Havent you read this forum? There is no problem at all with phantom braking. Its all in Tesla drivers minds since the brake pedal itself doesnt move down towards the floor, its not considered "braking"
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Phantom Braking
- Thread starter 2101Guy
- Start date
COS Blue
Member
Well here's one reason: Tesla Model 3 Failed To Save Dummies In AAA Emergency Braking Test
The video is pretty fun. After you watch a few dummy pedestrians get run over, you can skip to 3:20 to see what they have to say about the M3.
And if you put a 5 year old behind the wheel, will he/she stop? I'm not saying computers can do it right now, but given enough training and software development, they will. Just like a teenager is not as safe of a driver as an adult who has 10+ years of experience.
We don't say you can't get a drivers license until you've had 10+ years experience. But how to you get 10+ years experience? That 15 or 16 year old has to get behind the wheel of a car and drive to get experience. Will they make mistakes and get into accidents? Yes, but they learn and get better.
Computers will learn and get better too - and likely much faster than humans can. Because that 15 or 16 year old driver only has their experiences to teach them. Computers can get experiences from other vehicles to teach it instead of only learning from its own.
I personally may never see a round-about where I live - but my car will know how to drive around one because other cars have used them. I would approach it and have to "guess" what to do.
Oh, and BTW - that video was funny - thanks for the mid-day chuckle
I'll counter with another funny video:
COS Blue
Member
I think maybe he was being sarcastic and pointing out that many of us who complain about PB are simply told that PB isn't real, it's just letting up on the accelerator, if we think it slammed on the brakes we are just wrong because this one time this one guy measured something and it was only 0.2G, and therefore we are wrong, and if you are afraid of PB it's just like someone who is afraid of ghosts.
But I could be wrong
.
OncomingStorm
Member
i MAY have been responding to the wrong user hahahaI think maybe he was being sarcastic and pointing out that many of us who complain about PB are simply told that PB isn't real, it's just letting up on the accelerator, if we think it slammed on the brakes we are just wrong because this one time this one guy measured something and it was only 0.2G, and therefore we are wrong, and if you are afraid of PB it's just like someone who is afraid of ghosts.
But I could be wrong
sleepydoc
Well-Known Member
...and they even mention the problem of phantom braking (4:10)
The author also brings up some very good points about the impact of software revisions on test results, how Tesla could release new software that improves the scores, or even release new software that makes things worse (but that could NEVER happen with Tesla, right? /s)
Humans are special and better than computers will ever be, in ways we don't understand. Since WE really don't understand how we do everything we do, computers and software cannot understand. Plus, "safer" than human driving still will not be as comfortable as how we drive, so in the end, not desirable (by us). Human comfort is a messy nuisance that big tech cares very little about.
Beyond "simple" computers and software, artificial intelligence WILL be able to evolve to the point of autonomy. But hopefully, that's far away, and i think it is. I believe Elon when he says AI will surpass us and we will become irrelevant unless we merge with it. Even at this early stage, we see Tesla setting the tone for just how irrelevant the customers are, in favor of the technology.
Computers replacing humans - 100% agree with you. It's a long way off before AI can truly function like a human. But we're talking about driving a car. Which is a very limited set of operations for a human. There is a wheel we turn left or right, a pedal we push with our foot to make the car go faster, or remove our foot to make it slow down. Another pedal we push with our foot when we want to slow down more rapidly and stop. And a few sticks, conveniently placed around the wheel we can apply pressure to in order to indicate our intentions to other drivers. The vast majority of operations in today's cars are computer controlled, and happen autonomously. Air bags, seat belt systems, ABS, Traction-Control, AEB, headlights that turn on and off automatically, and even mirrors that adjust themselves when we reverse or a bright headlight hits them.
I'll give you an example - how many times have you gotten into the car, likely coming home from work, and are totally distracted with a problem at work - or perhaps something in your personal life. Or you take an important phone call (hopefully using your hands-free bluetooth system). Perhaps you were lost in the music and singing at the top of your lungs to your favorite album. Then you get home and think to yourself - how did I get home? You don't even remember driving home - you did it totally on your brain's version of "autopilot". The only time your brain snapped you back into driving focus is when something out of the ordinary happens. Someone slams on their brakes, a car makes an unexpected lane change, or a pedestrian jaywalks in front of you. That "autopilot" your brain had is where FSD is heading rapidly. It won't be long before it can handle the basics of getting you home, only asking for your attention when unexpected things happen. And then it will learn about those unexpected things, as fleets of vehicles report those events back to a central database that all cars can tap into and learn.
