Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable

[Tim S]

I'm reading that Gensler, the new head of the SEC, is going to be confirmed on Monday (FINALLY). Can anyone confirm outside of this article?

Don’t Let Them Kill Digital Assets - Bitcoin Market Journal

The LBRY case lends credence to the presumption that some U.S. officials are discouraging innovation within their borders.

www.bitcoinmarketjournal.com

Edit:

Found the link.

https://www.senate.gov/legislative/LIS/executive_calendar/xcalv.pdf

More: Schwab Expects 'Activist' SEC Under Gensler; Senate Sets Confirmation Vote Date | ThinkAdvisor

Interesting...
"While Gensler’s background as a professor of blockchain, digital currency, financial technology and public policy at the Massachusetts Institute of Technology will help inform his agenda in areas like cryptocurrency,... "

 

[Singuy]

Prefab superchargers arrived, installed, and online within 2 days. Total construction time was around 3 weeks (minus permiting).

https://www.reddit.com/r/teslamotors/comments/mn53kd

 

[goinfraftw]

Interesting...
"While Gensler’s background as a professor of blockchain, digital currency, financial technology and public policy at the Massachusetts Institute of Technology will help inform his agenda in areas like cryptocurrency,... "

It's worth watching his MIT courseware lectures, back from 2018, over the weekend.

 

[Carl Raymond]

I told my cousin to freaking stop monkeying around.
Does anyone seriously doubt Musk and Tesla won’t get FSD up and running soon, and reap the rewards monetarily and for the good of society in lives saved?

My goodness, my cousin is trading stonks and playing pong FFS!!
View attachment 652445

I’ll be the unpopular one and say what I think many are perceiving incorrectly.

I see a lot of people getting excited about zero intervention A to B journeys, because it suggests FSD is near.

It matters not a jot what the car gets right 99.99% of the time. What matters is what the car gets wrong 0.01% of the time. It’s what we don’t see and won’t see on youtube, because a human always intervenes. I think we’re still a good way off.

This doesn’t discourage me a bit from being long TSLA. Nobody else is in a position to beat Tesla to (non hi res pre mapped) FSD. In Automotive, legacy will be eaten alive because they are chained down by ICE stranded asset and the cost of EVs is at the tipping point - a better car for less.

 

[SOULPEDL]

Also bouncing headlights under the car in front of you doesn't help your camera see an otherwise visually blocked car

Strobe headlights 120 Hz and look into the darkness in between. Just a random thought for a Friday.

 

[buttershrimp]

@Artful Dodger i liked your list, but I thought I should do one that is unwise but accurate

"Buttershrimp’s Ten Rules for Investment": 1st edition

1. Stop wasting time, invest as much as possible and as quickly as possible
2. Emotion is fine if you don’t act on it alone, math is hard, keep a calculator handy
3. Buy in a few Tranches during dips, or just buy it already
4. don’t sell TSLA, don’t pull a Gary Black.
5. Often the Best Action is 4 years ago, so stop living in the past and keep investing as long as TSLA keeps being awesome and innovative
6. Volatility puts hair on your chest; ask @Krugerrand to tell you to stop being a little B*****, you’ll eventually stop whining when prices fluctuate
7. Market Makers are best played against one another, always pay your bills so you aren’t a deadbeat and have no leverage
8. Use Margin if rates are good and if you can easily answer calls if things dip a lot, otherwise go nuts
9. Troll the “News” and/or Electrek when someone contradicts your world view of TSLA
10. Have Unreasonable Goals; indulge in fantasy, whatever makes you more motivated to accumulate more TSLA shares, Delude yourself strategically to offset Wild swings in emotion, they say lithium is a mood stabilizer.... so if you buy enough TSLA, the lithium content in the batteries will even you out
11. Diversification is what your 401k is for. Your only goal is to accumulate irresponsible amount of TSLA shares.

 

[SOULPEDL]

It matters not a jot what the car gets right 99.99% of the time. What matters is what the car gets wrong 0.01% of the time. It’s what we don’t see and won’t see on youtube, because a human always intervenes. I think we’re still a good way off.

Computer make mistakes? Well, this kind does and will yes. Time telling.
But I think the car should apologize. "Oops, that was a bit abrupt. We noticed, thanks for being patient."

 

[Curt Renz]

CNBC - this evening:

 

[Tim S]

It's worth watching his MIT courseware lectures, back from 2018, over the weekend.

Are you assigning us homework?

 

[Knightshade]

If you try to think up some scenario where this is not true, and then try to think how vision could get around this(by for example keeping a long distance to the vehicle in front in those situations), I think you will pretty soon exhaust your imagination. And if you still have some example, consider if this scenario happens meanfully often.

There's a big truck in front of you the cameras can't see beyond travelling at speed... ahead of him is another vehicle that unexpectedly brakes very hard (reason is irrelevant but maybe an accident, animal crossing the road, debris, whatever).

The cameras can't see the guy in front of the big truck braking hard.

Radar can.

If the guy in the big truck is slow to react, so will the cameras be slow to react. Not so the radar.


There's been actual examples of this in Teslas preventing accidents.

