There are many videos of FSD on UA-cam called “FSD is amazing!”, but your video is the most realistic, and shows the challenges that automated driving must overcome. It's very interesting. Thank you. I'm looking forward to the future evolution of FSD.
Great video! Certainly it looks like NY high traffic situations are too difficult for FSD now but actually I'm worried just about a single disengagement, the one at 5:25 - what's most important now is safety and other interventions were not safety related. Comfort can be improved while the robotaxi network is already deployed and this will boil down to the cost of remote control drivers, which Tesla would definitely need, and user satisfaction versus revenue (they will definitely want to get to much lower chance of comfort intervention rate than what we saw in this video). Anyway, an incorrect decision such as at 5:25 could lead to a safety problem when on another road type where the opposite traffic is faster moving. In such situation I think it is paramount for FSD to be able to move to a safe location and avoid high speed incoming traffic. It looks like Tesla focuses entirely on avoiding getting to dangerous positions, which is of course correct, but they should definitely acknowledge that some car *will* land on such situation eventually and it should be able to move even to an otherwise illegal place, like for example drive into a lawn or a sidewalk, in order to avoid accident. To do that I think they need to train the FSD on accidents data, when another car approaches fast and I'm not sure they are doing this.
Yeah, having the car turned into the opposite lane wasn’t definitely not the right thing to do and needed a disengagement. Other ones in the video were definitely NJ drivers assertiveness lol. But with 12.5.1.3 it took too long to stop for a pedestrian crossing and I had to stop myself. I’m sure FSD wouldn’t have hit the pedestrian but it might have stopped a little too close had I let it figure it out.
"I am unable to reconcile Tesla's claims regarding V12.5.2. Their assertion of a ~3x improvement in miles between necessary interventions is not substantiated by my findings. This video serves as compelling evidence to the contrary."
Yeah and Tesla should let us know what they consider an intervention so we can actually see if that’s really the case. Recorded another video of 12.5.2 and I still have to press on the accelerator multiple times.
Every time i watch your car, i just feel like its a completely different brand of car running on a different model of fsd compare to some of the other OG testers. I don't doubt your honesty by any stretch of my imagination, I'm just confused why is your car behaving so differently, TimDoes performed a few rush hour rides around las Vegas, pretty tough drives and his car nailed it. Wholemars same thing, TehcGeeks HW3 did incredibly well in Chicago rush hour traffic. sorry i did not mean to piss anybody off but i feel something is wrong with the car. reset DAS ,cameras and recalibrate cameras and then do a few drives (of course if this hasn't been done before). thanks again for your efforts and making these videos.
because FSD is garbage and very unreliable...man from mars is an outed FSD fluffer who makes money from favourable reviews..ever wonder why that clown from San Fran never seems to drive in fog?? my experience is like "just's"
You try driving in NJ, we got inpatient people and cars coming from every direction, it works very smoothly but some of these intersections it’s gonna get confused, mx 4
Youre getting honked at around the 2 minute mark…. Do they actually want you to push out into traffic thats flowing? Id be pissed if my car did that. FSD is making the right decision here.
Good video -- this is in stark-contrast to Whole Mars Catalog videos where he has no interventions. I think the difference is due to regional driving style differences. I live in CA and understand why FSD appears to work better there -- the ROAD Design and marking are much better and in general the drivers are more 'chill' . On the East-Coast and particularly in NJ -- the roads are older and the drivers are considerably more aggressive.To make FSD successful, TESLA will need to pay more attention to FSD Operation in the North East.
I think this is a very well written comment. I agree 100%. I dont understand why FSD can drive in vegas, san francisco even new york with almost 0 disengagement. It could be because there are teslas in those cities and therefore can collect more data? I dont know. but yeah good to see there are still some improvements to be made
@@Frenchy78180 - Thanks -- I am suspect of the zero-disengagement videos and have very-closely studied them to see if I can find edits or gaps. I could see nothing that would indicate they modified or staged -- it just appears that FSD works really well in those environments. Your point about available 'training data' may be the difference. I live in the heart of Silicon Valley where maybe +20% of the vehicles are TESLAs.
It's been documented tesla gives slightly different builds to some youtubers. They should stop because it gives false impressions of actual fsd. I have 2024.5.2, same as this guy. Most drives are perfect, but some need multiple disengagements and/or interventions.
