Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

TSLA Market Action: 2018 Investor Roundtable

This site may earn commission on affiliate links.
Status
Not open for further replies.
*snicker* Stop driving when the landscape is completely covered in snow? Heh, you wish. ;)

Icelandic-Ring-Road.jpg


Take a look at the road makers we use. Notice how the left hand side has two reflectors, while the right-hand side has only one? Any guesses as to why that might be?

It's so if you're driving in whiteout conditions, and end up driving off the road without realizing it, you'll see either two dots to your right or one dot to your left, and know that you messed up.

Its the millions of little nuances like this to driving, which vary from location to location, that make true full self-driving difficult. One of my favourite examples is: how long do you think it'll be before Autopilot learns to even recognize what a sheep is, let alone how to classify them into ewes, lambs and rams, let alone to understand that if you have a lamb on one side, and a ewe on the other, that's a much more dangerous situation than any other combination of sheep, because the lamb will run to its mother when your car approaches?

It's great that AP adds in other senses that humans don't have. But we also want it to understand the sorts of things that we human drivers know.

That's an awesome picture....beautiful country!
 
*snicker* Stop driving when the landscape is completely covered in snow? Heh, you wish. ;)

Icelandic-Ring-Road.jpg


Take a look at the road makers we use. Notice how the left hand side has two reflectors, while the right-hand side has only one? Any guesses as to why that might be?

It's so if you're driving in whiteout conditions, and end up driving off the road without realizing it, you'll see either two dots to your right or one dot to your left, and know that you messed up.

Its the millions of little nuances like this to driving, which vary from location to location, that make true full self-driving difficult. One of my favourite examples is: how long do you think it'll be before Autopilot learns to even recognize what a sheep is, let alone how to classify them into ewes, lambs and rams, let alone to understand that if you have a lamb on one side, and a ewe on the other, that's a much more dangerous situation than any other combination of sheep, because the lamb will run to its mother when your car approaches?

It's great that AP adds in other senses that humans don't have. But we also want it to understand the sorts of things that we human drivers know.
That logic seems quite simple to program. How hard can it be for computer imaging to recognize shapes/sizes when an AI supercomputer has analyzed a million images of lambs, ewes and rams befobe they even put the finished software in your car. The programmer tells it in a line of code what you just told the world about ewes being say 90 percent more likely to run across....it will be 10x smarter than humans who might be tourists and not know that. The computer technology to do this is already in place.
 
  • Disagree
Reactions: Mader Levap
That's an awesome picture....beautiful country!

You better believe I'm going to drive the heck out of my Model 3 here when I get it ;) There's a reason we're a tourist destination - just check out a google image search for "Iceland nature". That's actually what the countryside looks like. :)

I'm planning on having a "winter / offroad" config for going to remote areas (have been taking part in other threads about how to get the car extra clearance). I want to show up in places that are the last place anyone would ever expect to encounter a Tesla. ;) And then when they're all camping in tents or whatnot in miserable weather, sleeping inside in a climate-controlled cabin.
 
That logic seems quite simple to program. How hard can it be for computer imaging to recognize shapes/sizes when an AI supercomputer has analyzed a million images of lambs, ewes and rams...

That's not the issue. The issue is that this is an example of the countless different things that we humans take into account when driving. Yes, you can program in any specific case; the problem is the number of cases, and how these cases differ from one driving location to the next. E.g. the "rules" for when to decide to drive down the centreline of a road versus staying in your lane could fill a book, and involve, among other things, how likely you are to die if you stray off the road.
 
Take a look at the road makers we use.

Excuse me, but that is cheating (and hardly unique). Naturally, one could just as well introduce road markers that a self-driving car could always sense, regarding of snow depth and other environmental factors.

But it is a nice photo. As I remember, Iceland has basically one major road that pretty much follows the coast and most other rural driving is serious off-road. Trekking there was about as demanding as in Himalaya, with high-speed snow storms mixing into sulfuric steam for midsummer. The absence of pollution was a big plus.
 
  • Like
Reactions: wipster
That's not the issue. The issue is that this is an example of the countless different things that we humans take into account when driving. Yes, you can program in any specific case; the problem is the number of cases, and how these cases differ from one driving location to the next. E.g. the "rules" for when to decide to drive down the centreline of a road versus staying in your lane could fill a book, and involve, among other things, how likely you are to die if you stray off the road.
I don't think people dedicate that much effort to driving. 70 percent of the ppl I pass are texting. This will still be way safer even if there is an AI learning curve while all these scenarios are being taught. Plus the virtual training can literally run through millions of scenarios that a normal person would never see...and give those scenarios to every car in the fleet. That sort of exponential learning power blows away. Mammals are crappy drivers as someone once said....it's true.
 
