I have pure visions of bureaucratic incompetence!
-
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
X Fan
Active Member
Do a little research or just google phone hacking scandal.
He is not a virtous guy….
X Fan
Active Member
Lucid has an awesome vehicle. Their problem is scaling up.
navguy12
Active Member
I really hope it drives better than (the current wide release version of) autopilot on two lane highways with speeds at 100 kph, lots of curves, lots of wide loads in the oncoming traffic whilst not following another vehicle.
They'll start faxing it all over this afternoon.
Pat from Intel was just on 60 Minutes saying they're building 2 more FABs nearby at $10B each. TSMC is also building new FABs here in the Phoenix area. Tesla would need to get in line with the equipment suppliers who are likely overwhelmed. The equipment (Tools) are essentially built by hand. But that's just for the FAB (where wafer's are made), not including things like making the raw ingots and cutting them into wafers.
Then you have Sort/Assy/Test (Test/Sort, Mount, Scribe, Saw, Attach, Encapulate, Cure, Mark, Burn-in, Test, Test some more, Pack, and Ship). That's the simplified process. It's lower-cost but many different suppliers. But would love to see Tesla's version. (I've seen tray flipper Tool that just accepts a stack of trays and flips it. It was called... "The Flipper". It would go through that process step about 4 times maybe more. I don't know the full Flippin' story, but I'm pretty sure someone loaded a machine and forgot to flip it first. Nobody seemed to questions why all this Flipping. Absolutely no Tool industry standards there. So... I'd have to advise against Tesla building their own Semiconductor Factories - recommend partnering with Intel or others.
However, I still believe Lithography in space could shrink the chips a lot more. So a SpaceX FAB? Maybe something there IDK. Interestingly, FAB Tools/Floors are on vibration dampers because a truck driving down the street could misalign the optics... and that was implemented at least 20 years ago.
And then, by the time the FAB is built (takes years normally) will we even need that many cars? Robots maybe someday, but not cars I don't think. Just need to increase utilization of the Tesla Fleet through FSD or incentives. Machine Utilization (MU) will be the indicator to follow, just like the factories.
Last edited:
I know you are joking however consider this - the reason why nobody thinks it is a good idea is because the barrier to entry, the cost of current state of the art is too high. But from a first principle standpoint, does it have to be? Maybe just yet another disruption opportunity.I'm telling ya guys, Elon is prepping the field for a new crop of Tesla home-grown semi-conductors...
With the new product line-up, we're gonna need that FAB.
Softwall Cleanroom - Quickest install, lowest cost. | Flow Cleanrooms ?
Last edited:
I don’t think so, fab volumes need to be ginormous to make sense and Tesla isn’t anywhere close to that big. And then you need a new fab technology every 2 years. So, nah. Best they can do is negotiate supply agreements ahead of time.
Short term shortage would be the best possible outcome. Seems like pushing roadster off to ‘23 may be a hint as to duration or could be when a fab is fully transitioned over to the new owner or I could be completely wrong.
I agree on Dojo being outside for now. Optimus could dwarf present needs however. And then there is the StarLink earth station requirement. Gona need lots of chip to do AI IMO.
All depends on what problem they are trying to solve,,,
Step 1.
Initially I can see Tesla acquiring a smaller company that already owns one or more fabs, making older generation chips.
The purpose is just to shore up supplies of fairly bulk standard automotive chips, not make the FSD and Dojo chips at this stage.
Step 2.
If Tesla eventually wanted to build their own fab for higher end chips, then expertise from the acquisition might come in handy..
We have to fast forward 5-7 years into the future to determine if Tesla can justify building their own high end fab.
That depends on possible orders from SpaceX, car and robot volumes, also any new products Tesla might develop in the next 5 years.
I'm not even 100% sure about Step 1, but it may happen.
jhm
Well-Known Member
The "PRODUCTION DATA" and "REQUEST NUMBER TWO DATA" will give NHTSA analysts the ability to measure exposure to both aided and unaided driving and to link this to subject crashes ("NUMBER TWO"). The crashes can be labeled aided or unaided (or likely some more detailed classification of the state of driving assistance just prior to the crash. Essentially, analysts will be able quantify the odds ratio of a crash (or certain type of crash) with and without driver assistance. Tesla has long asserted that crashes are more like (like twice more likely) for unaided driving. So long as Tesla has been doing an adequate analysis all along, NHTSA should be able to replicate these results.
