lafrisbee
Active Member
if we go that far can we use the lasers to make the other drivers ill-tempered too!
-
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
Just drove by the freemont factory on I880 north and they have a fence up with tarps covering the fence that say "new project coming soon". Is that just the pilot facility for the new batteries or does Tesla have something else up their sleeve on the cusp of reveal that I'm not aware of?
Here we go, crypto pump and dump schemes is what this forum needs.I'm with @Beltsbear. Basically it comes down to this. We have (at least!) two multi-generational aberrations disrupting the US stock market:
1. COVID-19 --> another Great Recession + Massive US National Debt.
2. The Red Man with the Orange Tan. See #1. [sounds like the name of a children's book!]
I am able to wait patiently for #1 and #2 to end and will not touch my investment in $TSLA. Instead, I periodically watch youtube videos of the tremendous progress being made at GigaShanghai, GigaBerlin, and GigaAustin, etc. and visualize all of the units that will flow from these factories - many using Tesla 4860 cells - generating more massive growth and profits. I'm thrilled $TSLA is cash flow positive and has about $14B in the bank to weather any storm.
My wife and I do not plan to touch our $TSLA investment in our lifetimes. Along with other assets, it is going into a living trust with Schwab Trust Co of Nevada with a rule that they cannot sell more than 1% of it per year for "diversification" after we die, since we have REITS for income. This is generational wealth. Starting to buy $TSLA in 2011 Q3 and adding on since is like those lucky people that invested in Warren's BRK.A in the early days.
What is happening in the world is actually playing right into $TSLA's hands, weakening ALL competitors while Tesla grows stronger by the day. Sure, we might see huge panic selling if Covid-19 craters the market and supply chains are disrupted, millions evicted, etc. I would not be surprised to re-visit the $200's again if it gets really dire here in the US. ICU's turning away Covid-19 patients to die (like happened in Italy for a time) will be a horrible sight to see.
I think that Bitcoin is the new gold to hedge the slim possibility of an American social and economic collapse. Even if that happened, $Tesla will have factories on two other continents. Right now, we have massive social and economic disruption. It really hurts huge segments of the population, but we will come out of it stronger than ever. I mean we have to or else... Society has already changed technologically in significant ways. We will never return to the way it used to be pre-Covid-19.
I have younger LA friends into BTC. I got a Blockchain app wallet and bought a small position (NOT investment advice). It has been rising all year against USD, and really shames the S&P 500 and Gold since October 2020, which is interesting in and of itself:
View attachment 604381
What's great about BTC is it is readily convertible into any major currency. Just don't lose your wallet keys!
lafrisbee
Active Member
it ain't the pilot, the pilot is already up and running in Kato
Bet TSLA
Active Member
And yet miraculously Tesla did "fight geometry" and "line of sight" issues at least once already that we know of. They bounce radar under the vehicle in front of you to, in a superhuman feat, detect the behavior of the car in front of that car that you probably can't see. I am reminded of...
And yet somehow you're sure they will need a hardware upgrade to handle this one obvious situation you've discovered, that somehow they decided they couldn't handle. So instead of being able to simply turn on FSD for most of the existing fleet they've always planned to call them all in for a hardware upgrade? Right.
That is the bottom line, Tesla will improve the sensor suite if it is necessary, and probably even if it isn't totally necessary.
Getting far along the march of 9s will identify any upgrades that are needed. In comparison to getting FSD working, upgrading sensors is no big deal.
But obviously they are starting out with a suite of sensors they think can do the job. While we can speculate, they know a lot more than we do, we can trust Tesla to handle this.
I think there are two problems that need to be solved:
- Looking far enough, may be up to 100 meters in either direction
- judging the speed of those oncoming cars
It is not clear if the cameras with NN processing can do that. Our brains with stereo vision can do. Also wish the cameras are positioned a bit more to the front. The acuteness of this problem of taking a right turn on a fast moving cross traffic can be easily see in many of the current FSD beta videos. Often the car lunges forward when there is a car approaching real fast and the driver had to step on the breaks and take control.
The left turn is even more challenging.
The difference between a good driver and an average one is that a good driver sense the danger of a road cluster configuration and can avoid them and stay away.
The ex wife of my brother had about 1 or 2 accidents yearly from situations she said she was not at fault. Like a drunk driver making a Uturn at a green light she had, people not stopping on the red light, people not making their stop.
in 20 years of driving I had no accident over 200 000 miles of driving and I had about 15 situations I saved my car from near misses. A guy in North Carolina overtaking the left lane when I was there, I expected it and just moved over on the left half of the wheels off road.
3 days ago I was taking a call for work and I was about to go on my green light but saw the car in the cross section jerking his car before I started accelerating, I waited before accelerating and he ended up passing on his red light and crashed in the car in the lane besides mine.
I wonder how the AI will be trained for these unfrequent unusual situations that the driver just see that there is a clusterf&$% about to happen.
Not saying that I am a good driver and could drift in a rally and win LeMans, but I try to avoid at all cost people driving like boneheads either passing them as far as possible to be 1 mile away of them or moving completely out of the way when a Mustang is speedin at 180mph when I’m getting out of my rental car out of Miami airport. These moves might prevent some near misses that can happen 1 or twice a year when you drive over 30 000 miles yearly. I wonder how much FSD can actively place you in a position safer away from these situations. It’s a life saver against crashing your car from getting asleep behind the wheel but I am eager to see how it can be a life saver from other human drivers driving recklessly. Once all car will be FDS, the accident rate will be near 0. It’s the transition period I am interested into.