COS Blue
Member
I'm sure this will get posted elsewhere but pretty relevant to the whole idea of getting computers to figure out how to drive like humans. Haven't watched the videos yet but it's a very interesting read: https://www.washingtonpost.com/technology/2022/02/10/video-tesla-full-self-driving-beta/
sleepydoc
Well-Known Member
Have you actually driven in inclement weather? The fact that you have to ask the question and your response tells me you have no comprehension why and an explanation will probably be useless, but for starters, why does a car need 8 cameras vs he person's 2? Plus ultrasonic sensors. A human is far better at pattern recognition. A human brain is far better at adapting to situations and can figure out unfamiliar situations.May I ask why?
Why do people think that computers with software can't handle these situations like a human can? Humans have two HD cameras (eyes) on a swivel mount (neck), which can only swivel about 90 degrees left and right, and about 45 degrees up and down. We have a brain, which is just a biological computer that accepts sensory input (eyes, ears, nose, etc.), processes that data (pattern recognition), accesses historical databases (memory), and uses decision trees based on those databases to determine an outcome. BTW, the human brain has about a 120ms lag - which should totally freak you out when you think about it - everything you see and think is happening RIGHT NOW, actually happened 120ms in the past. But that's another discussion.
Since we can build computers and processors that can read sensory data much faster than we can (human eyesight is roughly 60fps), access databases much faster than we can, with much more accuracy, we should be able to handle anything a human can handle. We've proven that computers have a massively faster reaction time versus humans. While it takes you 120ms to process that child's bouncing ball coming into the street, and then your personal reaction time to decide how to handle it (swerve, slam on the breaks, etc.), a computer can see that ball and react 10x faster, applying the breaks or turning the wheel.
Most cars have traction-control, which senses when the car is slipping, as in your example of black ice, and attempts to alter power from wheel to wheel to regain control. Many times this happens before we even know what's going on - for example in hydroplaning. You hit a puddle of water and hydroplane, but the car reacts and maintains control - you feel an odd sensation of the car changing direction, or vibrating as it pumps the breaks dozens of times per second.
And snow covering the cameras? Why can't the windshield wipers automatically engage and clear the view for the cameras? If the cameras (which are inside the car) can't see the road, it's very likely your human cameras can't see the road either. What would you do if you couldn't see the road? Slow down as safely as you thought possible and possibly even stop the car - turning on your hazards? Why can't the computer in the car do the same thing? It attempts to clear the windshield, and if it can't then it slows down the car, turning on the hazards, and comes to a stop.
Oh, and on top of all that, we as humans do all our driving with 2 cameras. Our Tesla's have 8 cameras that can see all around it. We have to check our blind spot by turning our heads, which means we no longer can see what's ahead of us. The Tesla can see the blind spot and what's ahead of it, and what's behind it at the same time.
In the last snow storm, I could barely distinguish where the road was. when I met another car on the road we slowly passed each other, each making sure s/he wasn't too far towards the middle or too far towards the shoulder. On several occasions I had to deduce where the shoulder was based on how the wheels felt.
I let FSD try driving on the (empty) snow covered streets in my neighborhood the day after the snowstorm. Despite the fact that it had to use traction control, it didn't slow down at all. When the car came to an intersection it slid ¾ of the way through.
I can go on, but like I said - if you don't get it now, you won't get it.
and that's what you don't get. A car has 6 functions - stop, go, forward, backwards, left, right. It also has potentially infinite circumstances in which to perform those functions. It's not the functions but how to use them to deal with the circumstances.
Last edited:
But our Lord and Savior Elon has it all figured out. He told us “NYC to LA in any kind of weather with no human input by 2019”.
And he’s always right.
sleepydoc
Well-Known Member
I think we need to put Elon as a pedestrian on streets navigated by his autonomous cars!
Do your eyes not move? Add that to the neck and a bit of torso movement and we can see in a complete sphere, 360° horizontal, 360° vertical even while seated.May I ask why?
Why do people think that computers with software can't handle these situations like a human can? Humans have two HD cameras (eyes) on a swivel mount (neck), which can only swivel about 90 degrees left and right, and about 45 degrees up and down. We have a brain, which is just a biological computer that accepts sensory input (eyes, ears, nose, etc.), processes that data (pattern recognition), accesses historical databases (memory), and uses decision trees based on those databases to determine an outcome. BTW, the human brain has about a 120ms lag - which should totally freak you out when you think about it - everything you see and think is happening RIGHT NOW, actually happened 120ms in the past. But that's another discussion.