See video from a Tesla that avoids an accident explicitly because of this here:

Tesla Autopilot avoids a crash before it happens

Tesla's Autopilot 8.0 has a particularly clever feature: it uses radar to track road activity two cars ahead, helping it avoid danger that you wouldn't normally see. And it now appears that this tech just averted a disaster. Dutch Model X owner Frank van Hoesel has dashcam footage showing his...

www.engadget.com

I don't understand the science behind all of this to have a strong opinion. But I know someone who does. And I trust him to decide.

In fairness, the same guy a few years ago went on at some length about how much better radar makes the system.

Being radar it can see through rain, fog, snow, dust, and it recovers quite easily, so even if there was something where you’re driving down the road and visibility was very low and there was a big multi-car pileup ahead of you, you can’t see it, but the radar worked, it would initiate braking before your car being added to the multi-car pileup.

We’re confident we can use radar to look beyond the car in front of you by bouncing radar off the road and around the car, so even if there was something that was obscured both by vision and radar we can use the bounce effect of the radar to look beyond that car and still brake.

In both quotes he cites something cameras physically can not do as a reason radar adds value and safety.



So far the best explanation for the change I've seen has nothing to do with the fact radar is adding info the cameras can't-it's 2 limitations Tesla might be running into with continuing to COMBINE that added data with the 4D re-write.

Specifically-

First, it processes all camera input as 360-degree panoramic video frames, and Tesla doesn't have 360-degree radar.

Second, and more importantly, the ratio of deep learning research published on processing camera images vs. processing radar signals is probably like 10,000:1 or 100,000:1. A hard research problem that Tesla has been working on is how to get a neural net to compare sequential video frames, thereby incorporating temporal information into its predictions (rather than treat every still image as belonging to its own universe. It may just now be possible to do this with camera images, but I'm fairly certain it's still impossible to do this with radar signals.




Credit for the above suggestions goes to user Shrineofachance and I'd suggest further discussion might better move to the thread I'm reposting those ideas from-(post 16 specifically)

Poll: when will FSD Beta v9.x go fleet-wide?

You have 7 days to vote. You can change your vote until then. Votes will be displayed publicly.

teslamotorsclub.com

 

[Big Time]

Perhaps elon and the team have invented some sort of new camera that combines many types of lenses(thermal, infrared, telephoto....others?) etc etc and are about to disrupt the camera industry.
Look out Nikon.....

 

[BornToFly]

I was shocked to come home and read that the ER is on a Monday (never happened in the 8 years I've been holding TSLA). This will make options on the 23rd very interesting. I was going to roll my 800 strike covered calls from the 16th to the 23rd on Monday to pocket more time value, because I was assuming the ER was going to be on the 28th. Now I will just sit tight and let them expire worthless on the 16th and wait until after the ER to do anything else. FOMO might cause a nice spike on the 23rd....

 

[MTL_HABS1909]

I was shocked to come home and read that the ER is on a Monday (never happened in the 8 years I've been holding TSLA). This will make options on the 23rd very interesting. I was going to roll my 800 strike covered calls from the 16th to the 23rd on Monday to pocket more time value, because I was assuming the ER was going to be on the 28th. Now I will just sit tight and let them expire worthless on the 16th and wait a week to do anything else. FOMO might cause a nice spike on the 23rd....

Thought the same thing. It’s ALWAYS on a Wednesday. Wonder why they decided to switch it up....

 

[buttershrimp]

There's a big truck in front of you the cameras can't see beyond travelling at speed... ahead of him is another vehicle that unexpectedly brakes very hard (reason is irrelevant but maybe an accident, animal crossing the road, debris, whatever).

The cameras can't see the guy in front of the big truck braking hard.

Radar can.

If the guy in the big truck is slow to react, so will the cameras be slow to react. Not so the radar.


There's been actual examples of this in Teslas preventing accidents.

See video from a Tesla that avoids an accident explicitly because of this here:

Tesla Autopilot avoids a crash before it happens

Tesla's Autopilot 8.0 has a particularly clever feature: it uses radar to track road activity two cars ahead, helping it avoid danger that you wouldn't normally see. And it now appears that this tech just averted a disaster. Dutch Model X owner Frank van Hoesel has dashcam footage showing his...

www.engadget.com

In fairness, the same guy a few years ago went on at some length about how much better radar makes the system.







In both quotes he cites something cameras physically can not do as a reason radar adds value and safety.



So far the best explanation for the change I've seen has nothing to do with the fact radar is adding info the cameras can't-it's 2 limitations Tesla might be running into with continuing to COMBINE that added data with the 4D re-write.

Specifically-

First, it processes all camera input as 360-degree panoramic video frames, and Tesla doesn't have 360-degree radar.

Second, and more importantly, the ratio of deep learning research published on processing camera images vs. processing radar signals is probably like 10,000:1 or 100,000:1. A hard research problem that Tesla has been working on is how to get a neural net to compare sequential video frames, thereby incorporating temporal information into its predictions (rather than treat every still image as belonging to its own universe. It may just now be possible to do this with camera images, but I'm fairly certain it's still impossible to do this with radar signals.




