Ark Invest CEO spanks the lizard (aka Cory Johnson). Delicious...
Ark Invest CEO Explains Why She's Excited About Tesla
Ark Invest CEO Explains Why She's Excited About Tesla
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
- Status
On the "supercomputer server". I'm sure Tesla has this kinda thing in their RND team, perhaps several. Those are used to train their algorithm. The algorithm then OTA to cars. For the computing ones on each car, no server would be needed.
I know Cathy and Tasha (she mentions in the piece)...I like the way they think...Tasha in the white paper (which I posted in the Tesla Mobility thread) contends that the current valuation for Autonomous driving today, based in future growth, is greater than the current valuation of all OEMs combined (Ford, Volkswagen, Toyota) at a little over $1Trillion, and will dwarf them 10-1 in a little over a decade. There is a lot of synergy between the ARKK white papers and the Bloomberg New Energy Finance projections for "peak oil".
Only good thing about Corey was that window pane jacket.
everman
Member
Dang it Curt! What a trip down memory lane! I just got caught up in that thread and read it from $38 through the run to $60 and the earnings +30% pop to $80+. What amazing days! Hope we have another run on the books for 2017, year of the Model 3!!
My first job was $1/hour working as a soils mechanic/draftsperson with a civil engineering firm while in high school. Minimum wage then. Saved enough to pay for about one-third of my tuition at MIT then. Along with other student leaders we complained one year the 'toot' was raising tuition from $900 a year to $1,100 while Harvard wasn't raising its tuition. We asked, "what about Mr. Sloan?" a great supporter of MIT and an 1895 graduate. Jay Stratton, who was acting president, said "Harvard has many Mr. Sloans." Q.E.D.
Times change. Another BFO.
Moderator buttinski/name-dropping time: Jay was a very close friend of my Dad, even closer to Uncle Shep, and over lunch one weekend gave this 17-yo the mental wallop I've never forgotten...and have used on this forum two or three times: "You can't induce accuracy with precision."
Last edited by a moderator:
racer26
Active Member
While I did jump a bit in response to you originally and I could see why it would be interpreted that way, I wasn't getting worked up.
You're correct, some of the highest end nVidia cards are priced around $1k. In quantity 1. In consumer trim. With retail packaging and the interfacing and connectors to be able to plug it into a computer.
A wholesale deal alone in the sort of quantities needed to build Teslas? Probably cuts the pricetag in half or better. Then cut out all the unneeded stuff?
I don't have any hard info, but my R+D experience in the small-run consumer electronics industry tells me they don't need the absolute tippy top of the line to have sufficient performance (because basically no application for anything that is possible ever does), and that volume discounts and customized packaging that keeps only the bits you actually need saves big value.
Lots of people don't realize it, but general purpose computing power is not very expensive at all. Some of what I do for a living is writing code for microcontrollers. These are the small microchips that are in everything from your TV to your coffee maker. Chances are, if it has buttons and a display and was produced in the last 20 years, there's a microcontroller in it. Simple ones can be had for a few dollars, and really high end ones that can connect to the internet or to USB with all sorts of fancy peripherals are only about $10.
Good microcontroller developers have to understand all sorts of tricks to squeeze maximum performance out of these little devices. These kinds of devices often have limitations on speed, or memory, or pin count, and there are lots of ways to improve performance if you know what you're doing. For example - the cheapest units often don't natively support floating point math operations. If you just write code that uses floating point numbers, the compiler will generate code to produce the correct answer by breaking it down into many memory stores and retrieves and non-floating point math operations on the different pieces. It will get the right answer, but its very slow - adding two floating point numbers together might be tens or hundreds of times slower than a simple integer addition operation. In many cases, if you only care about precision to two decimal places, you can simply multiply all of your numbers by 100 and use integers instead of floating point numbers and you'll get much better performance.
Most programmers today are used to programming on a octo-core monster of a machine with truckloads of RAM and storage space. They're not used to needing to think about what the processor is actually doing, because the theoretical maximum performance of the machine is bigger than what they actually need by orders of magnitude, wasting cycles or bytes here or there doesn't matter on a machine like that.
