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

Here's What's Missing from Self-Driving Cars: TRUST

This site may earn commission on affiliate links.
Your takeaway should probably be that GM has delayed introduction of Super Cruise, period. There's no evidence w/ regards to how reliable it will be when released.

do you take the same stance when Elon/Tesla delayed any one of its billionth features that he at some point promised?
or did you repeat elon's excuses and promises verbatim...i highly distaste hypo-critics.
 
Last edited:
no way i can trust any computer to drive my car around....never.....who will you blame when you get smashed up? someone gets killed?...car gets dinged? the computer? tesla? lol.....i can see why insurance will go skyhigh with ap cars...
And by association... how does someone live with themselves after they allow the car to make decisions for them, which causes an accident where one of their passengers - a friend or family member - dies. AP2 does lots of crazy things, not the least of which is throwing your vehicle into the opposite lane of traffic which has been documented numerous times. One day that will cause a head-on collision. A human driver in the same situation would likely not have caused this.
 
  • Like
Reactions: NerdUno
What's missing here is patience and actual statistics. I think pro-self-driving folks are often overly optimistic about capabilities and timelines. Some say Tesla lies. I don't think they do. I think they are just overly optimistic and can't meet their own ambitions as stated.

The important matter for self driving or autopilot technology in the wild is that the statistics show it is generally safer than human driving. But even then, you have to ask yourself, are you an average driver? If you are an above average driver, you may still be better than autopilot or self driving technology. Neither will ever be entirely flawless, and crashes will happen in both cases. It's just a question of trying to accurately assess the risks of you driving versus the car driving itself, and accepting those risks and whatever consequences may result.

If you crash when AP or self driving would not have, do you lament that you didn't turn over the driving to the car? If you crash when AP or self driving is driving and you would not have, do you lament that you didn't drive yourself? It's all hindsight. Sometimes $*** happens. You just have to assess the risks as best you can, make the call, and live with the result, good or bad.
 
  • Helpful
Reactions: NerdUno
This is impressive, more so that the Cadillac president is doing the demo at 73 miles an hour on curvy roads of I-280/
But I don't get how the lidar mapped data is going to help when a road hazard appears, like stopped vehicle partially blocking lanes, someone cutting in front, animals, lane diversion, water logging under overpasses, etc. Hopefully the software is mature enough. An accident can happen in a few seconds, so the alert after a few seconds of distraction may not help. GM did have more autonomous miles under their belt in California autonomous driving. Still, I remain fearful.

GM seems to be getting pushed to the edge by Tesla's bold marketing to bring this to market.
Is GM Supercruise using MobilEyeQ3?
 
My takeaway is that GM has delayed introduction of Super Cruise until it actually worked reliably
IMG_1060.JPG
 
do you take the same stance when Elon/Tesla delayed any one of its billionth features that he at some point promised?
or did you repeat elon's excuses and promises verbatim...i highly distaste hypo-critics.

I do. I don't tend to comment on the reliability of any unreleased feature of EAP or that of FSD as a whole(as it's entirely unreleased). To that point, I don't recall ever seeing any positive claim about the reliability of future updates/features by anyone else either(other than Musk, himself). The people making claims about how reliable it will be tend to be those claiming it won't work.

And I highly distaste tu quoque arguments.
 
I do. I don't tend to comment on the reliability of any unreleased feature of EAP or that of FSD as a whole(as it's entirely unreleased). To that point, I don't recall ever seeing any positive claim about the reliability of future updates/features by anyone else either(other than Musk, himself). The people making claims about how reliable it will be tend to be those claiming it won't work.

And I highly distaste tu quoque arguments.

really? you completely missed all of @stopcrazypp @Reciprocity and @JeffK posts...
 
I do. I don't tend to comment on the reliability of any unreleased feature of EAP or that of FSD as a whole(as it's entirely unreleased). To that point, I don't recall ever seeing any positive claim about the reliability of future updates/features by anyone else either(other than Musk, himself). The people making claims about how reliable it will be tend to be those claiming it won't work.

And I highly distaste tu quoque arguments.
really? you completely missed all of @stopcrazypp @Reciprocity and @JeffK posts...

Exactly, I can claim as fact that with regards to FSD hardware that in the future cameras will still work like cameras and radar will still work like radar.

I can also claim that Tesla has said that, if need be, the unit which is performing the inferences, in this case a custom DrivePX2 based board is easily upgradable.

We can also use statistics to measure performance of a neural network.
 
Exactly, I can claim as fact that with regards to FSD hardware that in the future cameras will still work like cameras and radar will still work like radar.

I can also claim that Tesla has said that, if need be, the unit which is performing the inferences, in this case a custom DrivePX2 based board is easily upgradable.

We can also use statistics to measure performance of a neural network.

To be absolutely clear, we cannot use statistics to measure the performance of a neural network that does not yet exist ;)
 
To be absolutely clear, we cannot use statistics to measure the performance of a neural network that does not yet exist ;)

Sure, but we can use statistics to measure the performance of any networks that do exist. We can also test the ones for self driving in shadow mode without negatively impacting drivers.

There are several models that do exist already though (object detection).
 
  • Informative
Reactions: NerdUno
Is that implying you work for Tesla? If so, then I'll defer to your insider knowledge. If not, then we can't test anything; only Tesla has knowledge of what is/isn't in that data.
We humans, we researchers, and no, not just Tesla.

You can go take some random footage of a road and run it through any self driving or partially self driving networks (ADAS) and measure results. George Hotz was doing this with cell phone footage.

You can go to openAI get some tools and even do this with video games.
 
We humans, we researchers, and no, not just Tesla.

You can go take some random footage of a road and run it through any self driving or partially self driving networks (ADAS) and measure results. George Hotz was doing this with cell phone footage.

You can go to openAI get some tools and even do this with video games.

Ok... I'm lost. I was just saying we can't say anything about the reliability of future EAP/FSD features that have not yet been made by Tesla.
 
Ok... I'm lost. I was just saying we can't say anything about the reliability of future EAP/FSD features that have not yet been made by Tesla.
Yes, you are lost. You are concerned with only the software in which stats will be given to regulating bodies before it's ever allowed to drive itself.

You seem to not believe in statistics. Just because you, personally, don't have insight into the proprietary data doesn't make it any more or less effective.
 
  • Like
Reactions: NerdUno
  • Like
Reactions: EinSV