1x Tsla Discussion Threads

[dominiccummings] 17/10/2024 18:42

So we adapt

[hosede] 17/10/2024 13:55

Dominic
But while history continues to repeat itself every (say) 8 generation, Climate change of this order is something totally new to Homo Sapiens

[dominiccummings] 17/10/2024 12:31

Climate change predates all humanity by aeons,
A broken financial system is rebuilt in a few years at worst.
Maybe we deserve to be poorer?

[hosede] 17/10/2024 11:16

OK - I accept there are exceptions, but when the debt bomb goes off there will be little left - The oil-rich gulf states and maybe Switzerland and Norway. For the rest of us Argentina here we come.
Oh and then there's the little matter of climate change !!

[dominiccummings] 17/10/2024 10:08

Microsoft have successfully moved their client base onto 'rental'. So their revenues are never going to be lower year on year whilst customers are locked in by organising their life and/or business utilising the Microsoft world of products.
Similarly, Google are nowadays locked in the recurring advrtising revenues such that only a mighty change in habits would weaken them.
Just two examples where massive PE's ARE justified.

[hosede] 16/10/2024 14:27

Lets see what Microsofts share price is in a years time . Well under 100 I suspect

[cfb2] 16/10/2024 14:08

In 2005 Microsoft's P/E ratio was 32 with a share price of $28. Current share price is $410.

Writing NEVER in capitals doesn't make it a fact.

[simon gordon] 16/10/2024 08:13

FT - 16/10/24

Investors should be wary of Tesla’s robotaxi hype

The pay-off from fully self-driving cars is likely to take much longer than expected

Elon Musk has said Tesla’s valuation could go as high as $5tn as it pivots to autonomous driving and artificial intelligence — seven times its value today. The company’s shares are down more than 10 per cent in the past week after its robotaxi event disappointed investors. But even if Tesla had delivered on the hype, should robotaxis be that big a deal for investors today?

Despite recent declines, Tesla shares trade at 80 times forward earnings — a steep premium to global peers. That reflects hopes that its heavy AI spending will make it a key beneficiary of a shift to fully self-driving cars and that the transformation will arrive fairly quickly. But the pay-off is likely to take much longer than expected, with the beneficiaries spread out over a wide range of industries and companies.

Robotaxis have the potential to be transformative — should they become mainstream. For that to happen, mass production and commercialisation must come first, which will mean many more years of development and testing. It could also require a whole new approach to car manufacturing.

Musk did not provide details on the technology behind the robotaxis or how he would cut the cost of self-driving cars. The latter is key: carmakers need to offer a compelling cost advantage for customers and robotaxi operators to justify the cost per mile of an unproven technology.

Critics have been sceptical about how soon Tesla’s promised self-driving taxi, priced under $30,000 and less than its cheapest Model 3, will hit markets. It has taken Chinese tech group Baidu, which operates fully driverless cabs in China, seven years of development, extensive road testing and six generations of robotaxis to get its latest model down to around $28,500.

For now, the economic argument for robotaxis remains weak. There is a limit to how low costs can go because of the high prices of the cameras, radars, chips and lidar sensors that are in them. Traditional automakers would need to invest heavily in software development to catch up.

Mass adoption is also a long way off. Development and testing will take time. Addressing safety concerns will require even more. Operating driverless cars on highways and in urban environments, with pedestrians, animals and bicycles and other hard-to-predict obstacles, comes with a high safety bar.

Meanwhile, fares for robotaxi rides are typically lower than for traditional cabs. Robotaxi networks, which also need to provide cleaning and remote assistance, are not yet profitable. Baidu’s operation, which has completed 7mn rides as of July, has yet to break even, despite using its own robotaxis at rock-bottom prices. Waymo, which runs more than 100,000 paid robotaxi rides per week in the US, is also lossmaking. GM’s autonomous vehicle business, Cruise, booked a $5.8bn loss from 2021 to 2023.

