-
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
Doggydogworld
Active Member
Regulations make V2G very difficult, but as I indicated that is a long way off (if ever). Regulations do not prevent the other stuff. Super-simple example: have EV/chargers respond to grid signals and reduce or increase their charge rate accordingly. Dispatchable demand is a HUGE benefit to the grid and with enough EVs can enable much higher renewable penetration.
Thanks for the link. Showed it to my husband and he's watching it now. Jeff is always an interesting listen. BTW interesting Twitter account.
Regarding AI and FSD I think he has absolutely no idea though... like most analysts.
Let's summarise his points and then break them down to show how wrong he is:
> He starts with saying AI is not artificial intelligence, it is artificial stupidity, and his basic premise is that AI will not be able to handle 5% of driving situations. He says for just a 20 min drive, the system would need a human intervention.
Reminiscent of Jack Ma's argument style on AI. It's pretty easy to get to 99.99 level of reliability. I don't remember the figures of what would constitute acceptable FSD - maybe 99.9999 or 99.99999. It gets exponentially harder for every 9. The first 2 decimal points will happen very rapidly. Though I think it's more useful to compare the system to accidents from a human on a miles driven basis. Either way his 5%/20 min drive shows very poor understanding of the numbers.
> Cannot call AI systems intelligent because they are just follow a set of rules. Driving cannot be solved by a computer. Driving requires un-qualifiable philosophic tools and an AI cannot be creative to solve situations it has not seen before.
Neural networks are not really rule systems. They are designed to mimic the human brain. He has a somewhat passing understanding of strong AI vs weak AI, and correctly infers FSD is a form of weak AI, such as a chess engine or Siri/Alexa. But I doubt he appreciates what weak AI can solve. I used to watch Magnus Carlsen stream and he gave his thoughts on chess engines. He described Stockfish (brute force minimax) as an idiot savant which looks like it has no idea what it is doing, and yet luckily still manages to beat you every single game. AlphaZero (neural network) on the other hand played with sacrifices, and without even knowing whether the sacrifice would work. It played positionally and strategically, rather than just tactically like Stockfish. AZ felt like playing a human. MC and others said playing AZ very much felt like there was a human playing. Pierre, on the other hand, wants to take a concrete position on the differences between human and machine intelligence... a position which is neither proven, likely or relevant to the how far weak AI can get in FSD.
> For example:
* Stuck behind truck making a delivery
* Dropping off someone in forbidden place
* No room in a street
* No clear cut priority
What reason does he think these common situations cannot be solved by an AI? You will have to get a lot more creative that these examples to reach some of those 99.99999 level scenarios for which Tesla will not have not trained their AI for. His general idea of the AI only being able to handle what it has encountered before is only somewhat accurate. The AI will be able to handle situations where it has encountered similar but not identical scenarios previously... that is the whole point of training data after all. It is similar but not identical. To expand on Pierre's scenarios, maybe the system had never been trained on being stuck behind a truck making a delivery, but had been stuck behind a car dropping someone off, and it would correctly apply the same behaviour. In this regard there is some flexibility, creativity to solving new problems. Also, Pierre is overestimating the intelligence required to solve what are essential many, many, simple problems, which in the world of deep learning just comes down to data + training + time. And yet he still thinks Waymo is ahead. If there is one thing he needs to learn is the importance of data/data extraction/data quality vs simulated.
> Might happen in .01% of situations, becoming less and less but still happens all the time.
He seems to understand there is a direction of improvement (contradicting his earlier 5%), yet fails to see that at some point it becomes viable/preferable to human drivers.
____________
Overall, I liked some of his ideas, but I can't help but lose all respect for an analyst who gets AI/FSD so wrong.
Last edited:
davecolene0606
Member
briefly,
single crystal electrode material, the correct electrolyte additives (more than 2 is better) makes massive improvement to cycles and storage life of cells.
I think we see this and Maxwell dry electrolyte process coupled with precision injection of into a massively sped up production rate that leads to the 2 terrawatt path.
Fire Away!
(It’s the batteries, Stupid!)
Unfortunately, tests don’t stop the spread of diseases. And there’s not enough test kits to be testing broad populations.
Interestingly, Dahn only talked about NMC chemistries. He mentioned NCA only in passing. His expertise was in NMC chemistries, so perhaps he is only talking about his immediate interest. But I wonder what if anything this lack of mention signifies.
Artful Dodger
"Neko no me"
NASDAQ data shows that 301,654 shares traded for $674.00 at 7:07 pm. BTW, that's 45.95% of all 656,440 shares that traded during the After-hrs session on Fri, Feb 28, 2020, a huge proportion of trading in a single minute.
Then at 7:08 pm, just 36 shares traded at $673.49, so the SP actually went down after the huge trade made in the previous minute, and volume returned to a trickle.
