Awesome- I'm LOVING 10.5 in Chicago as well. I'm having zero intervention drives about 1/3 of the time. I've had a few times friends didn't even know I was using it for the whole drive. It's come a LONG way in a short time.
@@fsdbetaboy1403 I’ve had it say use is restricted in the rain but it doesn’t actually kick me out. Roundabouts however…. I’d say it does about 80% of them well - the other 20 it clearly marks them with a red circle but then proceeds to drive directly through them. Never hit the report button faster :)
Wow, big congrats, Frenchie!! What a beautiful ride. It even didn't stop for that second immediate stop sign!! Thank you! Also, you deserve more subscribers, man. We will keep engaging audiences!
Wow, that's an awesome milestone! Your test route is far from easy, so this is a big deal. I still think it's worth repeating though because the exact road situations are constantly changing.
I too have had some surprising zero disengagement drives on 10.5. It's exciting seeing the improvements, but still has a ways to go before it's ready for public release.
Awesome performance. Maneuvering around parked trucks, *not* stopping for the double stop sign, slowing for a speed bump. Just amazing. I feel like I'm watching history in the making. Your visuals are excellent too, the car screen on top of the live forward camera. Would like to see you get around downtown... they have the Christmas Market set up on Daley Plaza. Barriers are up to block the street but you can't see them until there's no traffic in front of you. You can see the traffic turning south ahead of you but the barriers are low and the other cars obscure it till everything is clear ahead. That would be challenging.
I am missing Chicago, normally I am once every year there, so its great to the the streets and the improvements in the video. I hope FSD beta is coming to Germany soon.
Well. Hats off to you for your steadfastness, patience and dedication. A big day, maybe even a historic milestone, and deserving of recognition I think amongst your RUclips colleagues, and I hope that those who will undoubtedly be watching this in decades to come will echo those thoughts. Cheers.
Driving in and around Chicago can be such a pain! I'm amazed how good FSD 10.5 performed. You drove past Google in West Loop. It did well navigating under the L tracks too. Impressive stuff.
3:26 GPS is a little off this time. By the time it can clearly see the second median and where it is OK to turn, according to GPS it thinks it has passed the intersection. That seems to be why it just went straight at that point.
I always wonder about GPS and the FSD. In other videos when according to GPS a turn is coming up it often acts like it's not there until it can "see" it.
@@tecwzrd It has to understand what it sees well enough to be sure how that matches up with the GPS/Nav information, always with the possibility that the GPS could have tens of meters of error in positioning. In future I'm sure it will use street signs, addresses, and other landmarks, along with tracking the odometer and direction from last known position, to supplement GPS/map data.
Scott Graham: This is another example of why multiple modes, such as HD mapping, are helpful, re edge cases, for confidence. But of course, NOT trying to turn into the barriers or the tents and rerouting is infinitely better than what it used to do there. If I were a robo-taxi customer, I wouldn't even CARE about such a re-route, unless it caused a delay of annoying length.
@@rogergeyer9851 "HD mapping" (hi-rez 3D model of static terrain) makes a number of things "easier", but at a tremendous resource cost of creating and maintaining the HD map. Humans however do quite well on all sorts of weird roads and intersections even if we have never seen them before, and we can find out way with nothing but a crude hand-drawn not-even-to-scale map, or written turn-by-turn landmark & street sign based instructions. Tesla is expressly avoiding the HD map crutch in order to fully focus on the best possible vision AI as soon as possible, which is the best/fastest way to get to autonomy. The more and better it understands what it sees, the better it will be able to adapt to any location or situation and make the right decisions without any prior hd mapping being done. Tesla is working off a crucial insight into AI: that intelligence is inextricably linked to data compression. Amongst other things, the human brain is an mind-bogglingly efficient data compressor. HD maps are ridiculously inefficient use of data storage, which in turn means wasted compute to process all the unneeded data. By focusing on vision AI first, Tesla will eventually be able to have their fleet of cars generate an efficient "neuromap", which will be vastly more efficient and more useful, roughly analogous to how humans remember places we've been, tricky intersections, and so forth, except this will be a cloud-based shared memory among all vehicles. But it also will not be *required* for the vehicle to function, as they must function reliably while disconnected from the internet and lacking an up-to-date neuromap. Ideally it will usually be able to navigate even with GPS disabled. I'm not particularly interested in what is acceptable to a robotaxi passenger. I'm interested in what this situation tells me about the progress (and weaknesses) in the vision AI. FSD skips the turn and re-routes because it cannot clearly and accurately interpret the layout of the intersection until the car is almost to the middle of the intersection, showing that there is yet a lot of improvement to be made in the pure vision AI.
