds to me like they haven't started on it.
This prompted me (much like AI) to look up some info on the Waymo Pole accident in Phx. Reddit people locally seem to be a lot more supportive of Robotaxi than I thought. In my models and projections (not on paper yet), public resistance is (was?) my biggest risk factor. I have to then wonder if Waymo is helping to grease our wheels with basic social conversation. A form of therapy and sharing that is the very essence of the S-curve. So leverage this!
"<Waymo> also has ride-hail autonomous service in Austin, Texas, Los Angeles and San Francisco that provide a combined 50,000 paid rides a week, according to the company"
IF Tesla were to pilot in the US soon, I wonder which city - Austin, Phx, or SF? (Nobody dare say they will deploy everywhere at once, I don't subscribe to that theory and it's not my point here so don't deflect it please.)
Chandler is still my favorite due to weather, road friendliness (wide lanes, dividers, markings...). The public here in the Phoenix area has a well established baseline. From what I "read", it would likely get a resounding applause. Likely, the Chilliest Chilly Chill mode would still impress this group. It would immediately gain attention of the existing Waymo wannabees (where it's not quite in their network but they were waiting for them to expand - like Ahwatukee just recently). That group is likely large while Tesla could do the full surrounding area and leapfrog it overnight... Waymo would not survive, end of story.
Are there enough available Tesla's here now? Easy IMO. I have 2 ready to go and would take that income and roll it into CT. There are only about 600 Waymos in Phx total. Easy to triple that in one day with likely improved night service at first - when we don't need it sitting in the garage, and especially how people really don't want to meet some stranger at 2AM in an Uber.
Elon reiterated the usefulness of the existing Fleet for RT and even how it could be rescued using R/C (although IMO he seemed to make that up live on camera, or very reluctant to share, and I can't decide).
So I may day trade my 1% now and again to keep the spirits up, but my time out of TSLA is getting more limited - by conscious choice. Notice that I picked up more before the Annual Meeting. In the past, I would likely sell a little, then BTD next day. I guess I'm saying it's a ticking time bomb now more than ever, even though we can't see the timer. But you can hear it ticking louder now.
... 
Pretty sure if Tesla is starting this, it is in a geo fenced area they have tested for awhile. They may announce the first area of service on 8/8. And no, this geofenced area will not have 6 lanes worth of 65mph cross traffic for a UPL turn. Apparently some thinks that needs to be solved first or else robotaxi 20 years away.
cliff harris
Member
I definitely agree that they will not suddenly launch a nationwide robotaxi, but I do think a single city (or a handful of cities), launching with safety drivers could be a thing on 8/8th.
The conservative approach would be a single city, with standard model Ys or model Xs, and a safety driver, just to work out the app and the logistics, while work continues on new vehicles, and perfecting FSD. The more adventurous approach would be to do it with prototype robotaxis (with steering wheels & drivers).
Knowing Elon, he will show a prototype RT with and without the wheel, and announce the launch on 8/8th with a few dozen vehicles and and safety drivers.
Actually they already have some experience with this in the las vegas loop. It would be cool if in the future you had to install tesla's robotaxi app to use the las vegas loop. That should drive a bunch of installs.
The conservative approach would be a single city, with standard model Ys or model Xs, and a safety driver, just to work out the app and the logistics, while work continues on new vehicles, and perfecting FSD. The more adventurous approach would be to do it with prototype robotaxis (with steering wheels & drivers).
Knowing Elon, he will show a prototype RT with and without the wheel, and announce the launch on 8/8th with a few dozen vehicles and and safety drivers.
Actually they already have some experience with this in the las vegas loop. It would be cool if in the future you had to install tesla's robotaxi app to use the las vegas loop. That should drive a bunch of installs
.
Or sooner, and then reveal the RT vehicle on 8/8 - it's anyone's guess. This is actually possible. It is non-zero.
Paracelsus
Active Member
“This Delaware judge probably did more damage to Tesla than short sellers did to Tesla between 2013 and 2019, but only in the short run. In the long run, this battle has probably strengthened Tesla retail shareholders by forcing us to wake up and fight. This will be useful in future shareholder battles around AI, even in Texas.At least short sellers had to cover and give us their money during short squeezes. This judge squandered a few valuable years by creating a ton of uncertainty about the future of Tesla, keeping the company from moving forward with a new compensation plan for Elon. It was a massive distraction and waste of time for which we will never be compensated. The best thing to come out of this situation was the uniting of Tesla shareholders not through our holding of the stock (with the short seller situation), but through our votes.”
Similar threads
- Locked
- Locked
- Locked