However, there is a lot of granularity in the subject crash data. Analysts can repeat the above sort of analysis looking as just narrow slices of data. For example, just twilight crashes where the Tesla was headed into the sunlight while the temperature below 32 degrees F, ice was on the road and Mars was in retrograde. An analysts could estimate the odds ratio AP was engaged in just such narrow circumstances. If this is done responsibly by analysts across a reasonable set of conditions, it could provide useful feedback on what sort of accidents AP is preventing and where it may perform worse than unaided human rivers. My expectation is that Tesla performs this sort of data mining exercise internally to focus learning efforts. But, in the wrong hands, this could be come a fishing expedition for analysts that, bent on finding problems, narrow conditions so much that there is no real statistical safeguards against false alarms. In other words, that data could be tortured in such a way as to yield false findings that cannot be replicated with further accumulation of data. Let us hope that NHTSA analysts have more integrity than to go on such a fishing expedition.
Even so, there is no real harm in Tesla engineers being alerted even to false alarms. On march of 9s to a system that is more than 10X safer than human drivers, every crash scenario needs to be well modeled, augmented by simulation, and used to train and test NN and other software. What would be cool is, if for any sort of scenario NHTSA might point to, Tesla can show them the video data and simulated data they have curated. While NHTSA analysts might engage some data dredging, Tesla is building out an advanced library of training data that is way more advanced.
While Tesla may be singled out for extra scrutiny from regulators, this will payoff in the long run. Tesla can set the bar so high (a Dojo full of training data) that it will be hard for competitors to demonstrate the same rigor. Tesla has the potential to create a regulatory mote by setting a high standard for compliance.
Krugerrand
Meow
You have a hard time distinguishing when someone is being funny, huh?
Surely this is bulk data aggregated by model and year. etc and fairly high level..
I don't think Tesla has much to far from presenting bulk data, I'm sure the average performanced of Autopilot is accurately represented by the safety data Tesla provides. So the few random instances of things going wrong will not be significant in the bulk data.
If that bulk data is lots of individual vehicles in different categories, it will add up to the same result.,
totally agree with this.
Tesla will set a Gold Standard for the regulation in assisted driving. Regulators will have to compare Tesla standards to all the companies that want to deploy autonomous or assisted driving. Since Tesla is probably far in front of the other in it’s engineering aspect, it will be a problem for others to deploy their version if they do not pass those tests too. Hope they develop standardized test for peculiar risky situations.
US asks Tesla how Autopilot responds to emergency vehicles
The U.S. government’s highway safety agency wants detailed information on how Tesla’s Autopilot system detects and responds to emergency vehicles that are handling traffic crashes.
apnews.com
Timing on this request for data has me wondering as it coincides with Beta 10 release and the likely headlines that spin it tomorrow. Maybe they're sincerely just trying to get ahead of this FSD enigma. Sooner or later, someone's going to be asking all those ethical questions about robotics we've all been pondering since before even Robocop.
did anyone ever come up with a decent theory for why IV just absolutely crashed alongside a pretty strong move in the underlying?
I have June ‘2023 strike 1200 calls that I bought on 7/15. They are down 8% while the stock went from 650 to 730 (An annualized gain of ~130%). I feel totally jacked by the Russians. This is really bothersome and I’m already inflamed by the irrationality of these meme stocks.
I have June ‘2023 strike 1200 calls that I bought on 7/15. They are down 8% while the stock went from 650 to 730 (An annualized gain of ~130%). I feel totally jacked by the Russians. This is really bothersome and I’m already inflamed by the irrationality of these meme stocks.
Carl Raymond
Active Member
The video is pure speculation, and the tweet has no source.
Nevertheless, worth keeping an ear to the ground for news of Tesla making their own chips. They didn’t shy away from FSD, or cell making, or mining. I can’t see why chip making would be any different. It’s hard and it’s expensive and existing suppliers are slow to scale/adapt… right up Tesla’s alley.
Nevertheless, worth keeping an ear to the ground for news of Tesla making their own chips. They didn’t shy away from FSD, or cell making, or mining. I can’t see why chip making would be any different. It’s hard and it’s expensive and existing suppliers are slow to scale/adapt… right up Tesla’s alley.
It is really odd. I wonder if it's about to set up another gamma squeeze for the next leg up. I'm super happy about it at the moment, as I've been considering converting shares to leaps for a few weeks, but haven't pulled the trigger. I also can't believe the decay I've seen on a $750 June 2023 put that I sold two months ago. It dropped another $8 today, despite the share price barely moving.
Similar threads
- Locked
- Locked
- Locked