@@TheRobertlonskiI agree. That’s why it seems like videos from the west coast area do so well. Which can explain why aren’t there any Waymo taxis here in NJ/NY, it’s just so much harder with roads and the driving style here.
Removing a waypoint and having the car reroute in the middle of a turn is not the best idea. That's why FSD hesitated at 16:00. Overall it performed decently in heavy traffic.
Yeah sometimes I wait too late to remove the location. Which is on me. But sometimes if I remove it too early, it changes the whole route that I thought it would take. Will try to do better on removing the stops earlier, but not too early
Thanks. At this point videos with several disengagements are actually more interesting than the "zero intervention" ones. Also like that you report these situations to the Tesla team.
Good to see that version struggle. I’m one version below (12.5.1.5) with AI3 in Los Angeles, and it is far from a relaxing experience. As others will say, I love the Tesla FSD team and the work that they do. But having ridden Waymo and seeing it slowly progress further and further into more cities, it’s putting pressure on my expectations for FSD.
@@JustAnotherTeslaGuy Agreed. I'm holding FSD to the Waymo standard that I am experiencing in Los Angeles. And it still has a long way to go. But I'm a Tesla investor so I must keep the faith.
At 20:28. Wasnt that a little harsh. You could argue that going into that intersection is not really the safe thing to do. If you're backed up too far back the car basicly has no escape.
Yup, you're right. I am totally for waiting it out to make sure we're clear before we move forward. However, at that intersection though, NJ drivers are going to make that left turn even if there was a tiny gap lol. At that point, I could tell there was enough space for FSD to go and make it in (the cars behind me follow in as well), so stopping or slowing down where FSD did wasn't good.
This channel - both the video and comments - should be seen by every $TSLA investor to evaluate the progress of FSD and the wisdom of their investment thesis.
My car has same issues.. regressions.. car stops at some stop lights then doesn't go when green sometimes. Sometimes making left turn it is smooth then jerks for no reason when completing turn. Sometimes when people on side of road, car just stops in middle of street. I really don't understand what happened. 12.5.1.3 was fantastic. 12.5.2 has gotten better slowing at stops and drives better in parking lots. Highway driving has also improved.
Yeah your experience was what I experienced when FSD first came out. I haven’t had 12.5.2 that long yet. But the stopping at red light or stop sign for too long is something I’m experiencing
There's no way FSD will ever be able to merge into fast oncoming traffic in NJ.. I get stressed out all the time as human driver because you have to seize that split few second opportunity and no way FSD will be able to make that type of decisive decision safely..
Not true. Fsd has made several very quick merges that suprised me because i realized i couldnt have made the same merges. Car was on auto speed. I realized the car had advantage of seeing in all directions at once and made merges that I couldnt see cause I was looking at other fast moving traffic.
I don’t know why the comments are pessimistic about FSD… 😅 I watched many videos of yours but FSD was quite great and improving. These are high traffic jam situation and intervention situations are not necessary in the video. It’s just the driver’s taste. Maybe many other drivers might feel dangerous to squeeze in those small room between cars. Everyone has their own driving taste, and even in reality many passengers aren’t comfortable with some uber drivers. I think FSD will be servicable in next Q2 definitely (actually, with little sacrifice and remote control team, in present it is possible also). Well I know many of you guys are mad because like whole mars catalog tsla influencers definitely shows very easy situations for FSD, but after monitoring years of progress in FSD, I’m being confident it is improving exponentially(with some little backward and forward).
Also, the mistakes that FSD is making in video is anyone who is not familiar with specific regions can do… I’m continuously thinking why that should be an intervention while watching video… FSD can overtake the front car and maybe turn back to the right lane. Currently, I think many users aren’t just comfortable with FSD cause it’s a wild and never seen before tech. So whether actually FSD succeed in those jobs, they aren’t just comfortable because that is not my driving style. I think many tesla investors should focus on marketing and people being familiar with FSD, because we all know that FSD is the no-comparable best autonomy tech compared to waymo or other companies.
