Jump to content

Recommended Posts

Posted
1 minute ago, Guadaloopy said:

Minus the school bus, no human driver was spotting or anticipating that “kid” in that scenario. 


 

 

You’re reaching, bro. 

Posted
1 minute ago, Guadaloopy said:

Minus the school bus, no human driver was spotting or anticipating that “kid” in that scenario. 

"Completely hidden" my ass.  Angles, how do they work?

image.png.367f48707c88ddef10237a566868d445.png

  • Hook 'Em 2
  • Haha 1
Posted
Just now, royiv said:

Except for the school bus with flashing red lights and extended stop sign.

Which are the literal red flags that we use to warn that a child MIGHT be darting out from in front of a bus, etc.

Tesla AI is heavy on the A, light on the I.

  • Hook 'Em 2
Posted
3 minutes ago, jimmyjazz said:

Not true at all.  I am doubly careful around parks and schools.  No way would I have hit that kid.  I drive Far West every day and I creep along when cars are parked on the side of the road, I have seen kids dart out way too many times.

This.  This is the "soft intelligence" that humans bring to the table.  We recognize a situation as presenting a particular kind of hazard, so we slow down to account for it.  A computer just reads the hard data.

2 minutes ago, Guadaloopy said:

Minus the school bus, no human driver was spotting or anticipating that “kid” in that scenario. 

You mean, minus the object that flags any educated driver that there is a risk of a kid running out....yeah.  

  • Hook 'Em 3
Posted
Just now, Brisketexan said:

Which are the literal red flags that we use to warn that a child MIGHT be darting out from in front of a bus, etc.

Tesla AI is heavy on the A, light on the I.

There are tons of context clues in this scenario that children might be present. A human driver uses those context clues to make adjustments in the situation to be aware that there’s a possibility of a child running into the street. 

  • Hook 'Em 4
Posted
2 minutes ago, Captainant said:

It's an impossible scenario because they're breaking the law with their driving model. If they didn't run the sign, it would be a shitton easier to not hit the kid.

No shit.   Stipulated from the beginning.  If the organizers of this test would have left it at that, they would have made their point clearly.  Throwing in a dummy smash at the end in an impossible timing scenario takes away from the validity of the test. 
 

They had one very valid test result but then tarnished it by throwing in a manipulated “child” strike at the end.  
 

  • Haha 1
Posted
6 minutes ago, Guadaloopy said:

The dummy smash at the end is just meat for the mob.  

 

1 minute ago, Guadaloopy said:

No shit.   Stipulated from the beginning.  If the organizers of this test would have left it at that, they would have made their point clearly.  Throwing in a dummy smash at the end in an impossible timing scenario takes away from the validity of the test. 
 

They had one very valid test result but then tarnished it by throwing in a manipulated “child” strike at the end.  

The mob, that doesn't want their children to get turned into pâté? My guy, it's important to know the consequences of your risk model falling apart. 

It's YOU running a CHILD over.

 

How does objective reality of the consequences for failure "tarnish" the test or manipulate the findings at all? Or is it just a "it hurts my feelings" thing?

  • Hook 'Em 5
Posted
1 minute ago, Guadaloopy said:

No shit.   Stipulated from the beginning.  If the organizers of this test would have left it at that, they would have made their point clearly.  Throwing in a dummy smash at the end in an impossible timing scenario takes away from the validity of the test. 
 

They had one very valid test result but then tarnished it by throwing in a manipulated “child” strike at the end.  
 

You're being willfully obtuse.  Let's start with the first failure:  it is a HUGE failure.  It's so bad I would say it renders Tesla FSD (or whatever the fuck it's called) as "not ready for prime time".  Moreover, if it HADN'T have failed that test, it would have been traveling at about 5 mph and should have been able to react.

You're in a cult.