Last edited:
That’s not the roadrunner facility. That is on Kato road. I haven’t been by there to see the fencing you’re talking about but I heard a rumor about one of those purple “tencent” surprise boxes. I’ll swing by tomorrow and take a gander.
maybe ill swing by tomorrow if someone else doesn't get one up by then. it was only visible for like 10 seconds of drive time and i wasn't going to circle back
I am not claiming any discoveries here. It was obvious that the camera placement was clearly a trade off and the current setup is most optimal for 8 cameras. But even in Palo Alto, and LA I am sure these corner cases are not too far off.
And hey, I already took my car in once for a hardware upgrade to get the FSD. And a second one is very likely going to be in my driveway, unless the engineers develop software that makes radar turn corners or there is blanket wifi that tracks all roads at all times and relays live cross traffic data to my car in realtime to let it make a blind (to the car) turn.
Really? This video made the rounds here when it first happened. Also, it's pretty common for human drivers to creep forward to take a better look before turning. Tesla's FSD just isn't as smooth at doing that yet but I fully expect that it will given some time to learn. As far as the difference of where your eyes are versus the B pillar camera, the difference is about a foot under normal circumstances. It's even less when you stop to look before turning. It's not unusual to angle your car to start into your turn while riding the brake a bit to look before you go. That angling brings your eyes closer to in line with the B pillar camera while looking at oncoming traffic.
Last edited:
I explained to my son when he was a teenager learning to drive that he had better reaction time than I did at my age.
However, I also said not to rely on that entirely as you improve your ‘anticipation’ time as you gain driving experience.
I’m sure some of experienced drivers’ knack for driving will be inferred along with more immediate driving skills during FSD training. Perhaps even giving FSD a bit of anticipation time.
I also expect that some tricks of the human driver’s mind will elude the machines.
Conversely, the FSD system will have some strengths humans lack (hurry up Neuralink!
).Remember, the actuaries and regulators will be looking at aggregate data. As long as there aren’t egregious problem cases, FSD could still be considerably better overall — even if there are areas where it isn’t quite as good as humans.
Bet TSLA
Active Member
You're utterly underestimating the AI. It will have the equivalent of at least 100x the driving experience of any human driver, in a far greater variety of situations, by the time it is qualified. Humans will seem like rank beginner drivers compared to the AIs. And then it will get much better as it gains experience.
It won't be any different than what the best chess and go players are seeing. They look at what the AIs are doing with the games they are experts at with wonder and incomprehension. We drivers will be the same. AIs will get from A to B with speed, precision, and efficiency that humans can't come close to replicating. This is how superhuman learning works. Then they'll spec out what tomorrow's vehicles should look like in order to enable them to do even better.
People who are saying that they'll never be able to go autonomous is Vietnam, or China, or Bangalore have no idea what they've talking about. It will take longer to get there because it requires more skill to operate in an environment with very flexible rules, but AIs thrive in situations that require skill. Humans have fairly limited abilities. And AIs have far better situational awareness, faster reactions, a much superior sensorium, don't tire or become distracted or angry, and the list goes on and on. People are just pretty bad at most things.
Edit: Oh, yeah. At some point lights, stop signs, and speed limits will be eliminated as pointless. After all, the vehicles have to be safe even while assuming that others don't follow the rules, so there's not much point in rules. A few years down the road people will just walk across busy streets trusting that the traffic won't hit them. People who grow up with this will marvel that it was ever any different.
Last edited:
StealthP3D
Well-Known Member
If after you go full margin, your brokerage announces less margin credit on stocks like TSLA (since stocks are obviously in a freefall), you get an instant margin call. I've never been there but it can't be a fun way to live (dependent on others for your solvency).
Life is too short to mess with using other peoples money. It's a cocky and/or desperate thing to do. It might work out but it might not. Plus, you're letting the lender skim sure profits while you're taking all the risk.
Humans are good at creativity and original thought, both things not required for driving.
Once AI is doing all the driving, AI will find it even easier to understand the environment, an amount of creative unpredictability will disappear.
Up to that stage, AI will drive by copying what humans do, with more sensors and attentiveness.
JusRelax
Active Member
In the game Go, AI learned to make moves that seemed completely unorthodox to humans, but which resulted in better outcomes. If I'm not mistaken, I believe the very best Go players have started to learn from these moves and implement them in their own play. I think it is possible that FSD may behave in ways that seem unorthodox as well, but may end up being proven to be safer than human drivers.
EDIT: also, FSD has different hardware than humans as well, with 8 eyes compared to humans' 2 eyes, but likely less persistence of data (the better human drivers keep a map of where vehicles are around them, and if a vehicle 'disappears' from their map, they try to locate it; not sure if the memory within a tesla is large enough to be able to maintain this persistence of data). This could result in different behaviors as well in optimizing for both it's advantages and disadvantages in it's hardware in comparison to humans.
Last edited:
StealthP3D
Well-Known Member
Exactly! Autonomous cars will be bone-headed drivers but they will be far safer than humans due to their alertness (360 degree cameras always on) and ability to react more quickly. Certain situations will be beyond their capabilities but they will more than make up for those cases by being consistently aware and ready to stop or swerve at moments notice. The accidents they DO have will be accidents that a human probably wouldn't have. But the final measuring stick will be the number or accidents/deaths per 100 million miles and, on this metric, they will take home the gold.
In time they will continue to improve capabilities to the point where there is hardly ever an accident of any consequence. It will be considered a freak event and make national news (err...wait, if it's a Tesla on Autopilot, it already does
).Similar threads
- Locked
- Locked
- Locked