And wtf is going in 16:00 ? You changed the destination while turning right… FSD is perfectly right to have responded like that… You are making intentional interventions. Suppose you are riding on uber and while turning right, you say “oh sorry my destination is behind us! Then uber driver will be hesitant to turn right and think whether he should go back to straight line or just turn right. That’s a hesitation moment for human as well .. maybe more longer if not skilled
I agree that most of the interventions from this video are not critical - if you are sitting in the backseat and minding your own business on your laptop then it doesn't really matter that much that the car will arrive at the destination a little bit later or that some decisions won't be perfect. But this is true only on the condition of the right safety level. In my opinion it looks like Tesla doesn't yet train FSD on accidents data and the car would not be able for example to quickly avoid a high speed incoming car by driving into a sidewalk or a lawn. Being able to correctly react to such difficult situations is necessary. Of course it's paramount to avoid getting into danger but the fact is that the cars **will** eventually end up in those situations.
@@soul_lovers4539 my thinking behind at the 16 mark was autopilot was completed was the pop up so I removed it. At the same time the route actually never changed so it just had to continue its turn. I’m trying to time it right so that the route doesn’t change when I remove a waypoint. Cause if I remove it too early, the route may change. So I try to wait till the last minute but I waited too long on some of them, so just gotta keep on trying.
hah, no worries! Tesla just does not use the training data from NJ yet! it is just about Tesla's FSD development priorities. Currently Tesla must consentrate on thoses cities where they think most worth accomplishing FSD
Thank you for your honest video reviews. Sometimes I feel that there must be something wrong with my own evaluation of FSD. I often see glowing reviews online, with testers singing endless praises (although sometimes there are multiple errors during their own drives). I received 12.5.2 a couple days ago. Took a 200+ mile round trip yesterday and had to disengage at least three times (sorry, I lost track). My first disengagement was about 10 minutes after leaving home. I was on a two-lane road traveling around 55mph when FSD hit the brakes abruptly (I wouldn’t be too far off if I said “violently”) and jerked left and right several times as if it was trying to avoid something in the road. (Fortunately there was no one directly behind me). There was nothing in or around the road except for some dark tire skid marks. That’s the only thing I could see, which may have confused FSD. Later in the trip I was approaching a stop light. Left turns were not allowed at the intersection. The map navigation displayed the correct route, which was to turn right and make a U-turn a couple hundred feet down the road. But FSD got in the straight-thru lane and turned on the left-turn signal, totally ignoring the mapped route, and attempted to make an illegal turn. There’s more, but geeesh, I don’t even want to think about it, except to say that speed management is still abysmal in 12.5.2. It’s a constant battle on two-lane roads to maintain speed; it keeps dropping, holding up traffic. I am dismayed. Bought my first Tesla in May of this year with the hope of seeing full autonomy in the not too distant future. Don’t get me wrong, I think that FSD is amazing for what it does well!! But there are way too many things it does poorly, and sometimes egregiously. I keep hoping to see major progress. But even if FSD is not fully autonomous in my lifetime (I’m 77), I would be happy if it just got anywhere close to achieving it. … And yes, to answer your question, personally I consider a pedal press an intervention. Sometimes interventions are matters of preference, but at other times FSD is not doing its job.
@@VasuJaganath I’m not yet ready to give up hope. I believe that FSD may very well eventually achieve full autonomy. I just think that the challenges are being way underestimated. But that is no reason to give up. It may or may not happen in my lifetime, but it will happen. This is simply my opinion, and there is so much I do not know. I remain hopeful.
@@fedderdthank you for your support on my videos! That phantom braking is annoying and makes us feel uncomfortable which I still experience myself when I’m using it and not recording. I have also seen it not follow the map route and it needed to reroute. So what’s the point of the route on the map? Same with the speed. The limit was 55 on a two lane freeway and it was constantly going at 45. And yeah, despite all this. I still wouldn’t get a car with out FSD because the things it can do, it’s already amazing. This wasn’t something any car could do a few years ago. So I do hope that FSD will reach level 5 and we can all have our cars drive us everywhere unsupervised. I use to think accelerator presses were not an intervention but because of how often I’m pressing it now. It feels like it is lol.
I guess self driving is never there... I wait every 3 weeks just to hear you say negative things about it.. I’m gonna check your channel in 1 month lol just to hear the same problem 🙄🙄🙄
first i m gonna say i watched the whole video, and you test is probably the most difficult one of all FSD testing videos, i love it that you put FSD through this kind of tough situation. But, but you really don't know how to drive, do you? You care to much about people honking at you even when FSD is doing the right thing, like that slow-down when turning left, it's because FSD saw the traffic is jammed up and did not want to block the intersection, it's the right thing to do, and then you pushed the peddle for it to block the intersection when theres no space to go. And the first shot in your video when you took over to cut the traffic in front of you to turn left, that was one of the most assholey thing to do in traffic, you should wait till all the traffic clears and then turn left, the people honking at you is clearly an asshole, why would you even care about him? This whole video is like, FSD always trying to do the right thing, and you kept pushing it to do the wrong thing. You just wasted a lot of good training data.