Indeed, if we look at other trades before and after 7:07 pm we see there is no reaction or change in Share Price and/or Volume due to that outsized trade:
This implies strongly that this was a prearranged trade which did not go through the order book, else it would have had substantial effect upon the SP. It did not. Perhaps @Cherry Wine has some insight into the Order Book at the time of this large trade?
The SEC needs to pull the logs for Trades done at 7:07 pm on Fri, Feb 28, 2020 and trace the parties involved. If one of those parties was also involved with naked short selling before 4:00 pm Fri Feb 28, 2002, daily totals for which was reported by FINRA on the same day, that is a violation of the Market Makers Exemption on the prohibition against naked short selling.
Proprietary, for-profit naked short-selling is NOT legal, even by Market-Makers, and on Fri, Feb 28, 2020 this practice significantly distorted the market for TSLA and damaged shareholders. This must be investigated to restore confidence in the market.
#SEC @Hock1 @Fact Checking
Regards,
Lodger
Last edited:
Tests are the primary method of stopping the disease. How can you diagnose, appropriately quarantine someone and trace their connections if you don’t test them?
Yes, I also noticed that earlier...
I've had a long standing hunch the Semi might be NMC,
The Semi needs a chemistry which cycles daily for a long time,,, no babying the pack on a work truck ... it needs to survive hard work.
It is never going to be perfect, but whenever you can trace, test and quarantine people you at least slow the the spread of the disease..
The aim of these measures is to buy time while we work on a vaccine and reduce the impact..
At this stage of the progression it is still well worth doing this, China has apparently had a measure of success.
Pras
Member
A few thoughts.
I am guessing that Tesla knew how bad this coronavirus *sugar* was and decided to raise money.
China has already shown the world how to deal with the situation. So the world can now deal with the situation more effectively. Mean much smaller # of cases.
china Q1 GDP will be negative and US GDP will be negative potentially. So we may be in recession already.
the stock market may fall another 10% or even 20%. Time to be defensive (although I am 90% invested so I need to worry on this one, Monday musings).
Tesla story doesn’t change and long term it still remains attractive no matter what happens to market.
The market is split in two categories, stocks that have overperformed (think Apple) and stocks that have underperformed (think material). I think the market will rebalance.
....just my thoughts
I am guessing that Tesla knew how bad this coronavirus *sugar* was and decided to raise money.
China has already shown the world how to deal with the situation. So the world can now deal with the situation more effectively. Mean much smaller # of cases.
china Q1 GDP will be negative and US GDP will be negative potentially. So we may be in recession already.
the stock market may fall another 10% or even 20%. Time to be defensive (although I am 90% invested so I need to worry on this one, Monday musings).
Tesla story doesn’t change and long term it still remains attractive no matter what happens to market.
The market is split in two categories, stocks that have overperformed (think Apple) and stocks that have underperformed (think material). I think the market will rebalance.
....just my thoughts
Did we watch the same video? He mentioned that he expects energy storage to be a $30 billion business in 2025.
Artful Dodger
"Neko no me"
Nice! Thanks for the heads-up. This will promote the notion that "EVs are FAST" within the hardcore, gearhead, gasohol crowd.
Here's the highlight from the race on Youtube: "Race Highlights - 2020 Marrakesh E- Prix" (7:00 min)
Cheers!
Last edited:
Um. No.
Recession:
“It is typically recognized after two consecutive quarters of economic decline...”
Recession Definition
"Tests are the primary method of stopping the disease." That's one of the most ridiculous things I've heard of when it comes to medicine/science. So, you've found the source of a community acquired case. Then what? Do you think the transmission is halted during the investigation?
By the time a sick person voluntarily seeks medical care, they’ve already been shedding the virus everywhere. Subclinical infections are never accounted for. How do you think community acquired cases happen?
Once you have an unknown community acquired case, containment has to be thought much broadly. One of the bigger reasons why China is able to bring down rate of new cases is because China's authoritarian style of government can force broad lockdown quarantine. That level of preventative measure most likely won't happen in countries such as the US, where civil rights is a huge deal. (And especially in an election year.) Advising self-quarantine helps but is not the best.
Last edited:
Dahn made it pretty clear that a lot of his research was about fundamentals and the choice of his materials was often more about convenience rather than "this is what we plan to use". For example, he was using 532 pouch cells to test electrolyte additives, not because Tesla plans to use 532 pouch cells, but because he could buy them by the thousands from China without an electrolyte, then inject thousands of different electrolyte combinations and measure their performance.
NMC and NCA are quite similar chemistries, and I'm not sure how important it is to make a big deal of the difference between one and another. Both manganese and alumium service the same oxygen-scavenging role.
Similar threads
- Locked
- Locked
- Locked
