voip-ninja
Give me some sugar baby
I could be mistaken but I have a feeling that there are many of the players involved in lending money to Tesla who were also lending money to Amazon,
-
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
I could be mistaken but I have a feeling that there are many of the players involved in lending money to Tesla who were also lending money to Amazon,
True, but if you’re a traditional big investment firm, it seems a big change of direction. Before, they and many others, were thinking high margins and Apple type skimming of the most lucrative cream of the market.
My guess is that their biggest concern is what will happen to margins, and will the price drops be enough to totally alleviate any demand concerns.
Of the 20+ people to whom I’ve shown my Model 3, only one seemed likely to buy. Although everyone thought it was great and was wowed, the EV hurdle is just too great. They are pretty much fixed into the 5 minute fill up that gas vehicles provide and the familiarity. I always show the Supercharger map and explain, but almost no one has been aware of the range and the Supercharger network and just unable to make that large of a leap.
Disclaimer; I’m talking about the southern U.S.
I agree, the obvious thing to do is to identify weaknesses in the algorithm, likely through testing or disengagement reports, and then set the global fleet to collect full data of similar situations. This data will be labelled and added to the neural net. Potentially similar situations currently in the neural net will be reviewed and replaced with better examples. After the data set has been refined, then the whole neural net will be retrained. I don't know how many days of training the neural net takes, but multiple are likely to be in training at the same time. The new neural nets will then be tested, potentially in real world, but also in simulation by feeding the full neural net data already collected from the fleet for different driving situations and then quantifying the neural nets response. Just speculation, but this is roughly how I would do it.
Fact Checking
Well-Known Member
Tesla's NNs are effectively autogenerated from the labeled training data during each re-training run, which is done on a huge cluster. (I suspect they are not using deep learning techniques yet.)
I.e. the old NNs are typically discarded for anything but debug/comparison purposes.
I don't know how long Tesla's training run takes, but I'd expect them to have enough hardware for this to take less than a day, with automatic daily results generated.
Mistakes in the training dataset, bugs or inefficiencies in the training algorithms and labeling processes of course show up in the NN - I think Tesla's AI group uses a significant portion of their time to find such meta problems.
But "mistakes" in one NM don't get carried over into new re-trained networks, unless you are doing something fancy or are cutting corners due to lack of training HW.
To be fair, I don't think there's ever a good timing to cut the price of a car from €176k to €118k - I bought my MX in September 2017 and I feel quite off about it, so those that just took delivery...
Totally agree. My conspiracy theory: Looks like a classic Larry E. 'Gorilla Move'. Go Big or Go Home. If it succeeds Tesla becomes the Gorilla and the mission Statement is nearer completion.......If it fails, the SP drops to a point where going private is 'cheap' and Tesla moves on without retail investors to complete the mission.
I prefer the former outcome. I do believe EM cares about the retail investor but he cares more about mission over money (his and ours)
Do you think the following are insufficient to explain this? Because I think the following are sufficient to explain it:
1 -- tax credit pullforward of demand (well documented phenomenon) means they were running above steady-state US demand in Q4, and are runinng below steady-state US demand in Q1
2 -- The $35K model was a longstanding promise and the company's credibility depended upon actually getting it out the door, preferably before the $3750 tax credit goes away.
Yes, I believe Q1 is essentially a blip due to filling the pipeline *and* the tax credit pullforward effect. However, I think Musk finally took a good look at the stores and said "We are spending how much on these and how much value are they adding" and the answers were "a lot" and "very little". I mean, I may be biased because I've actually been saying that for years.
Tesla definitely doesn't have as much cash as they would like; there's no cash for starting a European Gigafactory or a Model Y Factory or a Semi Factory at the moment. They should have enough to expand service (I do know this is happening; and the wait times for spare parts have dropped by fixing the logistics) and Superchargers.
From your link
Current is the number of electrons passing a point in a given amount of time.
Drift velocity is analogous to a river: A wide river has a lower flow speed for the same volume than a narrow river does. A fast moving river has more water molecules passing at a time. However, water is not-compressible whereas a wire has way more electrons than the number moving due to current.
Perhaps a better example is gas through an orifice. More flow is more molecules, due to higher density, but they are still moving at the speed of sound (until the chamber pressure on Merlin or Raptor get high enough, then it goes supersonic and non-compressible and starts increasing the exhaust velocity, thus imparting more momentum onto the Falcon/ Starship, but I digress).