Thanks again for your support and glad you enjoyed these tests. I agree with you that FSD was doing the right thing in making sure we are clear to make that first left turn. But at the same time as NJ driver, people expect you move as soon as there’s a gap. For example the human driver that was ahead of me took the space and went for it. Also yes, while the person that was honking me may have been an asshole for doing that, I don’t want to incite any road rage incidents if I can avoid it. If me moving forward and completing the turn was all it takes, then I’ll do that. Additionally, at that moment the cars on the left aren’t really able to move forward anymore, so with that gap it was fine to go through. On that last part of the left turn into the intersection - I am totally for waiting it out to make sure we're clear before we move forward. However, at that intersection though (being there all the time), drivers are going to make that left turn even if there was a tiny gap. At that point, I could tell there was enough space for FSD to go and make it in (the cars behind me follow in as well), so stopping or slowing down where FSD did wasn't good. All that to say, yes I agree with you the FSD is doing the right thing each time. But drivers here in NJ are aggressive and drive a certain way and expect other drivers to drive that way too. It may be a bit calmer in other places which why when FSD does the right thing, it works. FSD needs a NJ assertive mode.
again thanks for an honest, real driver (not economically ties to Tesla) review..I just saw somewhere where Tesla is touting something about 3x improvement in "necessary interventions"...clown car company, as everyone of my interventions is necessary..ASS burger stuff, does Tesla honestly believe FSD owners drive around making "uncessary interventions"..what a clown car company, this suggests to me they are very defensive about the state of FSD..PS Tesla should specifically define what a necessary intervention is, and IMO it is independent of FSD disengagement, as it is a pain to reingage.
@lor8ancg thanks again! I just repot and show FSD as it. I want it to do great and drive me everywhere unsupervised, but if it doesn’t work, I’ll be sure to show that. Yeah, hopefully Tesla can let us know what they consider an intervention. In that past, I used to think accelerator presses were not really an intervention but it’s happening so frequently now, that I feel like it is.
Haha love the way you think😂. NJ drivers are just too impatient and if they see a gap, they expect drivers to move. I don’t want to incite any road rage. If moving forward was it takes then I’ll do it to avoid trouble with other drivers.
@@JustAnotherTeslaGuy I totally get it. Also, the older I get, the less I care about saving a few seconds. Unless there's pizza or pasta involved, I'm not in a hurry. 🙂
Haha yeah. Driving in NJ seems like all the accidents were people going to fast here or not letting people cut in. Like if me being one car slower because I let someone in and that can avoid a headache insurance claims, then sure. I also get honked at when FSD goes slower because it see other drivers blinking to merge in.
There are many videos of FSD on UA-cam called “FSD is amazing!”, but your video is the most realistic, and shows the challenges that automated driving must overcome. It's very interesting. Thank you. I'm looking forward to the future evolution of FSD.
Thank you for your support! Will continue to make more videos of these rush hour traffic videos
Great video! Certainly it looks like NY high traffic situations are too difficult for FSD now but actually I'm worried just about a single disengagement, the one at 5:25 - what's most important now is safety and other interventions were not safety related. Comfort can be improved while the robotaxi network is already deployed and this will boil down to the cost of remote control drivers, which Tesla would definitely need, and user satisfaction versus revenue (they will definitely want to get to much lower chance of comfort intervention rate than what we saw in this video). Anyway, an incorrect decision such as at 5:25 could lead to a safety problem when on another road type where the opposite traffic is faster moving. In such situation I think it is paramount for FSD to be able to move to a safe location and avoid high speed incoming traffic. It looks like Tesla focuses entirely on avoiding getting to dangerous positions, which is of course correct, but they should definitely acknowledge that some car *will* land on such situation eventually and it should be able to move even to an otherwise illegal place, like for example drive into a lawn or a sidewalk, in order to avoid accident. To do that I think they need to train the FSD on accidents data, when another car approaches fast and I'm not sure they are doing this.
Yeah, having the car turned into the opposite lane wasn’t definitely not the right thing to do and needed a disengagement. Other ones in the video were definitely NJ drivers assertiveness lol. But with 12.5.1.3 it took too long to stop for a pedestrian crossing and I had to stop myself. I’m sure FSD wouldn’t have hit the pedestrian but it might have stopped a little too close had I let it figure it out.