With economic returns sufficiently uncertain, makers of self-driving cars will end up competing on cost for mass-market adoption. That could require a wholly different approach to making electric vehicles. Instead of each carmaker building their own from scratch, as Tesla is doing, outsourcing manufacturing will start making more sense. Standardising as much hardware production as possible and focusing on differentiation through software would squeeze costs down faster.

Waymo, the driverless car company owned by Google’s parent company Alphabet, is an example of the right path. It entered a partnership with Hyundai Motor this month to integrate Waymo’s fully autonomous technology into Hyundai’s EVs. Elsewhere, Foxconn, the world’s largest contract assembler for Apple, has set up a platform to make tailor-made EVs on contract, offering autonomous driving modifications.

Foxconn already supplies parts to EV makers including Tesla and makes EV car parts such as camera modules, electronic control units and sensors at high volumes. Its scale means stable, low-cost parts procurement for clients. Such EV outsourcing is expected to reach $144bn, or 3.2mn EVs, by 2030, according to Goldman Sachs. That also means new growth for suppliers of autonomous driving-related components, for which the total market is expected to hit as high as $80bn by 2030, according to McKinsey.

Despite the impressive speed at which EVs and self-driving functions have progressed, great areas of the nascent technology, especially around safety and regulation, remain sketchy. Given all the present uncertainties, the adoption of robotaxis in everyday life is likely to remain slow and cautious. That path is one that investors should follow.

[hosede] 14/10/2024 10:26

Yes
I can't see any Carbon based life system being able to travel so far. Creatures hibernate for a few months, but hundreds or even thousands of years? And what would wake them.
Meanwhile a minor politician somewhere said
" the end of the world is charging towards us"
and nobody cares a jot!! Climate change will accelerate and in say 200 years we could be at the previously calculated record 90 Deg F as opposed to 60 now.
Last time we were here - at the end of the Permian - the seas heated and 96% of life died including - I guess - Phytoplancton that provide 50% of our Oxygen. Burning forests would destroy the rest.
We might be smart enough for a small residue of Homo Sapiens to survive - but very little else.

[cfb2] 14/10/2024 09:48

Yes, the Fermi paradox is asking why haven't we detected intelligent aliens given there are over 2,000,000,000,000 galaxies and on average a galaxy contains over 100,000,000 stars and most of those stars have planets orbiting around them.

There is something called the Drake's equation, which attempts to use the above constants and a number of others to estimate the number of intelligent alien civilisations we would expect to exist in the universe at any point in time.

One of the constants is how long a civilisation exists with the expertise and will to emit detectable transmissions. Chances are there should be quite a number of intelligent aliens around at any time but it should be steadily increasing unless the civilisations become extinct.

Whilst the speed of light is not part of the equation, the fact that our current knowledge of physics says communication can not occur faster than the speed of light meaning it would take many years, perhaps thousands of years, for an alien civilisation to detect us and send something back.

The great filter are challenges that an intelligent civilisation must overcome to continue existing, such as climate change making their planet uninhabitable. Another example would be an extinction level asteroid hitting the planet before they have the knowledge to divert or destroy it. Elon Musk's goal is to make our civilisation multi-planetary and so side step a number of possible great filters.

[hpcg] 14/10/2024 08:25

There is no Fermi Paradox. It's a bunch of made up numbers multiplied together to yield another made up number. Even then it requires some impossible assumptions like faster than light travel.

[hosede] 13/10/2024 22:15

Cfb
You mentioned the Fermi Paradox earlier. I assume you mean that any society as advanced as ours is almost certain to destroy itself. Sad but true I fear.
What do we do with our incredible skill/knowledge?
Build more and more powerful weapons

[cfb2] 13/10/2024 21:41

hosede: Our opinions differ because I see Tesla's robotaxis and EV cars moving the world in a positive direction and rather than being irrelevant they could provide low cost and convenient transport for everyone whilst reducing environmental damage.