*snicker* Stop driving when the landscape is completely covered in snow? Heh, you wish. ;)

Icelandic-Ring-Road.jpg


Take a look at the road makers we use. Notice how the left hand side has two reflectors, while the right-hand side has only one? Any guesses as to why that might be?

It's so if you're driving in whiteout conditions, and end up driving off the road without realizing it, you'll see either two dots to your right or one dot to your left, and know that you messed up.

Its the millions of little nuances like this to driving, which vary from location to location, that make true full self-driving difficult. One of my favourite examples is: how long do you think it'll be before Autopilot learns to even recognize what a sheep is, let alone how to classify them into ewes, lambs and rams, let alone to understand that if you have a lamb on one side, and a ewe on the other, that's a much more dangerous situation than any other combination of sheep, because the lamb will run to its mother when your car approaches?

It's great that AP adds in other senses that humans don't have. But we also want it to understand the sorts of things that we human drivers know.
Variations by location for sure.

You don't EVEN want to know what a sheep is recognized as in Alabama...just sayin'. LOL!

Dan
 
Bloomberg Tracker showing well over 6k Model 3's

I wish these were true, but they're guesstimates.
There are pretty reliable rumours (as much as rumours can be) that last few days have been slow because of part availability issues.

Just don't bet the farm.
However, I did bet small part of my house (HELOC) in addition to all my liquid assets; but it's better to be prepared than blindsided.
 
I wish these were true, but they're guesstimates.
There are pretty reliable rumours (as much as rumours can be) that last few days have been slow because of part availability issues.

Just don't bet the farm.
However, I did bet small part of my house (HELOC) in addition to all my liquid assets; but it's better to be prepared than blindsided.

The model is based on VINS registered and those seen in the wild as well as reported. Not heard of the rumours, what's the source?
 
  • Like
Reactions: RobStark
I wish these were true, but they're guesstimates.
There are pretty reliable rumours (as much as rumours can be) that last few days have been slow because of part availability issues.

Just don't bet the farm.
However, I did bet small part of my house (HELOC) in addition to all my liquid assets; but it's better to be prepared than blindsided.
At least you are only betting your house and not a farm. No one should be betting a farm on this, just a house (with mortgage paid off.);)
 
To all those sheep jokes (off topic) I know a newfie joke (the one where the plane's going down). Gotta be good, sorry.

Anyway, I hope to see the AI capability that can drive and react better than I can in my lifetime. Sounds arrogant, but driving safely vs. looking at all options and pulling a stunt level maneuver to escape real-time danger is decades out IMO. That's the time when we'll be paying for Software to drive in a particular style (Granny mode free) or over particular terrain $$$.

I also think data will be downloaded to the vehicle depending upon location (like maps). So no need for snow capability in Central America for example, but would load up if you headed north. I believe they're already doing this in primitive ways as indicated by Fleet level upgrade versions that are location specific (according to Service). So why not have the location "trivia" be provided as needed. We just don't need Camel details over here in the US, and maybe the best way is to go under them? (ouch)
 
Is this good or bad news? Real? Do they need it?
Think deep on this... could all be strategy driven.

It’s good that it’s not for “image repair” as alleged by Charles Gasparino, because it’s FUD that they need image repair in the first place.

Since this company was also involved in the SCTY deal, to me, it’s just one more big player that wouldn’t be getting involved if privatization wasn’t really happening.
 

They are hired by the special committee, not by Tesla and not by the full Board. They are being hired for PR for the going private transaction. If they were being hired for general PR or to improve Tesla's image generally they would have been hired by Tesla management or the full Board, not the special committee charged with handling the Go-Private proposal.

This is a logical step with lawyers and financial advisors already on board and is a very positive sign for the deal moving forward IMO.

Note that Joele Frank specifically rejected the suggestion in Gasparino's tweet that they were hired to help repair Tesla's image:
  • "A Joele Frank representative confirmed to Business Insider that it is working with the special committee, but said the company was not hired for image repair."
Edit: @MikeC beat me to the punch -- agree 100%.
 
Last edited:
Status
Not open for further replies.