"I am unable to reconcile Tesla's claims regarding V12.5.2. Their assertion of a ~3x improvement in miles between necessary interventions is not substantiated by my findings. This video serves as compelling evidence to the contrary."
Yeah and Tesla should let us know what they consider an intervention so we can actually see if that’s really the case. Recorded another video of 12.5.2 and I still have to press on the accelerator multiple times.
Every time i watch your car, i just feel like its a completely different brand of car running on a different model of fsd compare to some of the other OG testers. I don't doubt your honesty by any stretch of my imagination, I'm just confused why is your car behaving so differently, TimDoes performed a few rush hour rides around las Vegas, pretty tough drives and his car nailed it. Wholemars same thing, TehcGeeks HW3 did incredibly well in Chicago rush hour traffic. sorry i did not mean to piss anybody off but i feel something is wrong with the car. reset DAS ,cameras and recalibrate cameras and then do a few drives (of course if this hasn't been done before). thanks again for your efforts and making these videos.
because FSD is garbage and very unreliable...man from mars is an outed FSD fluffer who makes money from favourable reviews..ever wonder why that clown from San Fran never seems to drive in fog?? my experience is like "just's"
well, thats the point of FSD. It is very unpredictable at this moment. You can use it on the same road 5x and it may act different on the 6th time.
You try driving in NJ, we got inpatient people and cars coming from every direction, it works very smoothly but some of these intersections it’s gonna get confused, mx 4
what happened at 5:27, happens to me all the time
Youre getting honked at around the 2 minute mark…. Do they actually want you to push out into traffic thats flowing? Id be pissed if my car did that. FSD is making the right decision here.
When FSD can cleanly handle NYC and North Jersey roads, then I'll know it's ready.
Yeah exactly. That’s why there aren’t any Waymos here lol
@@JustAnotherTeslaGuy😂
12.5.2 has not been good for me. Phantom (panic) braking has been the main issue. It feels like a smaller dumber model with less confidence.
Good video -- this is in stark-contrast to Whole Mars Catalog videos where he has no interventions. I think the difference is due to regional driving style differences. I live in CA and understand why FSD appears to work better there -- the ROAD Design and marking are much better and in general the drivers are more 'chill' . On the East-Coast and particularly in NJ -- the roads are older and the drivers are considerably more aggressive.To make FSD successful, TESLA will need to pay more attention to FSD Operation in the North East.
I think this is a very well written comment. I agree 100%. I dont understand why FSD can drive in vegas, san francisco even new york with almost 0 disengagement. It could be because there are teslas in those cities and therefore can collect more data? I dont know. but yeah good to see there are still some improvements to be made
@@Frenchy78180 - Thanks -- I am suspect of the zero-disengagement videos and have very-closely studied them to see if I can find edits or gaps. I could see nothing that would indicate they modified or staged -- it just appears that FSD works really well in those environments. Your point about available 'training data' may be the difference. I live in the heart of Silicon Valley where maybe +20% of the vehicles are TESLAs.
It's been documented tesla gives slightly different builds to some youtubers. They should stop because it gives false impressions of actual fsd. I have 2024.5.2, same as this guy. Most drives are perfect, but some need multiple disengagements and/or interventions.
@@TheRobertlonskiI agree. That’s why it seems like videos from the west coast area do so well. Which can explain why aren’t there any Waymo taxis here in NJ/NY, it’s just so much harder with roads and the driving style here.
Removing a waypoint and having the car reroute in the middle of a turn is not the best idea. That's why FSD hesitated at 16:00. Overall it performed decently in heavy traffic.
that was not a good ide indeed
Yeah sometimes I wait too late to remove the location. Which is on me. But sometimes if I remove it too early, it changes the whole route that I thought it would take. Will try to do better on removing the stops earlier, but not too early
Thanks. At this point videos with several disengagements are actually more interesting than the "zero intervention" ones. Also like that you report these situations to the Tesla team.
Yeah seeing it do well is great and all but I want to see what happens if does get into situations where it is a bit more complex
Good to see that version struggle. I’m one version below (12.5.1.5) with AI3 in Los Angeles, and it is far from a relaxing experience. As others will say, I love the Tesla FSD team and the work that they do. But having ridden Waymo and seeing it slowly progress further and further into more cities, it’s putting pressure on my expectations for FSD.