Very encouraging. It often feels like the program is stalled and then you wait a bit and don't pay attention and come back and bam. It feels like it has made a great improvement.
At 6:43 there seems to be an error in the displayed position of that semi. That semi is well past the curb, but it's shown on the visualization as further back.
Semis usually don't line up right. They don't stretch the model of the semi to line up with the actual dimensions of the vehicle it detects, so the model just gets snapped to one side of the detected bounding box. Probably it should have chosen the front of the box instead of the back, though. At least we know it saw the semi, given that it stopped weirdly.
@@justinmallaiz4549 The car did NOT detect that semi with ultrasonics. You can clearly see on the UI when ultrasonics detect something, and they only go off as it's passing the semi. On the other hand, we know for a fact that the semi visualization is not stretched to match bounding box size, based on previous videos when they were actually rendering bounding boxes.
Very nice video!! FSD is getting really good. Let me say one thing: as a German I’m shocked at the state of your roads!! Your roads are BUSTED!!!! Our worst roads here are much better holy shit, what is going on?!!
7:42 It didn't stop for that stop sign on the left hand side this time :) Seems to be getting better with updates every time (except 10.4) It's also handling rerouting much faster and easier now. Hopefully the next update will improve getting it through cross traffic faster, especially when it's already in the middle of the road.
Its definitely learning, I used to do driving instructing in the UK, and watching the FSD getting better is almost like watching a learner improve lesson by lesson. Its learning pattern is eerily human like. Its in the phase now where a human is about halfway to test I feel. Its confident in its abilities, maybe overly so, its thinking more about what its doing in the moment and not really planning ahead. For example left turning into narrow roads at a shallow angle meaning its cutting the lanes a little. 99 times out of 100 you get away with it, but 1 in a 100 your going to find someone who wasnt going to stop at the sign and isnt prepared for you to be blocking their lane. Same with right turns, its making its turn often from the left side of the lane. 999 times out of a 1000 you get away with it, but 1 in a 1000 a cyclist, a motorcycle, skateboarder or similar doesnt spot your signal or assumes youve signalled wrong due to your position and shoots up the inside. FSD has an advantage that it can always see all around it, but even if it stops in time it doesnt mean that the other vehicle will be able to and might end up going over your bonnet. You touched on it in the video too, that its not considering far enough ahead when its choosing lane positions that will ultimately bring it behind parked vehicles rather than staying in the outside lane. Its reached the point where it knows how to drive. and if there was no on else on the roads it could dive all day without issue. Its just not taking into account the "what if's" that other road users cause. Its definitely improving though. I would say keep your wits about you. Most of the times i was closest to an incident with a learner was at the stage FSD is now as you think they have it all under control right up until they don't and it takes you by surprise the first few times! Keep up the content, its very interesting!
@7:38 is something they really need to work on. The vehicle stays way too center at stop signs. On narrow roads there isn't even room for a car to pass on the left and I have to disengage.
Main thing for me is that the jump in capability proves the approach and capacity for Improvement. Improvement won't always be big but step changes will occasionally occur, with minor improvements the norm.
Congrats, Frenchie! xD Gotta say, just a few months ago I used to watch your vids and think it was gonna be a long time before it could handle cities like Chicago. Did not expect it to be anywhere near this good so soon! Long way to go before it's good enough everywhere but it'll be sooner than I expected! Wonder how long until robos are approved for limited areas. Less than a year for the capability. Regulation, though... *shrug*
awesome! seems to be happening a lot like Elon said it would... less and less interventions needed... eventually it will be how many drives per intervention instead of how many interventions per drive... Looking like we are on the borderline of that now.
Looks pretty smart and assertive with no intervention. Would like to see how FSD handle more traffic after it consistently perform this good in your regular test routine.
The reroute at the road closed sign was great actually. I'm more impressed by it apparently thinking "Ya, not going there. I'll take the next street." than just barging it's way through. That was a very impressive drive. It encountered lots of difficult situations and handled them very well. Any news on the next update from Elon? I've got a few problem locations that 10.5 handles poorly so I'm very anxious to see if the next version (10.6? 11.0?) solves those issues. Like everyone, I'd dearly like waypoints too as I don't particularly like the way the navigation system routes me to and from my work (takes the scenic route).
