Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register
This site may earn commission on affiliate links.
'notha' fire cracker by mr. "PM YOUR NIPS PAPER" (guy who claims he works/worked in teslas ap program) over @ reddit:
Looking at other posts by this person, he's spent no real time in a data center, knows virtually nothing about machine learning, and is not qualified to be saying anything about Tesla's autopilot program.

He confirms in another post to be a college student...
 
Looking at other posts by this person, he's spent no real time in a data center, knows virtually nothing about machine learning, and is not qualified to be saying anything about Tesla's autopilot program.

He confirms in another post to be a college student...

I've said before in this thread that everyone on reddit is an expert and an insider. I wish people would stop quoting talk-only users as sources. It's a major insult to the REAL experts here on TMC, who produce actual material for people to peruse. On reddit, any college washout can go to a library, create an account, and pretend to be an authority on anything they want, as long as they throw out technical language and are inflammatory enough. I'm not sure how well people here understand the the culture there, but the place is supermagnet for such types.

As an aside, I've done my fair share of time on reddit. The environment there is extremely toxic for those with real technical knowledge. I have a Ph.D. in mathematics and am an expert in aircraft navigation, and I've still had obviously unqualified people fight me over basic facts from my areas of expertise. It makes some of the more vicious discussions here look like friendly tea time exchanges.
 
Last edited:
Looking at other posts by this person, he's spent no real time in a data center, knows virtually nothing about machine learning, and is not qualified to be saying anything about Tesla's autopilot program.

He confirms in another post to be a college student...

I remember all the people that dug up details about Eds and doubted him too.

Yet Eds knew. Many experts on TMC were made look like the proverbial fools by Eds' info.

I'd be careful about saying yay or nay about any anonymous source. But I'd pay attention.
 
  • Like
Reactions: davidc18
Consider this, the company he works for has fewer Tesla GPUs than mine and they host production models with Matlab...??

Let's break down the original post:
That video was pretty much faked. The car was hard-coded with specific GPS positions and (x,y,z) locations of traffic lights on a very controlled route. They held 3-4 trial runs and picked the one that worked the best.

All the rectangular bounding boxes and colored line/region segmentations on the "AI view" were done after the car had made the trip, for visualization and hype purposes.

I can't exactly say my sources, but if that video was true over a year ago, why is the current autopilot much worse than what is portrayed in the video?

  • The video itself is hard evidence that the video wasn't faked... it wasn't CGI, there's clearly a car driving itself.
  • As far as GPS coords... of course there was a planned route. There will be for every autonomous ride, otherwise how would it know where you want to go and how to get there.
  • Location of traffic lights are going to be part of high definition maps, they are part of the MobilEye maps too as described previously by Amnon Shashua
  • Yes they held many trial runs why wouldn't you? The disengagement data alone shows how little training was needed for that route.
  • It was also trained for that very route, there's no reason to think it wasn't, but that doesn't mean the video is fake. It shows what the system is capable of with the current sensors using a bit of training.
  • As for the "AI view" being done afterwards...perhaps or perhaps that's for engineers as previously stated. The computer doesn't need that and can represent the bounding boxes with four numerical values, five if you include the class.
  • From the video do I think it generalizes to all locations... heck no, they never said it did.
  • Do I think the current autopilot uses the same code base... It doesn't have to and when you look at how events took place with MobilEye, there's strong evidence that there was a full self driving codebase being worked on before Tesla ever knew they'd need to recreate autopilot 1 functionality.
Keep in mind algorithms like YOLO are super fast so there's no reason to think the embedded hardware is not capable of it.
 
I don't know how I feel about this as a stock holder. I articulate anticipate majority of Sr owners won't pay$9k to extend battery later.
Assuming they do:
Sure, but once they trade it in to Tesla, it can become a CPO LR with associated increase in sales price for free. As long as the difference in incremental cost of the two battery sizes with allowance for simplified manufacturing is less than the increase in CPO value averaged across the fleet CPO rate, it's a good idea.
Otherwise, the difference in LR/SR CPO price will cause SR take rate to be poor, potentially forcing Tesla to swap packs at full LR cost plus labor, but that is down the road when demand is less.
 
Well, be this as it may (I can't estimate the reliability of such a claim) now we at least have a very clear detail to check the reliability of this anonymous poster against after the fact. Nobody before has suggested AFAIK that Model 3's smaller battery version would have a software-limited long range battery, so we shall see if the poster knows or does not know...
 
Well, be this as it may (I can't estimate the reliability of such a claim) now we at least have a very clear detail to check the reliability of this anonymous poster against after the fact. Nobody before has suggested AFAIK that Model 3's smaller battery version would have a software-limited long range battery, so we shall see if the poster knows or does not know...

It has been discussed on here, can't remember the thread. In the absence of any battery lables on the 3 it's one of those things that can only be proven by Tesla offering range upgrades (or someone hacking/dismantling a 3SR). So an easy "prediction" to make without immediate reputational damage.

IMO the only chance of LR batteries going into SR cars would be if Tesla had a shortage of SR batteries and a tough target to meet.
 
... it's one of those things that can only be proven by Tesla offering range upgrades (or someone hacking/dismantling a 3SR). So an easy "prediction" to make without immediate reputational damage.

Tesla PressKit regarding Model 3:

Press Kit | Tesla

  • Curb weight:
    • 3549 lbs. (Model 3)
    • 3814 lbs. (Model 3 Long Range)
  • Weight distribution:
    • 47% front, 53% rear (Model 3)
    • 48% front, 52% rear (Model 3 Long Range)

Put it on a scale and we will know it.

Or better:

Send me one and I will do it :)
 
It has been discussed on here, can't remember the thread. In the absence of any battery lables on the 3 it's one of those things that can only be proven by Tesla offering range upgrades (or someone hacking/dismantling a 3SR). So an easy "prediction" to make without immediate reputational damage.

IMO the only chance of LR batteries going into SR cars would be if Tesla had a shortage of SR batteries and a tough target to meet.

I haven't seen any signs of this, though. As @andifu says, the weight actually points to the contrary.

If anything, if this was PM_YOUR_NIPS_PAPERS really posting this (I haven't seen the post itself - link? so we can judge if he really said that?), it makes me doubt him/her a bit more. It seems a bit unbelievable.

That said, if it turns out to be true, I would tend to believe PM_YOUR_NIPS_PAPERS more afterwards.
 
  • Like
Reactions: Swift and andifu