It is amazing how far it has come. But it still feel like human supervision will be needed for a much longer period
@@JustAnotherTeslaGuy Agreed. I'm holding FSD to the Waymo standard that I am experiencing in Los Angeles. And it still has a long way to go. But I'm a Tesla investor so I must keep the faith.
Same. I want it do well and drive me unsupervised everywhere. But if it can’t do that here for me in NJ, I’ll be sure to show everyone
At 20:28. Wasnt that a little harsh. You could argue that going into that intersection is not really the safe thing to do. If you're backed up too far back the car basicly has no escape.
Yup, you're right. I am totally for waiting it out to make sure we're clear before we move forward.
However, at that intersection though, NJ drivers are going to make that left turn even if there was a tiny gap lol. At that point, I could tell there was enough space for FSD to go and make it in (the cars behind me follow in as well), so stopping or slowing down where FSD did wasn't good.
What do you expect from NJ drivers. They have zero tolerance and expect even the smallest gap to be taken. FSD def needs to be more aggressive
This channel - both the video and comments - should be seen by every $TSLA investor to evaluate the progress of FSD and the wisdom of their investment thesis.
Thanks for the compliment and support!
@@VasuJaganath But there might be wisdom in the inverse-$TSLA thesis...
Good finds. 5:19 not the easiest to spot that they weren't parked cars. interesting edge case. Several interesting cases
I thought they were parked as well when watching this video.
I’m used to driving here so I know they weren’t parked cars. Hopefully my message can help make FSD better!
That's 3x disengagement improvement claimed by TeslaAI. /s
My car has same issues.. regressions.. car stops at some stop lights then doesn't go when green sometimes. Sometimes making left turn it is smooth then jerks for no reason when completing turn. Sometimes when people on side of road, car just stops in middle of street. I really don't understand what happened. 12.5.1.3 was fantastic. 12.5.2 has gotten better slowing at stops and drives better in parking lots. Highway driving has also improved.
Yeah your experience was what I experienced when FSD first came out. I haven’t had 12.5.2 that long yet. But the stopping at red light or stop sign for too long is something I’m experiencing
There's no way FSD will ever be able to merge into fast oncoming traffic in NJ.. I get stressed out all the time as human driver because you have to seize that split few second opportunity and no way FSD will be able to make that type of decisive decision safely..
Not true. Fsd has made several very quick merges that suprised me because i realized i couldnt have made the same merges. Car was on auto speed. I realized the car had advantage of seeing in all directions at once and made merges that I couldnt see cause I was looking at other fast moving traffic.
FSD definitely can, but it will feel much uncomfortable. It takes time for people to fully trust it
As they add more compute with AI4 and AI5, it will…. It needs more parameter count.
I don’t know why the comments are pessimistic about FSD… 😅 I watched many videos of yours but FSD was quite great and improving. These are high traffic jam situation and intervention situations are not necessary in the video. It’s just the driver’s taste. Maybe many other drivers might feel dangerous to squeeze in those small room between cars. Everyone has their own driving taste, and even in reality many passengers aren’t comfortable with some uber drivers. I think FSD will be servicable in next Q2 definitely (actually, with little sacrifice and remote control team, in present it is possible also). Well I know many of you guys are mad because like whole mars catalog tsla influencers definitely shows very easy situations for FSD, but after monitoring years of progress in FSD, I’m being confident it is improving exponentially(with some little backward and forward).
Also, the mistakes that FSD is making in video is anyone who is not familiar with specific regions can do… I’m continuously thinking why that should be an intervention while watching video… FSD can overtake the front car and maybe turn back to the right lane. Currently, I think many users aren’t just comfortable with FSD cause it’s a wild and never seen before tech. So whether actually FSD succeed in those jobs, they aren’t just comfortable because that is not my driving style. I think many tesla investors should focus on marketing and people being familiar with FSD, because we all know that FSD is the no-comparable best autonomy tech compared to waymo or other companies.
