TSLA showing cracks?
07-25-2016
, 08:01 PM
07-27-2016
, 03:53 AM
WANT!
Quote:
Musk said that once at full production, he expects the Model 3 to generate around $20 billion in revenue per year for Tesla with $5 billion in gross profit (or 25% gross margin).
It adds up to roughly 500,000 cars per year at a $40,000 price tag.
The Model 3 is expected to start at $35,000 for the base version, but with options and performance versions to be available, the average sale price will be higher.
The 25% gross margin is also an ambitious goal considering Tesla is having difficulty maintaining a similar margin on its Tesla Model S, a vehicle starting at $66,000:Tesla gross margin
But Tesla’s CEO also said that the Model 3 has been designed for being easy to manufacture and battery cost should be much lower once the vehicle hits its full annual production closer to the end of the decade.
During the event, Musk reiterated that he expects battery costs to fall below $100/kWh by 2020. They are currently below $190/kWh. The cost reduction will not only come from the battery cells manufactured at the Gigafactory, but also through lowering the cost of the overall battery pack for the third generation platform.
Finally, Musk reiterated that he is confident the Model 3 will enter production in summer 2017 for deliveries starting shortly after.
It adds up to roughly 500,000 cars per year at a $40,000 price tag.
The Model 3 is expected to start at $35,000 for the base version, but with options and performance versions to be available, the average sale price will be higher.
The 25% gross margin is also an ambitious goal considering Tesla is having difficulty maintaining a similar margin on its Tesla Model S, a vehicle starting at $66,000:Tesla gross margin
But Tesla’s CEO also said that the Model 3 has been designed for being easy to manufacture and battery cost should be much lower once the vehicle hits its full annual production closer to the end of the decade.
During the event, Musk reiterated that he expects battery costs to fall below $100/kWh by 2020. They are currently below $190/kWh. The cost reduction will not only come from the battery cells manufactured at the Gigafactory, but also through lowering the cost of the overall battery pack for the third generation platform.
Finally, Musk reiterated that he is confident the Model 3 will enter production in summer 2017 for deliveries starting shortly after.
07-27-2016
, 11:01 AM
What if Tesla used the industry standard accounting methods to determine COGS?
Cells manufactured at the Gigafactory? This Gigafactory?
The line in black represents the portion that's been completed since 2014.
Elon said the Gigafactory will be funded by revenue (lol) from Model X and S sales as well as a "small capital raise" so at least he is being consistent about one thing, too bad it happens to be diluting the share count every year since IPO.
07-27-2016
, 05:01 PM
Quote:
Norwegian site Autofil compares the new Mercedes E-Class "Drive Pilot" vs the Tesla "Auto Pilot" system. Their real world metric testing came to a similar conclusion as both motortrend and caranddriver testing - "Tesla’s autonomous system is far superior to the Mercedes."
The test revealed a scary situation where the Mercedes system crossed into the lane of oncoming traffic and reset itself there as if it was the proper lane. Although, the benefit of these non tesla systems is that they are ****ty enough that no driver will be convinced that he can watch a movie and or otherwise pay no attention to the road while using them.
The test revealed a scary situation where the Mercedes system crossed into the lane of oncoming traffic and reset itself there as if it was the proper lane. Although, the benefit of these non tesla systems is that they are ****ty enough that no driver will be convinced that he can watch a movie and or otherwise pay no attention to the road while using them.
Cliffs:
Tesla's AutoPilot is good and impressive.
Mercedes DrivePilot is a "disaster" and quote "This is actually a dangerous product".
Quote:
Drive Pilot: The only good thing about Drive Pilot is that your Mercedes will protect you from it. Did I trust it? Only at a crawl. Did I understand it? I don’t understand how Mercedes-Benz could release this to the public. I hated literally everything about it. It drove like a drunk ten year old, fighting for the wheel with a drunk fourteen year old. It was, in most conditions, dangerous.
Quote:
Autopilot: I loved it. A few hours in and one begins to learn a dance between looking out the window, looking at the display and using the stalk to manage speed. Once mastered, the pedals become largely unnecessary. It drives like a very good second year teenage license holder who really wants impress, and is getting better all the time.
