James Locke makes videos of FSD in the suburbs, and even the latest software continues to behave abnormally. The beta software is not worthy of trust yet, and maybe never.
Of course I wouldn't feel confident sitting in the back seat while it drives me around for one hour downtown. But for a few minutes if I know it's mostly straight and the traffic not too bad, maybe it wouldn't be too dangerous. As a driver assist, it's pretty good now.
@@freddydad1 you're wrong. The worst FSD would do on its own is something really bad. The long tail of dangerous rare situations never stop coming at you, and much more often in a busy city. Most of the time you don't see them, but then one pops up if you drive enough. Someone jumps out at you, a biker falls off right in front of you, someone cuts you off suddently, the low sun interferes with sensors when you're in heavy traffic, and endless other dangerous stuff. FSD drives around with no memory of the road network, so it doesn't know where the really bad intersections are that should be avoided, or where to slow down a little ahead of time. A complacent driver with no anticipation is not going to be safe over lots of miles. To be really safe it has to be reliable and robust over a million miles. For a large-scale robotaxi service, it has to be robust over hundreds of millions of miles.
Found you by chance and now subscribed. I'm a UK-based Tesla investor and it's good to see how the tech is shaping up given we don't have FSD yet over here. Disappointed to see it tuck in behind that stationary car and then have to go around, though I expect that will get resolved in the next few months - such is the pace of improvement. Thanks for having no music over your commentary - and your commentary itself is very useful for context. Great job!
Thanks for the kind words. I believe it should be coming to Europe soon. I’m actually very interested to see how it works with the roads there, versus traditional grid systems we see here in North America. The technology has definitely improved significantly since I first acquired it about a year and a half ago. Hoping it continues! Your observation is spot on.
That moving into turn lane seems baked in, you'd think if it sees a car parked it would just stay in lane until clear. Overall it seems tesla could offer a waymo type service now in every city in us/canada. Impressive technology. Great vid and commentary.
I think If the camera would've seen the car, it would have waited (hopefully). It was trying to get ahead of the turn... these are the conditions it needs to master before it becomes truly autonomous. Thanks for watching!
At 3:30 ish, the cameras MUST have seen the car parked in the left lane, yes? If so it's just a problem of the Tesla understanding the "logic" of lane selection on approaching a turn? (an ongoing problem... or is that better with 12.5.x?) Thoughts?
I'm not sure if the truck (size) blocked it from seeing that car on the left. It had to turn immediately after that car, so it's trying to calculate when it should move over to the left vs the parked car's actual position. I would think it's logic should first determine that there is a car there, and make adjustments to immediately turn left after passing the car. It did do that when i reactivated the FSD with the car in plain sight. Always room for improvement.
You're right that urban-center driving like this is the best FSD test. More challenges come per minute. But you still need to test like this over thousands of miles to get an idea of how good it is. A real robotaxi needs to be safe and effective over tens of millions of city miles to be a serious contender at scale.
Hopefully all the data collected from the thousands of FSD drivers on the road. will ge us there.... still a long ways to go, but much improved year over year.
@@freddydad1 I haven’t tried Waymo, but I’m curious to see how smooth it is vs FSD, particularly with the sensors (or maybe you just notice that for the safety aspect!) FSD is mostly a very smooth ride.
It’s much better than it used to be but it still sometimes can stay there until the end. The latest version coming out of FSD should finally merge city streets and Highway into a single stack, so hoping that will improve the merging performance.
From Europe's perspective this grid of parallel and perpendicular streets is still extremely easy for drivers. Don't you have any real raw urban environment there, just these carefully engineered safe-spaces? :)
@@peter.g6 North American street systems are definitely more grid-heavy! FSD in Europe or (some of) Asia would be a totally different experience! Thanks for watching.
Nice video. Two comments: At 3:58 better to see how the car handles the parked car in the left lane, rather than manually disengaging. Also, the screen in the car is rather low resolution and I like to follow the details normally. Thanks.