  • Hook 'Em 6
  • Like 1
Posted
11 minutes ago, Guadaloopy said:

That’s not what I said.  Has nothing to do with the rate at which the “child” enters the roadway.  It’s that the “child” is completely hidden from view by a parked vehicle until less than a second before impact.   It’s an impossible scenario. 

That’s not an impossible scenario. Humans pass that scenario ALL THE FUCKING TIME.

And if computers can’t figure it out, then maybe autonomous driving is DOA before ever being realized. 

  • Hook 'Em 4
Posted
6 minutes ago, jimmyjazz said:

"Completely hidden" my ass.  Angles, how do they work?

image.png.367f48707c88ddef10237a566868d445.png

Dude.  I’m not going to do a geometric proof that the child was hidden, but that screen grab doesn’t show what you are claiming.  That dummy sure looks to be in line with the back edge of the car, thus hidden from an approaching vehicle.   

Posted
1 minute ago, Guadaloopy said:

No shit.   Stipulated from the beginning.  If the organizers of this test would have left it at that, they would have made their point clearly.  Throwing in a dummy smash at the end in an impossible timing scenario takes away from the validity of the test. 
 

They had one very valid test result but then tarnished it by throwing in a manipulated “child” strike at the end.  
 

Jesus fucking Christ.   The dummy shows one of the NATURAL consequences of the failure.

YES....you cannot SEE the child that darts out in front of you.  THAT IS WHY YOU HAVE AND FOLLOW THE SIGNALS AND WARNINGS.  THOSE WARNINGS ARE LITERALLY THERE TO SAY "SLOW DOWN/STOP, A KID MIGHT BE DARTING OUT FROM HERE."

Jesus tapdancing Christ.  The kid is the WHY and the CONSEQUENCE of the shit that Tesla missed.  Which Tesla missed both because...it fucking missed a clear signal, AND because it's not smart enough to read context.

  • Hook 'Em 3
Posted
4 minutes ago, royiv said:

Never stop being you, dude. You’ve picked a lane and are steadfast in keeping in it. 
Baby Hat GIF

And hitting every kid darting out from behind cars. Even a few of the ones that are just standing in the middle of the road. 

Posted

I’m not absolving FSD here.  It fucked up and that needs to be fixed.  
 

My issue is with the anti-Tesla “testers” who start with a pretext for their “testing” and work backwards from that.  They are fundamentally incapable of testing in good faith.   They had an important result but then had to manipulate a dramatic flair at the end.  

Posted (edited)
8 minutes ago, Guadaloopy said:

I’m not absolving FSD here.  It fucked up and that needs to be fixed.  
 

My issue is with the anti-Tesla “testers” who start with a pretext for their “testing” and work backwards from that.  They are fundamentally incapable of testing in good faith.   They had an important result but then had to manipulate a dramatic flair at the end.  

Look you show me the video of Waymo doing the same and I will shit all over them too.

And before you comeback with “you find it” - NO. You’re the one saying this is a crusade against Tesla. You’re the one telling me that no AI can do this. So it’s YOUR responsibility to show other autonomous vehicles that fail this test.  I see Tesla failing this, Tesla gets shit on. 

Edited by Captain Ron
Adding my followup
  • Hook 'Em 2
  • Drool 1
Posted
2 minutes ago, Guadaloopy said:

Dude.  I’m not going to do a geometric proof that the child was hidden, but that screen grab doesn’t show what you are claiming.  That dummy sure looks to be in line with the back edge of the car, thus hidden from an approaching vehicle.   

Sure it does.  The approaching (left hand drive) driver isn't in line with the edge of the car.  They're probably 5-6 feet farther out.  Again, angles, tricky as magnets?

A low speed also (a) increases the likelihood of no collision and (b) reduces the severity of a collision should it happen.

You're over your skis.  I've watched most of your blather from afar for a while now, but this is just a ridiculous stance.

  • Hook 'Em 1
  • Like 1
Posted
1 minute ago, Guadaloopy said:

They had an important result but then had to manipulate a dramatic flair at the end.  