It's definitely safer than a human driver alone, assuming you use it as intended. Without a doubt, it is the best ADAS system on the market. It’s incredible Level 2, but also a facsimile of Level 3 without sufficient advance warnings, which is where problems have risen. Ignore the warnings like a child and you will be disappointed, or worse. Heed the warnings like an adult, and it’s the best thing on the market.
It's definitely safer than a human driver alone, assuming you use it as intended. Without a doubt, it is the best ADAS system on the market. It’s incredible Level 2, but also a facsimile of Level 3 without sufficient advance warnings, which is where problems have risen. Ignore the warnings like a child and you will be disappointed, or worse. Heed the warnings like an adult, and it’s the best thing on the market.
I would quote mine all the stupid things TS has said about Tesla's autopilot being super terrible and others crushing them to lol at but I'm busy with work.
Last edited by Cuban B; 07-27-2016 at 05:15 PM.
07-30-2016
, 03:22 AM
Quote:
I just finished watching the interview.
Executive summary:
But the fact he's going all-in on cellphone camera data, to me confirms that Uber is in a strong position if they can deploy any kind of sensor suite across their fleet (even just dashboard camera).
Executive summary:
- Comma.ai is focused on Level 3 autonomy, no Level 4 plans in sight.
- Hotz believes this is entirely a software problem, not hardware.
- Their software strategy is to apply deep learning to camera inputs.
- Hotz believes data from cellphone cameras today are good enough to train best-in-class Level 3.
- They hope to get training data at the necessary scale by giving away a free dashboard cam app for cell phones.
But the fact he's going all-in on cellphone camera data, to me confirms that Uber is in a strong position if they can deploy any kind of sensor suite across their fleet (even just dashboard camera).
https://commacoloring.herokuapp.com
They gather a lot of data from users mobile phones. Exactly how they plan to use this for "training" is not clear. Maybe it will be used as testing or to train a simulation that will synthesize training data. Imo they clearly need side camera data to do lvl3 so one single smartphone looking forward will not be sufficient for training.
When I asked geohotz about Mobileyes paper about end2end DL not being feasible his reply was pretty funny:
Quote:
Amnon Shashua wrote a paper claiming that End-End Deep learning for SDC is not feasible. http://arxiv.org/abs/1604.06915
Talk: https://www.youtube.com/watch?v=GCMXXXmxG-I
Any comments?
Talk: https://www.youtube.com/watch?v=GCMXXXmxG-I
Any comments?
Quote:
In other news, horse and buggy maker talks about why removing the horse from the carriage is just not feasible.
Some things are not possible, like FTL communication. Other things, like this, will just take time and cleverness.
Some things are not possible, like FTL communication. Other things, like this, will just take time and cleverness.
---
Tesla Gigafactory Opening livefeed:
https://www.facebook.com/tesla
07-31-2016
, 03:05 AM
That would be astounding if true, as if you scan through e.g. posts in the Quora "deep learning" category, the research community clearly believes it is a young field with many big unsolved problems.
Typical comments are like this excerpt from a Yann LeCunn post, "Deep learning has more or less cracked the most basic form of visual perception: classifying the dominant object in an image when the object comes from a constrained set of object categories. However, despite this success, deep learning is still very far from human-level visual perception when these constraints are loosened, or when the task is changed from image classification to detecting and segmenting objects. For these more challenging tasks, or when objects are small, or when the set of possible categories is expanded to include hundreds of thousands of objects, more research is needed."
But Hotz is putting his money where his mouth is, so I guess we can just wait and see if he succeeds.
Typical comments are like this excerpt from a Yann LeCunn post, "Deep learning has more or less cracked the most basic form of visual perception: classifying the dominant object in an image when the object comes from a constrained set of object categories. However, despite this success, deep learning is still very far from human-level visual perception when these constraints are loosened, or when the task is changed from image classification to detecting and segmenting objects. For these more challenging tasks, or when objects are small, or when the set of possible categories is expanded to include hundreds of thousands of objects, more research is needed."
But Hotz is putting his money where his mouth is, so I guess we can just wait and see if he succeeds.