And wtf is going in 16:00 ? You changed the destination while turning right… FSD is perfectly right to have responded like that… You are making intentional interventions. Suppose you are riding on uber and while turning right, you say “oh sorry my destination is behind us! Then uber driver will be hesitant to turn right and think whether he should go back to straight line or just turn right. That’s a hesitation moment for human as well .. maybe more longer if not skilled
I agree that most of the interventions from this video are not critical - if you are sitting in the backseat and minding your own business on your laptop then it doesn't really matter that much that the car will arrive at the destination a little bit later or that some decisions won't be perfect. But this is true only on the condition of the right safety level. In my opinion it looks like Tesla doesn't yet train FSD on accidents data and the car would not be able for example to quickly avoid a high speed incoming car by driving into a sidewalk or a lawn. Being able to correctly react to such difficult situations is necessary. Of course it's paramount to avoid getting into danger but the fact is that the cars **will** eventually end up in those situations.
@@piotrsuszynski9273 well great insight! Thanks for sharing your opinion!
@@soul_lovers4539 my thinking behind at the 16 mark was autopilot was completed was the pop up so I removed it. At the same time the route actually never changed so it just had to continue its turn.
I’m trying to time it right so that the route doesn’t change when I remove a waypoint. Cause if I remove it too early, the route may change. So I try to wait till the last minute but I waited too long on some of them, so just gotta keep on trying.
For the love of god just pick a location and let it go. The person in the back of the robo taxi won't be changing routes ten times.
Yup, totally right. I just want to push FSD where there is more traffic and see how it performs, so I try to find the busiest routes.
The first 2 minutes are so confusing, the guy is honking at you for what? You can bulldoze your way through but traffic in front has right of way.
Thought I was the only one 12.5.2 is a regression from 12.5.1.3 . I noticed so many hesitation and slowness
I have seen more hesitation with 12.5.2 as well.
hah, no worries! Tesla just does not use the training data from NJ yet! it is just about Tesla's FSD development priorities. Currently Tesla must consentrate on thoses cities where they think most worth accomplishing FSD
Needs to be a "city aggressive" mode 😂
Maybe a NJ state aggressive mode lol
did you use a hardware 3 or hardware 4 car?
Looks like a new model 3 w/ hw4. You can tell by the new steering wheel.
Yup, I’m using a 2024 model 3 with hardware 4 (AI4)
Thank you for your honest video reviews. Sometimes I feel that there must be something wrong with my own evaluation of FSD. I often see glowing reviews online, with testers singing endless praises (although sometimes there are multiple errors during their own drives).
I received 12.5.2 a couple days ago. Took a 200+ mile round trip yesterday and had to disengage at least three times (sorry, I lost track). My first disengagement was about 10 minutes after leaving home. I was on a two-lane road traveling around 55mph when FSD hit the brakes abruptly (I wouldn’t be too far off if I said “violently”) and jerked left and right several times as if it was trying to avoid something in the road. (Fortunately there was no one directly behind me). There was nothing in or around the road except for some dark tire skid marks. That’s the only thing I could see, which may have confused FSD.
Later in the trip I was approaching a stop light. Left turns were not allowed at the intersection. The map navigation displayed the correct route, which was to turn right and make a U-turn a couple hundred feet down the road. But FSD got in the straight-thru lane and turned on the left-turn signal, totally ignoring the mapped route, and attempted to make an illegal turn.
There’s more, but geeesh, I don’t even want to think about it, except to say that speed management is still abysmal in 12.5.2. It’s a constant battle on two-lane roads to maintain speed; it keeps dropping, holding up traffic. I am dismayed. Bought my first Tesla in May of this year with the hope of seeing full autonomy in the not too distant future. Don’t get me wrong, I think that FSD is amazing for what it does well!! But there are way too many things it does poorly, and sometimes egregiously. I keep hoping to see major progress. But even if FSD is not fully autonomous in my lifetime (I’m 77), I would be happy if it just got anywhere close to achieving it.
… And yes, to answer your question, personally I consider a pedal press an intervention. Sometimes interventions are matters of preference, but at other times FSD is not doing its job.
@@VasuJaganath I’m not yet ready to give up hope. I believe that FSD may very well eventually achieve full autonomy. I just think that the challenges are being way underestimated. But that is no reason to give up. It may or may not happen in my lifetime, but it will happen. This is simply my opinion, and there is so much I do not know. I remain hopeful.
@@fedderdthank you for your support on my videos! That phantom braking is annoying and makes us feel uncomfortable which I still experience myself when I’m using it and not recording.
I have also seen it not follow the map route and it needed to reroute. So what’s the point of the route on the map?
Same with the speed. The limit was 55 on a two lane freeway and it was constantly going at 45.