Its this sort of thing that I expect Tesla was talking about when they said they hoped to be able to do level 5 self-driving on a single DrivePX2 system. Clever programming and product design can squeeze what seems like extra performance from a chip by not being wasteful with the resources that exist. Tesla hires intelligent engineers, I'm sure this sort of thing is within the scope of their abilities.
Hope I haven't run off too far into the weeds. Seemed like a thing people might like to have some tree trunks for, to use Tim Urban's concept from the first few paragraphs of this page: http://waitbutwhy.com/2015/06/how-tesla-will-change-your-life.html ,
I haven't seen any mention of this one one here.
Gearing Up for a Self-Driving Future, Ford Drops $1 Billion on an AI Startup
Apparently Ford is buying less than 100% of a self driving car start up that only started business a couple months ago for $1,000,000,000.00!!!!!
To me this smells of serious panic. The Tesla shorts that think the other car companies are going to immediately crush them when they enter the market are crazy. This is just the latest proof that they do not have the juice to compete on anything to do with software. As far as I can tell they are paying $500,000,000.00 each, for two guys, who had to leave all the IP they have been working on for years behind at Google. They don't even get 100% of the company for that price!!!
Gearing Up for a Self-Driving Future, Ford Drops $1 Billion on an AI Startup
Apparently Ford is buying less than 100% of a self driving car start up that only started business a couple months ago for $1,000,000,000.00!!!!!
To me this smells of serious panic. The Tesla shorts that think the other car companies are going to immediately crush them when they enter the market are crazy. This is just the latest proof that they do not have the juice to compete on anything to do with software. As far as I can tell they are paying $500,000,000.00 each, for two guys, who had to leave all the IP they have been working on for years behind at Google. They don't even get 100% of the company for that price!!!
A few items from today's news confirm Musk's tweet about Rex Tillerson who may be responsible for two pieces today where it seems the new administration has reverted to earlier commitments by preceding administrations. Trump has said of his conversation with the President of China that the US remains committed to the one China policy in contrast to what he apparently discussed with the President of Taiwan in his first foreign policy discussion since the election. Further, CNN has reported the EU foreign affairs chief saying, after a meeting with Michael Flynn and Rex Tillerson, the Iran Nuclear Agreement will be fully respected by the Trump Administration. These are considered very good reversals among most in the foreign affairs intelligentsia. After all, the one China deal goes back to Kissinger's first trip and the Iran deal, though negotiated by Obama, has big power backing including that of Russia.
EU leader: US 'committed' to Iran nuclear deal
A third item proves the unwise position Trump has taken opposing our intelligence community over the Russian attempts to influence the election. It is pretty evident the recent revelations about Flynn's violation of the Logan Act in conversing with the Russian ambassador when he was a private citizen is based on a leak from the intelligence community. He may lose his job over this, at least once Pence becomes president. The NSA if not the FBI which has an ongoing investigation in the matter probably have transcripts of Flynn's conversations.
Cf: Logan Act - Wikipedia
Last edited:
racer26
Active Member
Even if they can throw enough money at the problem to get the software up to snuff, no amount of lighting barrels of money on fire is going to magick millions of battery cells into existence. To get the batteries they need a factory to be built to make them, and building factories takes time. That means you have to decide to build it and start throwing money at the problem now in order to have the solution ready 2-5 years down the road.
Remember, Tesla built gigafactory because they had to. Once fully operational, it will represent approximately half of the world's lithium battery production (and the other half is already spoken for by laptops and cell phones), and Tesla needs 100% of its output to build somewhere between 500k and 1M cars/yr. If the other automakers ever hope to build EVs in anything resembling the quantities they produce ICEs in now, they're going to need dozens of gigafactory scale battery factories to be built. Until somebody announces and starts building a battery factory on the same scale as the gigafactory, there are no credible competitors because there can't be. The raw materials simply don't exist.
I recalled Uber lost some Carnegie Mellon folks after their acquisition:
Peter Rander, Danny Lange, and Marco Campos.