It would have had to adjust and get back in the lenr. I know it would have but I would have seriously affected some rush hour traffic. They shouldn’t have to suffer for this content! 😎
@@RealBenAnderson Yeah, I passed that some time ago. This is a niche community, so any and all support is appreciated. Just love sharing these videos, they're fun to do, and I love the technology.
2:17 The man near the roadside was not detected until after he was passed. Look at the display. So, huge ERROR. Two reasons, one he is bent over and have a hoodie, looks like a trash can. Two, he is only partially lit up which most often confuses the image recognition. One of the ten original and unsolveable issues with FSD. Honestly, I cannot get why virtually none of you FSD testers actually look at your videos reviewing incidents. I fully understand why you don't see the real issues on the fly when driving. However, the purpose of testing is understanding and contributing to improving the product, right? It's easy to see why like Whole Mars that makes a lot of money from hiding mistakes do extremely bad analysis. But most of you guys is mostly doin it for fun I guess.
I partially agree, and think sensors would help there,.. wondering about the lighting conditions and what he was wearing.. Im not sure a human driver would pick it up much faster, but I see your point. FSD has been very cautious historically when it sees people standing close to the road and facing towards the road... as he was not walking in that direction, it did not see a reason to emergency brake. I did want to highlight it. I have no interest in sugar coating FSD, I dont make a dime off these videos!
Downtown is where Tesla’s FSD shows what it can really do! Would you feel confident driving it downtown?
James Locke makes videos of FSD in the suburbs, and even the latest software continues to behave abnormally. The beta software is not worthy of trust yet, and maybe never.
Of course I wouldn't feel confident sitting in the back seat while it drives me around for one hour downtown. But for a few minutes if I know it's mostly straight and the traffic not too bad, maybe it wouldn't be too dangerous. As a driver assist, it's pretty good now.
@charliedoyle7824 The worst it's going to do is annoy other drivers. Humans do that constantly.
@@freddydad1 you're wrong. The worst FSD would do on its own is something really bad. The long tail of dangerous rare situations never stop coming at you, and much more often in a busy city. Most of the time you don't see them, but then one pops up if you drive enough. Someone jumps out at you, a biker falls off right in front of you, someone cuts you off suddently, the low sun interferes with sensors when you're in heavy traffic, and endless other dangerous stuff. FSD drives around with no memory of the road network, so it doesn't know where the really bad intersections are that should be avoided, or where to slow down a little ahead of time. A complacent driver with no anticipation is not going to be safe over lots of miles. To be really safe it has to be reliable and robust over a million miles. For a large-scale robotaxi service, it has to be robust over hundreds of millions of miles.
@@freddydad1 Human drivers are orders of magnitude superior to the simple minded Level 2 FSD software. There is no comparison .
Found you by chance and now subscribed. I'm a UK-based Tesla investor and it's good to see how the tech is shaping up given we don't have FSD yet over here. Disappointed to see it tuck in behind that stationary car and then have to go around, though I expect that will get resolved in the next few months - such is the pace of improvement. Thanks for having no music over your commentary - and your commentary itself is very useful for context. Great job!
Thanks for the kind words. I believe it should be coming to Europe soon. I’m actually very interested to see how it works with the roads there, versus traditional grid systems we see here in North America. The technology has definitely improved significantly since I first acquired it about a year and a half ago. Hoping it continues! Your observation is spot on.
when you were saying car was signaling right turn to early it was actually signaling the shift to the right lane as it should.
But it kept it on through the intersection. 😮
That moving into turn lane seems baked in, you'd think if it sees a car parked it would just stay in lane until clear. Overall it seems tesla could offer a waymo type service now in every city in us/canada. Impressive technology. Great vid and commentary.
I think If the camera would've seen the car, it would have waited (hopefully). It was trying to get ahead of the turn... these are the conditions it needs to master before it becomes truly autonomous. Thanks for watching!
At 3:30 ish, the cameras MUST have seen the car parked in the left lane, yes? If so it's just a problem of the Tesla understanding the "logic" of lane selection on approaching a turn? (an ongoing problem... or is that better with 12.5.x?) Thoughts?