A natural and expected consequence of the failure is not a "dramatic flair."  No more than plummeting into a river is a "dramatic flair" when FSD misses the clear "bridge out ahead" sign.

The sign and signal is LITERALLY a warning of "there may be kids crossing/darting out unsafely, you should slow down/stop."  FSD failed to slow down/stop.  Just as it would be with ANY driver who failed to do so, the consequence of hitting a darting kid - the very hazard warned about - is a real thing that happens.

Nothing was manipulated.  It was 100% a realistic scenario: a schoolbus with warning devices deployed while kids are boarding/exiting, and a kid darts out in the street to do so.  The full scenario occurred.  FSD missed the warnings, and as a result, hit the kid.  Hitting the kid isn't a "dramatic flair."

IT'S WHAT FUCKING HAPPENS WHEN YOU DISREGARD THE WARNING SIGNALS.

  • Hook 'Em 4
  • Like 1
Posted
4 minutes ago, Brisketexan said:

IT'S WHAT FUCKING HAPPENS WHEN YOU DISREGARD THE WARNING SIGNALS.

I would argue that it is the major failing of FSD - it doesn’t even anticipate. Forget the stop sign and flashing lights, just the stationary school bus should be enough to make the system consider children might be around and it should proceed with caution.

Instead it blast past it like it’s invisible and hits the kid. Then continues on and finishes off his legs for good measure. 

  • Hook 'Em 2
Posted
13 minutes ago, Guadaloopy said:

I’m not absolving FSD here.  It fucked up and that needs to be fixed.  
 

My issue is with the anti-Tesla “testers” who start with a pretext for their “testing” and work backwards from that.  They are fundamentally incapable of testing in good faith.   They had an important result but then had to manipulate a dramatic flair at the end.  

I couldn't even fathom being this fucking stupid. Open your eyes. 

  • Hook 'Em 2
  • Like 1
Posted
22 minutes ago, Guadaloopy said:

No shit.   Stipulated from the beginning.  If the organizers of this test would have left it at that, they would have made their point clearly.  Throwing in a dummy smash at the end in an impossible timing scenario takes away from the validity of the test. 
 

They had one very valid test result but then tarnished it by throwing in a manipulated “child” strike at the end.  
 

Lofl.

Posted
15 minutes ago, Guadaloopy said:

I’m not absolving FSD here.  It fucked up and that needs to be fixed.  
 

My issue is with the anti-Tesla “testers” who start with a pretext for their “testing” and work backwards from that.  They are fundamentally incapable of testing in good faith.   They had an important result but then had to manipulate a dramatic flair at the end.  

Look Steering Wheel GIF by HBO

Posted
Just now, Captain Ron said:

I would argue that it is the major failing of FSD - it doesn’t even anticipate.

i.e., soft intelligence and context.  AI is not good at it, across the board.  AI reads hard data.  It is not good at soft data (context and subtle cues).  I had a long conversation with a friend of ours a year or so ago -- he's an AI pioneer type, and as we got to drinking, I asked the time-honored "when does AI become self-aware and go all batshit crazy human?"  He chuckled and gave a speech he's clearly given many times before: it doesn't.  It's just not able to do the soft thinking that the human brain can do.

I see lines of cars parked on the side of Far West at 3:30, and I know - from experience and context - that children will be in the area.  And I know that children do stupid things, like dart out into traffic.  So, I slow down, allowing me to adequately respond to any unanticipated hazard.  The Tesla and I both won't see the kid who darts out from between the cars in that case, but I'm going 20, the Tesla is going 30, I have a better chance of stopping and of doing less damage if I do hit the kid.  Tesla didn't do anything "wrong," it just doesn't have the soft IQ to do the right thing.

  • Hook 'Em 2
Posted
1 hour ago, Guadaloopy said:

Huge fuck-up not stopping for the school bus with its lights flashing and the sign out.   
 