Since we can build computers and processors that can read sensory data much faster than we can (human eyesight is roughly 60fps), access databases much faster than we can, with much more accuracy, we should be able to handle anything a human can handle. We've proven that computers have a massively faster reaction time versus humans. While it takes you 120ms to process that child's bouncing ball coming into the street, and then your personal reaction time to decide how to handle it (swerve, slam on the breaks, etc.), a computer can see that ball and react 10x faster, applying the breaks or turning the wheel.
Most cars have traction-control, which senses when the car is slipping, as in your example of black ice, and attempts to alter power from wheel to wheel to regain control. Many times this happens before we even know what's going on - for example in hydroplaning. You hit a puddle of water and hydroplane, but the car reacts and maintains control - you feel an odd sensation of the car changing direction, or vibrating as it pumps the breaks dozens of times per second.
And snow covering the cameras? Why can't the windshield wipers automatically engage and clear the view for the cameras? If the cameras (which are inside the car) can't see the road, it's very likely your human cameras can't see the road either. What would you do if you couldn't see the road? Slow down as safely as you thought possible and possibly even stop the car - turning on your hazards? Why can't the computer in the car do the same thing? It attempts to clear the windshield, and if it can't then it slows down the car, turning on the hazards, and comes to a stop.
Oh, and on top of all that, we as humans do all our driving with 2 cameras. Our Tesla's have 8 cameras that can see all around it. We have to check our blind spot by turning our heads, which means we no longer can see what's ahead of us. The Tesla can see the blind spot and what's ahead of it, and what's behind it at the same time.
My counterpoint to your whole argument is: FSD Beta. If everything you said is true, why does it still utterly suck at driving?
I agree there is every expectation that one day computers will be able to drive just fine, sure probably better than us. But not this version.
That last part of that is not correct.
What's included with FSD Beta is Tesla Vision, and this is used on all cars whether they have Radar or not. The radar isn't used with Tesla Vision. Basically once you go FSD Beta you're no longer benefiting from the Radar at all.
With Tesla vision you'll see the following differences:
Max allowable speed is 80mph instead of 90mph
Autobrights will default to on when turning on AP
Even TACC will be sensitive to rain while it will want to lower the speed due to weather.
I didn't make it clear, my apologies. I'm not saying that computers can handle it right now. Right now, it's like a child behind the wheel. But it can learn, and learn fast. Look at our world of technology and computer processing - now just go back 20 years. Heck, there is more computing power in your Apple iWatch than the entire Apollo moon mission. I'm not arguing that FSD is perfect right now. What I'm trying to convey is that the sentiment that "FSD will NEVER work" is just not realistic given our technological advancements. It was just a few years ago we developed a car's ability stay at the same speed, and keep itself in the middle of the lane. Then we were able to slow down and speed up the car while doing so. Now my Tesla can get me from my house to work without much intervention. A few days ago, for the first time ever for me, I let FSD (10.9) go from my office to my house (about 20 miles) with FSD all the way. It started on city streets for a few blocks, got onto the freeway, exited the freeway and went about 10 blocks to my house. It struggled a little on a few turns, made a few odd lane changes, which it just changed back. But overall it did it completely, and safely with no disengagements from me. Yes, it may have freaked out some drivers behind me (heck, it even had me cringe a few times - it gets awfully close to parked cars on residential streets) - which is why I have two large bumper stickers warning people behind me that I'm a self-driving test vehicle and they should use caution around me.
As you can read from my comments, I prefer logical arguments. I try to look at things from 30,000 feet - the big picture. So many people are black and white in their approach. "This will never work!" "It works perfectly, you're clearly an idiot" "Anyone that buys FSD is a moron and should have their licenses revoked" "Why does it still utterly suck at driving" So many passionate views, but not looking at the big picture.
Does FSD work perfectly right now? No. It's getting better - we can see the progress. Will FSD eventually work? Yes, very likely technology will catch up and allow humans to do other things with their time, while saving lives. Had one too many drinks at a company party? No worries, your car will get you home safely - someday. And the rate our technology is advancing, that someday will be sooner than later.
Just check out YouTube videos on the progress from FSD 9.0 to 10.10 (which is about 19 months). Here is a guy's channel that you can search for the various versions and see the progress over time:
I love this exchange from the TV show "The West Wing", where the chief of staff, Leo, is trying to convince the president, Bartlet, that the missile defense shield will work. And the president thinks it will never work.