Credit for the above suggestions goes to user Shrineofachance and I'd suggest further discussion might better move to the thread I'm reposting those ideas from-(post 16 specifically)

Poll: when will FSD Beta v9.x go fleet-wide?

You have 7 days to vote. You can change your vote until then. Votes will be displayed publicly.

teslamotorsclub.com

I would like to hear karpathy talk about this or I bet @jimmy_d will be able to drop knowledge on the radar deletion. Right now I’m betting on negligible benefit over very competent vision. Still it’s the platoon function for caravan of autopilot that it seems would be the most useful for radar.... and close distance caravan may be years and years away and even then, perhaps ultrasonics are just as good.

 

[Alphacrux]

The worst rain is nothing compared to certain types of sticky snow. I've driven in conditions where I've been on the edge of the seat, desperately peering out with the wipers going full speed and a near constant stream of subzero wiper fluid spraying. There's zero chance an autonomous car can drive in those conditions using cameras. FSD can be a thing 99.9 percent of the time and robotaxi can be a thing in certain geographies, but not all.

Agree. Also in conditions called “freezing fog” which is a rare but sometimes thing here in Texas. Defroster on the Tesla kept up at full,power, but only barely.

 

[Howser76]

I would like to hear karpathy talk about this or I bet @jimmy_d will be able to drop knowledge on the radar deletion. Right now I’m betting on negligible benefit over very competent vision. Still it’s the platoon function for caravan of autopilot that it seems would be the most useful for radar.... and close distance caravan may be years and years away and even then, perhaps ultrasonics are just as good.

Agreed. And speaking of Douma, I wonder whether further development and integration of Bird's Eye View (as outlined in his excellent Youtube video with Dave Lee) could have anything to do with being able to remove Radar - there is nothing obstructing your view from a bird's eye shot...

 

[insaneoctane]

Oct 2016, since Tesla broke up with Mobileye. The current FSD computer is a board swap for older models. If customers purchased FSD before the new computer was available, it is a free upgrade with a 1hr appointment at a service center. Most existing FSD customers have already received the upgrade.

Yes, but relative to ALL vehicles, regardless of option configuration, FSD computer has been standard as announced on autonomy day in April 2019. No board swapping required from that date on....

https://twitter.com/x/status/1380599694154362882

Almost ready with FSD Beta V9.0. Step change improvement is massive, especially for weird corner cases & bad weather. Pure vision, no radar.

Speed of improvement seems impressive. Very rapid. Can’t wait to see what dojo does to this equation!

With Elon’s comments here, it’s feeling like April could be epic. Probably a good time to check those seatbelts...

 

[heltok]

There's a big truck in front of you the cameras can't see beyond travelling at speed... ahead of him is another vehicle that unexpectedly brakes very hard (reason is irrelevant but maybe an accident, animal crossing the road, debris, whatever).

The cameras can't see the guy in front of the big truck braking hard.

Radar can.

If the guy in the big truck is slow to react, so will the cameras be slow to react. Not so the radar.


There's been actual examples of this in Teslas preventing accidents.

Ok. So lets work with this. First, would a human be handle this better than the autopilot? Probably worse because of worse reaction time. So this probably happens pretty rare as humans are still allowed to drive... Second the truck is travelling at speed, lets say 70km/h and then hits a stationary car it for some reason didn’t see(it’s not a Tesla semi). Even if this was a perfectly elastic collision, we are probably talking about a 10:1 mass differentional that the kinetic energy will be distributed into will probably not decrease the speed very much, maybe from say 70km/h to 60km/h. See quick

(You can play around with the numbers here: Conservation of Momentum Calculator )

The autopilot should have no trouble with this small decrease in velocity of the truck in front.

And even if we hit the car-truck-carnage our speed differential with the car-truck-carnage would probably not be too many miles per hour, that this would be the difference between life an death.

So would radar help? Maybe. Would it save any lives in these situation? Probably not. Would the benefit of a better vision system outweight this? Maybe, Tesla probably are better at estimating this.

I think if you brainstorm more scenarios you will probably see similar results.

 

[NicoV]

Robotaxi for Cities and Boring tunnels may look something like this design study (from "Westworld" season 3):

View attachment 652543

View attachment 652545

View attachment 652551

Model 1 / Robotaxi features include:

• LFP Bty pack in floor
• seating for 8
• modular wheel motors w. airless tyres
• air suspension for kneeling entry
• Infotainment system not shown

Cheers!

I doubt that. This design has a bottleneck with entry and exit of passengers. IIRC, when Elon talked about a potential mini-passenger bus (many years ago), he wanted to optimise the boarding issue. That can be done with a falcon-wing door at every seat.
Your proposal is more for a family, where everybody can socialize with each other. No need (nor desire) to do that in a bus mostly containing strangers.

 

[insaneoctane]

Dropping radar is hard to accept for some (myself included). I think it’s best to realize while adding radar ought to be a positive and makes sense in theory, if Tesla isn’t seeing positive results out of the NN, it simply may not be practically better. Vision from the 7? cameras might just be so good that the added complexity of the radar parameters doesn’t make a tangible improvement.