And yeah, despite all this. I still wouldn’t get a car with out FSD because the things it can do, it’s already amazing. This wasn’t something any car could do a few years ago. So I do hope that FSD will reach level 5 and we can all have our cars drive us everywhere unsupervised.
I use to think accelerator presses were not an intervention but because of how often I’m pressing it now. It feels like it is lol.
I guess self driving is never there... I wait every 3 weeks just to hear you say negative things about it.. I’m gonna check your channel in 1 month lol just to hear the same problem 🙄🙄🙄
well, I hope it does get better lol. I’ll definitely make more videos then, stay tuned!
ロボタクシーはまだまだ先ですね
Graham Hollow
first i m gonna say i watched the whole video, and you test is probably the most difficult one of all FSD testing videos, i love it that you put FSD through this kind of tough situation. But, but you really don't know how to drive, do you? You care to much about people honking at you even when FSD is doing the right thing, like that slow-down when turning left, it's because FSD saw the traffic is jammed up and did not want to block the intersection, it's the right thing to do, and then you pushed the peddle for it to block the intersection when theres no space to go. And the first shot in your video when you took over to cut the traffic in front of you to turn left, that was one of the most assholey thing to do in traffic, you should wait till all the traffic clears and then turn left, the people honking at you is clearly an asshole, why would you even care about him? This whole video is like, FSD always trying to do the right thing, and you kept pushing it to do the wrong thing. You just wasted a lot of good training data.
Thanks again for your support and glad you enjoyed these tests.
I agree with you that FSD was doing the right thing in making sure we are clear to make that first left turn. But at the same time as NJ driver, people expect you move as soon as there’s a gap. For example the human driver that was ahead of me took the space and went for it. Also yes, while the person that was honking me may have been an asshole for doing that, I don’t want to incite any road rage incidents if I can avoid it. If me moving forward and completing the turn was all it takes, then I’ll do that. Additionally, at that moment the cars on the left aren’t really able to move forward anymore, so with that gap it was fine to go through.
On that last part of the left turn into the intersection - I am totally for waiting it out to make sure we're clear before we move forward.
However, at that intersection though (being there all the time), drivers are going to make that left turn even if there was a tiny gap. At that point, I could tell there was enough space for FSD to go and make it in (the cars behind me follow in as well), so stopping or slowing down where FSD did wasn't good.
All that to say, yes I agree with you the FSD is doing the right thing each time. But drivers here in NJ are aggressive and drive a certain way and expect other drivers to drive that way too. It may be a bit calmer in other places which why when FSD does the right thing, it works. FSD needs a NJ assertive mode.
@@JustAnotherTeslaGuy lol NJ assertive mode
again thanks for an honest, real driver (not economically ties to Tesla) review..I just saw somewhere where Tesla is touting something about 3x improvement in "necessary interventions"...clown car company, as everyone of my interventions is necessary..ASS burger stuff, does Tesla honestly believe FSD owners drive around making "uncessary interventions"..what a clown car company, this suggests to me they are very defensive about the state of FSD..PS Tesla should specifically define what a necessary intervention is, and IMO it is independent of FSD disengagement, as it is a pain to reingage.
Sounds like you should buy a competing car company's FSD.
Pathetic cringe human. 😑
@lor8ancg thanks again! I just repot and show FSD as it. I want it to do great and drive me everywhere unsupervised, but if it doesn’t work, I’ll be sure to show that.
Yeah, hopefully Tesla can let us know what they consider an intervention. In that past, I used to think accelerator presses were not really an intervention but it’s happening so frequently now, that I feel like it is.
Why do you care about being honked at? If you honk at me, we're both staying there for another 30 seconds.
ahah
Haha love the way you think😂. NJ drivers are just too impatient and if they see a gap, they expect drivers to move. I don’t want to incite any road rage. If moving forward was it takes then I’ll do it to avoid trouble with other drivers.
@@JustAnotherTeslaGuy I totally get it. Also, the older I get, the less I care about saving a few seconds. Unless there's pizza or pasta involved, I'm not in a hurry. 🙂
Haha yeah. Driving in NJ seems like all the accidents were people going to fast here or not letting people cut in. Like if me being one car slower because I let someone in and that can avoid a headache insurance claims, then sure. I also get honked at when FSD goes slower because it see other drivers blinking to merge in.
This video seems very Dan O' Doddish to me
Just trying to show my experience with FSD here in NJ.