"The optimist thinks this is the best of all possible worlds. The pessimist fears it is true."

[dominiccummings] 13/10/2024 18:48

Cheery chap, hosede.

[hosede] 13/10/2024 18:45

CFB
All this stuff about Robotaxis - and flying across the Atlantic in an hour etc. etc. are (IMO)quite irrelevent. The coming depression (K winter) exacerbated by climate change means that we shall be looking in the main at survival. - and even that could by the end of the century - be in the balance.

[cfb2] 12/10/2024 22:00

Simon: An interesting article and largely correct except the software architecture of Tesla's system is wrong. I don't know enough about Waymo's software but there is a lot being hidden in the "Foundation model for perception" box in the claim #1 diagram, which is where they are supposedly using their neural net.

Claim #4: Waymo control centre does get involved in assisting cars on the road. There are 2 operators per Waymo car and they have the ability to send commands and respond to queries sent by the car.

The article is correct that Tesla will need to have a data centre for their robotaxi fleet. If for no other reason than to correct mapping faults detected by cars and relay them onto other vehicle as well as ensuring the robotaxi are evenly distributed over their operating area. Whether an operator will need to intervene with the cars will depend on how the software improves; my guess is yes because Musk will want to start the robotaxi service before the software is ready.

I see Tesla's biggest advantages as:
1. The amount of video data Tesla have captured by the fleet
2. The compute power to crunch the data in a sensible time frame (~2 weeks).
3. Their FSD supervised model has customers testing the software for free and the number of cars means that edge conditions are found faster.
4. The cost of Tesla's cars and FSD hardware means they will beat others on the cost per mile.
5. Tesla's technology has good portability so they'll be able to licence it to other manufacturers. The Cybertruck is a demonstration of how quickly they managed to do this with a completely different vehicle.
6. Tesla has over 4m cars on the road capable of operating as robotaxis, whereas Waymo has around 2,000 cars.
7. Tesla have over 2 billion miles of safety data ready for regulator approval.

[simon gordon] 12/10/2024 20:16

cfb2,

As far as I can see, that is his real name:



Here's his piece on Waymo:

Elon Musk wants to dominate robotaxis—first he needs to catch up to Waymo

I talked to Dmitri Dolgov, co-CEO of Waymo.

[cfb2] 12/10/2024 19:50

Ahh, this is not Sir Timothy Berners Lee, professor of MIT, but someone who writes his name that way to cause confusion.

[simon gordon] 12/10/2024 19:16

Hi cfb2,

This is the guy:

[cfb2] 12/10/2024 18:20

Interesting quote Simon. I couldn't find it via a search engine, can you provide a link please?

Lidar will always be more expensive that a photosensitive sensor because there is a lot more going on. Current best (10K+ volume) price is around $500 per sensor compared to less than $0.40 for a quality photosensitive sensor. Lidar manufacturers have been claiming they will eventually get the price down to $250. Hasn't happened yet.

Tesla have been saying that cars can drive with optical sensors because humans do it. They also say that a combination of sensors causes problems because at some point a decision needs to be made as to which sensor input you should believe when they differ.

[simon gordon] 12/10/2024 13:24

"Tesla fans put a lot of weight on the premise that lidar is going to be expensive forever, and therefore Waymo requires "niche hardware" that can never reach a mass market. But one of the most predictable things in the universe is that electronics gets cheaper with scale."

-Timothy B. Lee, 11/10/24.

[cfb2] 12/10/2024 10:30

hosede: Last time you mentioned putting tape over a car's sensor I pointed out this was vandalism and the car would have cameras to provide evidence.

Lidar uses laser pulsed light similar to radar so will just as easily be disabled with tape; the light catching a mirror would blind it, as would another coherent light source shined at it such as a laser pointer.

The biggest advance of Lidar is that it sees in distance. You get back a high resolution picture of the distance of every pixel. Tesla's vision works out the distances by analysing the image, the same way as our eyes work when driving.