Tesla Network prerequisites

[Buckminster]

Complete

1. Decent auto
2. Service network incl. mobile
3. FSD h/w
4. Sustainable funding / business model
5. Navigation

In progress

1. FSD s/w
2. App with payment for customers
3. App to loan auto to TN
4. Insurance
5. FSD trainer - Project Dojo?
6. Ability to remove steering wheel

Progress unclear

1. Snake charger rollout
2. TN auto s/w - cleaning / charging regime etc.
3. Auto cleaning network
4. Payment of tolls etc.
5. Cutover strategy for L3 to L4 to L5.

 

[Buckminster]

Yes, Karpathy specifically described Project Dojo as Tesla's own neural network training chip: "doing the same order of magnitude speed improvement on the training side, as we did on the inference side with HW3, at lower costs".

Internally they jokingly refer to Project Dojo as "Vacation Mode": it will allow them self-improving neural networks where training is fully automated and iterating so fast and so reliably that the AI team can go on vacation. (While the labeling team still labels images of course.)

Other interesting tidbits in Karpathy's presentation held at around October 10 (a month ago):

• The HW3 board is cheaper to make than the HW2 board,
• Smart Summon had ~700,000 uses two weeks after release.
• He described the size of the HW2+ AutoPilot fleet as "almost 750,000 cars".
• Nice top down "virtual LIDAR" data representation + videos in the presentation, "stitched together" real-time from 8 cameras by the neural networks themselves in the car, not by C++ code.
• Labeling includes traffic signs and road signs already.

Mood: he seems pretty content with AutoPilot progress.

 

[Buckminster]

Volkswagen unveils R2-D2-like charging robot that autonomously steers and refills electric vehicles | Daily Mail Online

 

[heltok]

Volkswagen unveils R2-D2-like charging robot that autonomously steers and refills electric vehicles | Daily Mail Online

Doubt it will be something like this. Using a middle step of a portable battery adds energy losses and infrastructure costs. Makes much more sense to use software to have the car drive into position and charge direct from the grid.

 

[Buckminster]

https://twitter.com/x/status/1618630423038525441

Good question. Tesla could state that H/W3 is only for private use.

 

[heltok]

The hardware is clearly capable. Already with today's software it's safer than humans and can drive [not always reliably] in every situation. I see now reason for why better software should not in theory be able make it capable of robotaxi. Will lawmakers approve it? A small possibility that the answer will be no in some jurisdictions if HW4 is even safer.

 

[Tam]

...Will lawmakers approve it?...

The question is not about lawmakers but about Tesla.

L3 Mercedes has been approved in Germany and Nevada, USA.

It told California DMV that it is L2 and not above so no applications needed.

Tesla has not submitted any applications for L3 and above approval. Because it doesn't apply, it blames the lawmakers.

A small possibility that the answer will be no in some jurisdictions if HW4 is even safer.

HW4 needs to pass mannequin tests from Dan O'Dowd first.

 

[Buckminster]

Tesla tries to get owners to give up 'unlimited free Supercharging for life'
This is bullish, suggests Tesla is actively prepping to ensure there aren't too many gold mines out there.

 

[Buckminster]

Fast forward 2 years: Prices could be:

• EAP $6k
• FSD for private use $25k
• Tesla Networked Robotaxi $100k (assuming owner receives 70% of revenue and Tesla don't own the fleet which is my preference)

EAP becomes the gateway drug at bargain basement prices.

Makes sense that Elon would continue to use agile methodology. Rather than unleash FSD to everyone everywhere - start on highways only. However, everything else points to a big bang.

 

[Buckminster]

https://twitter.com/x/status/1657518711291494400

 

[Buckminster]

https://twitter.com/x/status/1657911461543378952

 

[Buckminster]

Job openings for 2 Vehicle Operators in Tempe Az (Phx) - 2 shifts so round the clock'ish.
I smell Robo-taxi testing here. Are other cities hiring, anyone?

Key Tasks:

• Operate a vehicle in a designated area for data collection
• Start/Stop recording devices and do minor equipment/software debugging when necessary. Analyze/report data collected during shift
• Write daily drive reports detailing observations and issues
• more...

https://www.indeed.com/viewjob?jk=cbbafb2965676004&q=tesla&l=Phoenix%2C+AZ&tk=1h0m4lk02gkdr805&from=ja&alid=5aff71fd2e7c228b5de5193d&utm_campaign=job_alerts&utm_medium=email&utm_source=jobseeker_emails&rgtk=1h0m4lk02gkdr805

It’s happening!

 

[Buckminster]

Elon has spoken several times regarding the value of FSD and robotaxis to the future value of TSLA. This includes the predicted 5x utility if the car is self driving and out earning money. One unknown I would like to understand better is how we expect the liability to be managed and is the liability greater or less overall in cost. This is particularly relevant as liability will shift to Tesla in some form. Consider two scenarios:

1. Current situation where human driver is at fault and severely injures passenger in his car and in another car. His family member riding in the car probably does not sue the driver but insurance helps with medical bills. Family in other car will sue and there is individual insurance plan for that. Driver is happy he was driving a Tesla.
2. Self driving car causes same wreck with same injuries. Overall risk of this wreck occurring is less but now everyone is suing. There are insurance settlements and big lawsuits against one of the largest richest companies in the world (Tesla). The individual driver is no longer at fault and he is suing Tesla for making an unsafe product. His injured family member is suing Tesla for significantly more than the medical bills. Class action lawsuits crop up and lawyers specialize in suing Tesla for routine car wrecks. It is easy to imagine that any benefit in risk reduction does not save money overall if each wreck is more expensive. How should we design insurance products (and regulation) in this case?

 

[Knightshade]

It’s happening!

It's really not.

Same job they have had people doing for years- as pointed out in the thread you copied the post you're replying to from.

 

[Buckminster]

https://twitter.com/x/status/1666341081112367110

 

[Buckminster]

The new UNECE regs go much further than this and essentially would allow the current FSDBeta on our roads, not just motorways.

The main differences being that the system has to be approved first, and pass many tests before its deployment, rather than the USA/Canada approach where Beta/testing status is accepted in the hands of customers whilst it is developed.

There are also lots of things that will need to be incorporated into the UNECE version of FSDBeta, dealing with how it copes with different situations, be it reacting to speed limits, distances needed for automated lane changes etc.

this link gives lots of in depth info on the proposed new testing proposals:

https://unece.org/sites/default/files/2023-05/GRVA-16-39e.pdf

 

[Buckminster]

https://twitter.com/x/status/1667776312302505988

Cleaning required

 

[Knightshade]

Same stuff happens in normal cabs and drivered Ubers/Lyfts all the time... Unless it's super obvious to the driver (like someone pukes or they hear a shattering glass thing) they usually only clean out the car at start/end of shift

It does reinforce the idea the eventual-ready-for-public-use Tesla robot would be ideal to station at each supercharger both to plug/unplug the robotaxis AND to clean out the interiors while charging though.

 

[Buckminster]

I think the best way to do this is add a heat map of where successful drops were completed as instructed:

https://twitter.com/x/status/1668167793097965570

Snapchat:

 

[Buckminster]

https://twitter.com/x/status/1686816248703934464