President Josiah Bartlet : I want to know, when you're pushing me toward the missile shield, it's not 'cause you want me to look strong on defense?
Leo McGarry : I'm not.
President Josiah Bartlet : Don't.
Leo McGarry : I'm pushing you toward the shield 'cause I think it works.
President Josiah Bartlet : Based on what?
Leo McGarry : Confidence. And the understanding that there has been a time in the evolution of everything that works when it didn't work.
As you can read from my comments, I prefer logical arguments. I try to look at things from 30,000 feet - the big picture. So many people are black and white in their approach. "This will never work!" "It works perfectly, you're clearly an idiot" "Anyone that buys FSD is a moron and should have their licenses revoked" "Why does it still utterly suck at driving" So many passionate views, but not looking at the big picture.
Does FSD work perfectly right now? No. It's getting better - we can see the progress. Will FSD eventually work? Yes, very likely technology will catch up and allow humans to do other things with their time, while saving lives. Had one too many drinks at a company party? No worries, your car will get you home safely - someday. And the rate our technology is advancing, that someday will be sooner than later.
Just check out YouTube videos on the progress from FSD 9.0 to 10.10 (which is about 19 months). Here is a guy's channel that you can search for the various versions and see the progress over time:
I love this exchange from the TV show "The West Wing", where the chief of staff, Leo, is trying to convince the president, Bartlet, that the missile defense shield will work. And the president thinks it will never work.
President Josiah Bartlet : I want to know, when you're pushing me toward the missile shield, it's not 'cause you want me to look strong on defense?
Leo McGarry : I'm not.
President Josiah Bartlet : Don't.
Leo McGarry : I'm pushing you toward the shield 'cause I think it works.
President Josiah Bartlet : Based on what?
Leo McGarry : Confidence. And the understanding that there has been a time in the evolution of everything that works when it didn't work.
Last edited:
COS Blue
Member
I can respect your way of thinking. My big picture 30,000 view though is a bit different from yours.
- Autopilot has been around for around 8 years
- It still is known to slam on the brakes when there's a shadow in the road
- It still is known to veer onto the off-ramp when it should be staying in the right lane
- It still sometimes slows the car down because it thinks the speed limit is 45 when it passes a Speed Limit 65 sign.
- It still sometimes hits the brakes when a car is in the lane next to you because it is close to, but not over the lane line
- It still sometimes hits the brakes when there is an oncoming vehicle which is clearly in its own lane
- It can't consistently figure out when to turn the highbeams on/off (ok maybe not technically autopilot, but still)*
- It has been known to crash into emergency vehicles
- It smashed right into the fake pedestrians in the AAA test.
Doesn't necessarily mean it'l never work. But I don't think Full Self Driving is even close to reality.
There are definitely problems. I have experienced a few of them myself. And we could go back and forth on failures and successes. For every failure situation, I'm sure I could find one where it handled that same situation just fine. Even the Washington Post article circulating today, in their video showing the Model 3 hitting the fake peds in the AAA test, later in that video it showed the European testing worked perfectly.
Let's call attention to problems that need to be fixed, but let's also celebrate success when it's there. I've had to disengage my FSD many times, and still scratch my head from time to time on why it chose to do whatever it did at the time. I sometimes feel sorry for the drivers behind me. But I also have great joy when FSD takes me from one location to another smoothly and safely.
I'm reminded of a statistical joke: "Three statisticians go deer hunting with bows and arrows. they spot a big buck and take aim. One shoots and his arrow flies off three meters to the right. The second shoots and his arrow flies off three meters to the left. The third statistician jumps up and down yelling: I hit him!"
We all see failures in FSD - but we also see successes. Let's all try to meet in the middle and hope for the best - stay positive about the future of FSD.
The other day, I asked my MYLR/FSD to take me about two miles from one store to another. The route had two turns. At the first turn, FSD got into the right lane, put on its blinker, and then drove straight past the turn. That's one of those 6 basic functions - turn right - and it failed.
No one should be shocked by this, we know that FSD won't always be successful. But making a right turn successfully is a pretty basic function of FSD. And it does make quite a few of them successfully. But not all.
FSD is still working on how to handle those basic six functions. It's not even close to being able to evaluate the subtleties and nuances that we deal with every day while driving. I agree that it will get there eventually, but I think it has a very long way to go.
[Note: When I started writing this post, I had a lot more to say, but then realized that this is drifting pretty far from the thread's topic. Maybe it should be continued in its own thread.]
Similar threads