More electrons does not mean faster electrons. Or I'm wrong, physics was a while ago
How then do I read here Drift velocity - Wikipedia"
Drift velocity is proportional to current.
Also, look at Wikipedia's example:
u = I/(nAq)
with:
u: avg velocity of electrons (m/s)
I: current, (ampere, coulomb per sec)
n: avg electrons per m^3
A: wire cross section (m^2)
q: charge of an electron (-1.6e-19 coulomb)
So:
- more electrons does (indeed) not mean faster electrons, but more current does, which was the point I was making.
- the number of electrons in the wire is fixed (not-compressible, like mongo's river)
TL;DR: I stand by my statement
Jesus, that's terrible. Massive overdeployment of stores.
Frankly the stores needed to be cut back period. It's a no brainer and I wonder why they had so many to start with. I think every claim about why they're cutting them are excuses to make the fired employees feel better.
One per metropolitan area is more than enough.
anthonyj
Stonks
Lmfao wtf! I don’t even know what to believe anymore. This is the most confusing quarter in Tesla’s history
One thing I am gleaning about the stock market in my short term experience so far of investment & trading is that the market often reacts far more slowly to news than I would have expected. It’s as if they wait for a stock analyst’s report to decide on action, even though they may have been aware of the news for a day or two
Or am I misguided?
Or am I misguided?
I still don’t understand the price breakdown unless you invent
A bearish narrative to justify it.
Q1 loss is understood, one time charges too, the rest
Is a bearish concoction and fear.
A bearish narrative to justify it.
Q1 loss is understood, one time charges too, the rest
Is a bearish concoction and fear.
anthonyj
Stonks
Tutes and shorts dumping, then stop losses get hit, then longs get exhausted and sell low to the tutes who drive it back up
I *think* you’re referring more to incorrect data polluting the dataset, not actually retraining a network. Other than very small scale refinements, you don’t typically retrain an already trained network. You’d just start over with randomized weights and train from scratch(usually training many different versions with different hyper parameters at once and choosing the best one).
Well there's a bold statement that would surely be best discussed with some fine craft IPA's
The problem here is that SGD is going to find the quickest way to solve the problem it’s given. For any kind of simulation, that’ll likely mean overfitting to the particular parameters of the simulation, with any deviation from those causing it to completely break down.
For most retail investors, you're probably right - we're an exception here as we live and breathe Tesla.
Other other end of things are the tradingbots that almost instantly buy and sell based on headlines - given the amount of BS FUD around these days, that has no doubt initiated some sales.
OT
Picture a river feeding a lake feeding a river. A single water molecule will take years to get from river to river (drift velocity with large crosssection), but the outlet stream flows at the same rate whether the lake is there or not.
Increasing current reduces the time an individual molecule spend in the lake, but also the speed, because water in not compressible.
The number of electronics is fixed, but they do not all need to move to create flow (in a normal sized wire). So the velocity of the electrons is independent of the current/ number of them moving.
Being proportional to current does not impact the speed of the electrons that cross the measurement point. For a fixed resistance, voltage is also proportional to current, but also does not impact speed...How then do I read here Drift velocity - Wikipedia"
Drift velocity is proportional to current.Also, look at Wikipedia's example:
u = I/(nAq)
with:
u: avg velocity of electrons (m/s)
I: current, (ampere, coulomb per sec)
n: avg electrons per m^3
A: wire cross section (m^2)
q: charge of an electron (-1.6e-19 coulomb)
So:
- more electrons does (indeed) not mean faster electrons, but more current does, which was the point I was making.
- the number of electrons in the wire is fixed (not-compressible, like mongo's river)
TL;DR: I stand by my statement
Picture a river feeding a lake feeding a river. A single water molecule will take years to get from river to river (drift velocity with large crosssection), but the outlet stream flows at the same rate whether the lake is there or not.
Increasing current reduces the time an individual molecule spend in the lake, but also the speed, because water in not compressible.
The number of electronics is fixed, but they do not all need to move to create flow (in a normal sized wire). So the velocity of the electrons is independent of the current/ number of them moving.
Anyone else having trouble figuring out exactly what the context is of Elon’s reply? I’m not sure what gap he’s talking about...
Similar threads
- Locked
- Locked