Nice drive! The only question that needs to be asked is: “would you have reached your destination safely if you were asleep in the back of the car?” The answer on this drive is clearly: “yes”. You’ll need more challenging routes soon! 😁 I’m expecting some regressions when they introduce new systems like the generalised height detection, unless that was 10.4 already. But once that is in we should be close to beating the average driver in most circumstances. Looking good! Thank you once again, Frenchie. 🙏
Hey man. Just found your channel. I'm local to the Chicagoland area and Tesla investor. Haven't purchased a Tesla yet, but maybe sometime in the next year or so. Love the videos. If it can handle crazy Chicago streets and traffic, then it should handle anything :)
I think barriers, signs and timing etc should be prioritized, manner can be addressed later. They can teach FSD 'tactics' like 16 wheelers are slower and more patient even later, those would be in the logic rule layer not muscle memories.
Compared to earlier versions FSD10.5 is an incremental improvement....but only incremental,. Real traffic in a busy urbain setting (say, Chicago loop, Manhattan midtown, Paris place de l'Etoile, etc.) it is still a work "in progress".
This is great! I wish my car did this well. I don't understand why all the RUclips videos seem to be so much better than my experience. My car is also on 10.5 and I can't seem to go long without disengaging. It phantom brakes all the time, almost every time I run it.. I almost don't want to use it because it's so stressful with it slamming on the brakes all the time.
great drive. Its interesting how in future drives you may have to go out of your way to find disengagements... to "chase the 9's" as Elon says.... it shows how far its come!
rerouting on road works is definitely ok, but incorrectly detecting truck (car was shown instead and as consequence incorrect size of truck was detected, so it was beeping on conflict between vision and ultrasonic sensors (sensors was correct) 0:45, so on high speed in such situation would be a crash. 5:31 pedestrian on the right (he closing truck back door) is not detected until car passes him. so with such two mistakes that's pure luck that there was no interventions(required).
agreed. But from a glass half full perspective these are incredibly easy edge cases to fix. All 3 issues this drive 0:45 -open truck door 5:31 - truck lift gate and 6:43 - abnormal red / blue Truck with replacement panels were big truck classifications that were thrown off from edge cases. The big truck classification needs to be more varied, loosened up and retrain. -I'd bet Tesla could do that on a lunch break :)
@@FrenchieFSD That's terrific! Thanks for the part you're playing in helping refine this amazing advancement. The impact that this technology will have on global infrastructure over the coming decade and our way of life going forward is under appreciated.
11:03 Turning from right side to the left. Seems like there's 2 lanes (no lines on the road) but you know if there was a car it would be right next to you. Ify
This was 7-8 out of perfect 10. I did not think in one year it would do this. Another order of magnitude improvement in a year and then it's as good as a very good driver? That would be really a big deal.
On second crossroad Tesla doesn´t stop on STOP sign ... Not good. But overall it is GREAT an big step forward. FSD are actually driving similat like little bit experienced driver, so want to be safe and go slowly, don´t want to risk, which is good I think.
I don’t find any noticeable difference between chill, average, or aggressive. I’m driving a model X and it seems more “jerky” with turns than your car. Yours seems very smooth. Enjoy your videos.
I wonder just why my Model S Plaid does not do half as good as yours is doing? After calibrating cameras, it really fails to get in turn lanes about 85% of the time. I'm finding it useless. Hope 10.6 will be better?
I agree with Steven. I’m in a busy area in suburban Phoenix with 4 lanes plus a merge lane at 45-55 mph. I find right and left turns are very hesitant to get in position and jerky, particularly with oncoming traffic, while yours seemed faster and smooth.. It looked like you filmed in real time and it was much different than what I’m seeing. It’s certainly possible the higher traffic volume and number of interactions are a key factor in the hesitancy. I also recalibrated the cameras after the upgrade.
@@FrenchieFSD I'm in Indian Wells, CA. more or less well marked Country roads. Today I tried again, and had to take over every left turn in a 15 mile trip, also some Very hard phantom braking for no reason. Very late and hard Right turns and missed 4 out of 5 right turn lanes. It was completely confused near some construction cones and tried to go into oncoming opposite traffic. If 10.6 isn't much better, than I'm opting out. No sense in putting the car through such rough handling, braking and risk. Glad your having a far better experience than I.
I think you need to continue that route. Short stops, slow acceleration, not getting completely back over after going around objects, cutting a few lefts off a few feet, temporarily getting "stuck" behind a truck not in the path. Yes, it's a million times better but I watch for all of the details. It's a good benchmark and I'm happy you don't intervene with the go pedal because that throws most people's test off.