Uber has lost three of its top self-driving engineers
Based on your article, lo and behold who shows up working for Ford now:
"Carnegie Mellon robotics vets Bryan Salesky (a Google alum) and Peter Rander (an Uber veteran) will helm the project, and are currently recruiting a staff of 200."
So Uber's acquisition of CMU's self-driving team is essentially breaking up their expertise to the highest bidders. BTW if you have a chance, watch some of CMU's self-driving demo on Youtube. Very good stuff, for a quite a while now.
(Funny how they call him "an Uber veteran"...)
I agree, but, If they could get the self driving, and other software challenges right in an ICE car, it would give them a lot more time to source batteries and figure out the electric car. If there is only one credible brand of self driving car on the market for a number of years, and Tesla are able to lock down a large percentage of ride sharing.... I can see why Ford is panicking.
everman
Member
larmor
Active Member
So is current market pricing for Tesla including self driving AI engine?
[My naive take on the super-GPU on Tesla]
I think we've seen other times Elon take this approach: put an overkill hardware on the car ASAP, then take your time to develop the software later.
He even does it with batteries, and offers upgrades to customers later. Meaning that he actually loses money for a 75 battery that is software limited at 60.
Probably EM has done his calculations and has seen that economy of scale needs somehow this simplicity, he wants to reduce the moving parts,
so he gets the best hardware possible on cars to achieve the velocity and the "peace of minds" he needs.
Also, he gains a few years before the hardware gets obsolete.
When the AP1 fulfilled its potential, he had the AP2 hardware ready.
I guess this is how he likes to design things: I really like this "clean" approach.
I think we've seen other times Elon take this approach: put an overkill hardware on the car ASAP, then take your time to develop the software later.
He even does it with batteries, and offers upgrades to customers later. Meaning that he actually loses money for a 75 battery that is software limited at 60.
Probably EM has done his calculations and has seen that economy of scale needs somehow this simplicity, he wants to reduce the moving parts,
so he gets the best hardware possible on cars to achieve the velocity and the "peace of minds" he needs.
Also, he gains a few years before the hardware gets obsolete.
When the AP1 fulfilled its potential, he had the AP2 hardware ready.
I guess this is how he likes to design things: I really like this "clean" approach.
CALGARYARSENAL
Member
I didn't realize it was Cory when I was watching the clip. I heard this man full of venom and hatred and then boom.....his mug shows up.
With regard to the cost of the NVidia hardware, it isn't terribly difficult to make a stab at a general price. The Drive PX2 uses the Parker SOC, which is a generation newer than the X1 used in the NVidia Shield, which is a consumer-priced box with networking, USB, HDMI, etc. and sells for $200 (And is considered expensive by many). The Drive PX2 uses one or two CPUs and it looks to me like Tesla would need the dual setup in order to have enough inputs for the AP2 hardware suite, but otherwise the board should cost a fairly similar amount of cash. Assuming wholesale cost to Tesla in the 100's of thousands, I'd expect substantially below $300 is probably likely. Tesla would still be responsible for connections, power, cooling, housing, etc.
Cost of hardware is really not an issue here and as racer26 says, comparing to retail graphics cards with top dollar markup intended for consumers is VERY different from large scale supplier agreements.
I thought this thread might be interesting. Hopefully it is OK to post on the weekend. This is about a NOVA program that talks about how lithium ion batteries work and then also lithium metal and sodium ion. Lithium metal could double density and sodium ion could potentially be cheaper at scale for grid storage since no cobalt is needed. Sodium ion batteries are not dense so those are stationary only.
"Search for the Super Battery" on PBS NOVA this week
"Search for the Super Battery" on PBS NOVA this week
ruralectric
Member
This post captures my feeling on the matter quite succinctly, and as a more or less "all-in long" I feel it's important to say so. I am really looking forward to watching the demise of the ICE age unfold as the M3 hits the streets, and we are all very lucky to be in the position to benefit financially from the imminent disruption. Elon deserves every damn penny he makes for his effort and dedication to the cause, but I encourage folks here to give a thumbs up to Yonki's post to show your support and gratitude to men and women on the front lines who are working their asses off to make it happen.
Over and out.
- Status
Similar threads
- Sticky
- Locked
