Tslynk67
Well-Known Member
This book tweet
Is this a hint that the Maxwell acquisition is complete? Or am I seeing ghosts?
You'll have to explain that one, don't see it...
You can install our site as a web app on your iOS device by utilizing the Add to Home Screen feature in Safari. Please see this thread for more details on this.
Note: This feature may not be available in some browsers.
This book tweet
Is this a hint that the Maxwell acquisition is complete? Or am I seeing ghosts?
GhostsThis book tweet
Is this a hint that the Maxwell acquisition is complete? Or am I seeing ghosts?
They are? I'm in Texas now, and have lived in and/or traveled to many cities, and the only city where the traffic lights are what I'd call "different" is Winnipeg. In Winnipeg, at least the last time I was there, the red lights were always on permanently and then the green lights would come on in addition to the red, but the red never turned off.Different cities in different states have different driving dynamics, road signs, traffic signs. Heck even, traffic lights are different in certain states like Texas.
This is a very good question and is central to understanding Tesla's approach versus Waymo's approach. I don't have much time to get into the full on answer that this topic deserves, but I'll take a gander at a shorter version.
There are two main splits that are instructional. Back in TMC '16, Lex Fridman gave a presentation on self driving along with another speaker. Unfortunately, I don't remember her name. Anyways, he highlighted a schism between two main camps trying to deliver autonomous vehicles. He said there are those that believed that autonomous driving can only come about through internal test fleets and ship to consumer, either through ride sharing (w/o safety driver) or through vehicles once level 4/5 is achieved. Tesla, amongst others, believe in an incremental improvement approach. This drives a completely different development plan and we'll call one the direct birth approach and the other is the incremental approach. The other main split is what Mobileye calls map heavy versus map light. A map heavy approach means a team can get some pretty good looking results pretty quick. Extensively map an area... whether it be a small city neighborhood or an entire city. And then throw a suite of sensors on to try to avoid hitting anything, and voila - you can get hundreds of millions or maybe even billions in VC. It's a pretty good looking demo. But it doesn't solve any of the thorny problems in perception for generalized autonomous driving. The idea is that you get somewhere and then figure out the really hard parts piece by piece. So in that way, they are incremental.
So, there is direct birth versus incremental improvement approach and map heavy versus map light. Waymo is both direct birth and map heavy, while Tesla is incremental improvement and map light. Waymo's approach can get you low disengagement numbers quicker than Tesla's approach. For the hard problems with autonomous driving, we don't know which one will end up winners yet. Maybe both or neither. The CA disengagement reports are therefore exceedingly misleading. To make big movements, developers need to break stuff. They need to find where things are wrong and do it repeatedly. Lots and lots of miles without disengagement, when one is short of the goal, indicates either they are really close to the solution or they are not attempting to seriously tackle the situation. And given the distance from the goal, the real answer is that they are not seriously developing the solution on CA public roads.
The direct birth, map heavy approach leads one to structure the development program in a certain way, and the incremental improvement and map light approach causes a different structure. Tesla's Autonomy Day is best seen as Tesla showing their development process. And few can follow in Tesla's approach to the development process. If the direct birth, map heavy approach ends up hitting local maximum as both Waymo has indicated and Tesla has insinuated, then they are stalled. The incremental approach means lots of humans teaching the system. That means deploying lots of vehicles to lots of humans. The cost per vehicle cannot be very high. Say, in volume, Waymo's approach means the test development vehicle costs $200,000. And they deploy 20,000 of them. That's $4 billion dollars. Each vehicle has been driving about 10,000 to 15,000 miles according to their CA report. That's 300 million miles a year at best. Assuming they cannot get enough ride sharing to cover much of that cost, the labor cost likely approaches $1 billion a year. On the other hand, Tesla's approach is about $1,500 of incremental cost to a vehicle. Tesla's customers pay for that cost. Tesla gets its customers to provide most of the labor for free. They just have to come up with ways to use their customer's miles to be fruitful and they outline a number of their approaches to collect that data and utilize that data during the presentation. There are just over 400,000 AP2.x+ vehicles on the road. Each does roughly 12,500 miles a year. That's 5 billion miles a year. There is no labor cost for driving around and teaching the system, as Tesla owners are performing micro work for the effort. Of course, there are big R&D costs, but we assume that is roughly the same between the two teams. Actually, we pay Tesla for the privilege of doing the micro-work for them. Note that if the Autopilot sensor suite costs $5,000+ in hardware, this approach wouldn't work as not enough people would opt for the sensor suite in their vehicles.
Back on the issue of disengagements. The map heavy approach with LIDAR gets a team a good looking stat pretty quickly. It doesn't mean it can drive, much like when you see Disney's Liberty Belle Paddle Steamer, it isn't actually a boat, but a train sitting in water. Since that Paddle Steamer is riding on rails, it can never sink, it can never get grounded, but it isn't actually a boat.
Regarding Lidar, people who know AI vision already knew that even before Cornell paper. The paper was just coincidental, it was the analyst day presentation that did the trick. So now the attention has shifted to, yes it may work but not in next 10 years. If so, then why bloody give $0 value to this?
Good enough close. 240 held and it's a good setup to continue the bounce in case no other bad news is coming this week. Good InsideEVs numbers for April might support the recovery later this week and capital raise would do the same.
However, it's a long way to go. We recovered just half of the drop from Friday.
It was out during the market hours already. 0 effect on the SP.Old faithful Linette has another hit piece out there.
240 is an anagram of 420.Good enough close. 240 held and it's a good setup to continue the bounce in case no other bad news is coming this week. Good InsideEVs numbers for April might support the recovery later this week and capital raise would do the same.
However, it's a long way to go. We recovered just half of the drop from Friday.
Old faithful Linette has another hit piece out there.
Do you have the FSD computer in your car yet? No. Be patient. Next year is a long ways away.
Cathie Wood is sharp as they come on this with four analysts from diverse experiences combining to make smart bets.
Ark Invest's Cathie Wood defends her Tesla to $4,000 call
Cathie rocks. The other two on that panel.... SMH
Cathie Wood is sharp as they come on this with four analysts from diverse experiences combining to make smart bets.
Ark Invest's Cathie Wood defends her Tesla to $4,000 call
Cathie rocks. The other two on that panel.... SMH
I think she wanted to say more about the absolute lack of technical competition to Tesla's EVs but dialed it back. A point well illustrated by eTron specs."Ark Invest's Cathie Woods defends her Tesla to $4,000 call"