I'm not sure if the truck (size) blocked it from seeing that car on the left. It had to turn immediately after that car, so it's trying to calculate when it should move over to the left vs the parked car's actual position. I would think it's logic should first determine that there is a car there, and make adjustments to immediately turn left after passing the car. It did do that when i reactivated the FSD with the car in plain sight. Always room for improvement.
You're right that urban-center driving like this is the best FSD test. More challenges come per minute. But you still need to test like this over thousands of miles to get an idea of how good it is. A real robotaxi needs to be safe and effective over tens of millions of city miles to be a serious contender at scale.
Hopefully all the data collected from the thousands of FSD drivers on the road. will ge us there.... still a long ways to go, but much improved year over year.
@@FSDwithme it's on par with waymo surely?
@@freddydad1 I haven’t tried Waymo, but I’m curious to see how smooth it is vs FSD, particularly with the sensors (or maybe you just notice that for the safety aspect!) FSD is mostly a very smooth ride.
How do you find the merging at highway speeds on an on-ramp?...I've had the problem of it staying in the merge lane right until almost the end
It’s much better than it used to be but it still sometimes can stay there until the end. The latest version coming out of FSD should finally merge city streets and Highway into a single stack, so hoping that will improve the merging performance.
Good FSD video -- there was certainly a lot going on for FSD,.
It had a lot to manage. 😊
it drove better than i would in those conditions
It’s good to see an FSD video that doesn’t show the same thing over and over.
Thanks... to be honest, when I first got FSD, I would've been too nervous to trust it downtown, but it is so much better now in these environments!
From Europe's perspective this grid of parallel and perpendicular streets is still extremely easy for drivers. Don't you have any real raw urban environment there, just these carefully engineered safe-spaces? :)
@@peter.g6 North American street systems are definitely more grid-heavy! FSD in Europe or (some of) Asia would be a totally different experience! Thanks for watching.
@@FSDwithme Yeah, thanks for the video, still very informative.
What version? What city?
Toronto. 12.5.4.1. I list the versions at the beginning of each video. 😊
Nice video. Two comments: At 3:58 better to see how the car handles the parked car in the left lane, rather than manually disengaging. Also, the screen in the car is rather low resolution and I like to follow the details normally. Thanks.
It would have had to adjust and get back in the lenr. I know it would have but I would have seriously affected some rush hour traffic. They shouldn’t have to suffer for this content! 😎
I agree I would rather have seen what it would’ve done. Plus, is he giving audio feedback about these disengagement?
You didn’t give FSD a chance - I really don’t think it was that difficult or unusual of a situation.
👍 sub here
Bro you are almost at the sub count for monetization! How is your watch time?
I'm still such a small channel, but the recent growth has been very appreciated. Watch time is solid! It's a cool community we have.
@@FSDwithme close to 4000 hours yet? In my experience that is the more difficult hurdle
@@RealBenAnderson Yeah, I passed that some time ago. This is a niche community, so any and all support is appreciated. Just love sharing these videos, they're fun to do, and I love the technology.
Version?
12.5.4.1
2:17 The man near the roadside was not detected until after he was passed. Look at the display. So, huge ERROR. Two reasons, one he is bent over and have a hoodie, looks like a trash can. Two, he is only partially lit up which most often confuses the image recognition. One of the ten original and unsolveable issues with FSD.
Honestly, I cannot get why virtually none of you FSD testers actually look at your videos reviewing incidents. I fully understand why you don't see the real issues on the fly when driving. However, the purpose of testing is understanding and contributing to improving the product, right?
It's easy to see why like Whole Mars that makes a lot of money from hiding mistakes do extremely bad analysis. But most of you guys is mostly doin it for fun I guess.
I partially agree, and think sensors would help there,.. wondering about the lighting conditions and what he was wearing.. Im not sure a human driver would pick it up much faster, but I see your point. FSD has been very cautious historically when it sees people standing close to the road and facing towards the road... as he was not walking in that direction, it did not see a reason to emergency brake. I did want to highlight it. I have no interest in sugar coating FSD, I dont make a dime off these videos!