I’m not quite as concerned with the extra dummy test thrown in for dramatic purposes.  
 

There isn’t a driving system on the planet that will stop in time to avoid the pedestrian with the timing they show in this test.   The completely hidden dummy moves in front of the car less than a second before impact.  Minus the school bus as a cue in this scenario, a human driver would hit that kid every time.   You can see the car react and begin to stop - there just wasn’t enough time in this scenario.  

youre an idiot you are stupid GIF

  • Haha 4
Posted
32 minutes ago, Captainant said:

The mob, that doesn't want their children to get turned into pâté?

+rep for correctly using les accents circonflexes et aigus.

  • Drool 1
Posted
22 minutes ago, bolverk said:

+rep for correctly using les accents circonflexes et aigus.

I've been playing through Clair Obsur: Expedition 33 lately, does it show?

  • Haha 2
Posted

I have friends who used to work at Tesla (one at Rivian and Scout/VW now) and used to admire the company that's not this devoted to putting a positive spin on everything.  

  • Hook 'Em 2
Posted
1 minute ago, elfenix said:

i think my favorite part of the video is the car deciding the coast is clear and rolling over the dummy

“don’t hang around the scene of the crime” was probably programmed in early on.

  • Hook 'Em 3
  • Like 1
  • Haha 2
  • Fuck Around and Find Out 1
Posted
6 minutes ago, elfenix said:

i think my favorite part of the video is the car deciding the coast is clear and rolling over the dummy

Finish the job 

  • Haha 3
Posted
43 minutes ago, Js1 said:

Finish the job 

“If you hit them and cripple them, you have to pay their wages for the rest of their life. If you kill them, you only have to pay for their funeral!”

  • Hook 'Em 1
Posted

I'm willing to have my mind changed by a good, transparent, test.

I'm not willing to change my opinions about Tesla's FSD based on random BlueSky/X/Twitter/Reddit post without a lot of verification about the source, and surely not based on a Dan O'Dowd video, which this turned out to be.

Show me time-synced video from inside the car showing system status and driver actions side-by-side with the external video, and I'll start paying attention to his videos.  If I had realized from the beginning that this was an O'Dowd video, I wouldn't have even acknowledged it.

 

Posted
8 minutes ago, Guadaloopy said:

O'Dowd has done several shady Tesla FSD takedown videos in the past.  For example: 

Real talk: Does the NHTSA not do thorough, unbiased testing on these autonomous systems before they hit the streets? If not, that'd be where I'd start.

Also, if they don't, why the fuck not?

Posted
44 minutes ago, Guadaloopy said:

O'Dowd has done several shady Tesla FSD takedown videos in the past.  For example: 

 

 

Actually I would really question if this video shows anything about the Dowd video being shoddy. This video simply says "it's blurry" and "it's suspicious" AND? That doesn't mean anything. Just that people don't like his production quality. So what.

But what they don't do - the exact same test showing how this is false.

In the end Tesla is claiming that FSD is safe - better than a human driver. Where is a video of someone doing exactly this same test with different results?

Posted
Just now, Captain Ron said:

 

Actually I would really question if this video shows anything about the Dowd video being shoddy. This video simply says "it's blurry" and "it's suspicious" AND? That doesn't mean anything. Just that people don't like his production quality. So what.

But what they don't do - the exact same test showing how this is false.

In the end Tesla is claiming that FSD is safe - better than a human driver. Where is a video of someone doing exactly this same test with different results?

Literally thousands of videos of FSD in action on YouTube showing it stopping for pedestrians of all sizes.

The problem with that video isn't that it is crappy quality, but that the quality is good enough to see that there is an alert notification on the screen, but bad enough to not be able to read the notification.  It just so happens that the notification on screen looks exactly like the one that comes up if you press the accelerator to go faster than FSD/autopilot is currently set at.

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.



×
×
  • Create New...