To be honest I end up using it as a steering assist feature and basically using the accelerator to nudge it constantly and get it through a path - which is still "teaching" it
Awesome- I'm LOVING 10.5 in Chicago as well. I'm having zero intervention drives about 1/3 of the time. I've had a few times friends didn't even know I was using it for the whole drive. It's come a LONG way in a short time.
If mine were handling traffic as well as his, I'd only be having issues with roundabouts!
Well, that and Tesla's stupid new restrictions on rain =/
@@fsdbetaboy1403 I’ve had it say use is restricted in the rain but it doesn’t actually kick me out.
Roundabouts however…. I’d say it does about 80% of them well - the other 20 it clearly marks them with a red circle but then proceeds to drive directly through them. Never hit the report button faster :)
Wow, big congrats, Frenchie!! What a beautiful ride. It even didn't stop for that second immediate stop sign!! Thank you! Also, you deserve more subscribers, man. We will keep engaging audiences!
Thank you! I was so happy with the double stop sign. That was a pain
Wow, that's an awesome milestone! Your test route is far from easy, so this is a big deal. I still think it's worth repeating though because the exact road situations are constantly changing.
You are a part of history. When the documentary is made, all these videos will be in play
So much improvement since your first tests!
and that’s not even that long ago
@9:10 "onto the railroad crossing, budubudum" and this is why you're my favorite beta tester!!
really impressive drive... your videos are excellent.. would be great to see some tests in new areas also..
It would be interesting to see how it gets on with rush hour traffic on the same route.
These 10.5 videos are by far the most exciting! It’s getting scary good
I too have had some surprising zero disengagement drives on 10.5. It's exciting seeing the improvements, but still has a ways to go before it's ready for public release.
great video! The confidence of the visualizations are insane already and I can't imagine what they will be like in 2 years or something like that.
Awesome job! Lovely to see it improving and getting to zero's. Your enthusiasm and excitement is infectious. Thanks Frenchie
One small step for Frenchie, one pretty big step for FSD.
Wow that was impressive! Big kudos to the AI team.
Awesome performance. Maneuvering around parked trucks, *not* stopping for the double stop sign, slowing for a speed bump. Just amazing. I feel like I'm watching history in the making. Your visuals are excellent too, the car screen on top of the live forward camera. Would like to see you get around downtown... they have the Christmas Market set up on Daley Plaza. Barriers are up to block the street but you can't see them until there's no traffic in front of you. You can see the traffic turning south ahead of you but the barriers are low and the other cars obscure it till everything is clear ahead. That would be challenging.
I am missing Chicago, normally I am once every year there, so its great to the the streets and the improvements in the video. I hope FSD beta is coming to Germany soon.
I got emotional man. The moment FSD agressively turned left and then right at the end. Great video
Well. Hats off to you for your steadfastness, patience and dedication. A big day, maybe even a historic milestone, and deserving of recognition I think amongst your RUclips colleagues, and I hope that those who will undoubtedly be watching this in decades to come will echo those thoughts. Cheers.
Driving in and around Chicago can be such a pain! I'm amazed how good FSD 10.5 performed. You drove past Google in West Loop. It did well navigating under the L tracks too. Impressive stuff.
3:26 GPS is a little off this time. By the time it can clearly see the second median and where it is OK to turn, according to GPS it thinks it has passed the intersection. That seems to be why it just went straight at that point.
I like that explanation.
I always wonder about GPS and the FSD. In other videos when according to GPS a turn is coming up it often acts like it's not there until it can "see" it.
@@tecwzrd It has to understand what it sees well enough to be sure how that matches up with the GPS/Nav information, always with the possibility that the GPS could have tens of meters of error in positioning. In future I'm sure it will use street signs, addresses, and other landmarks, along with tracking the odometer and direction from last known position, to supplement GPS/map data.
Scott Graham: This is another example of why multiple modes, such as HD mapping, are helpful, re edge cases, for confidence.
But of course, NOT trying to turn into the barriers or the tents and rerouting is infinitely better than what it used to do there. If I were a robo-taxi customer, I wouldn't even CARE about such a re-route, unless it caused a delay of annoying length.
@@rogergeyer9851 "HD mapping" (hi-rez 3D model of static terrain) makes a number of things "easier", but at a tremendous resource cost of creating and maintaining the HD map. Humans however do quite well on all sorts of weird roads and intersections even if we have never seen them before, and we can find out way with nothing but a crude hand-drawn not-even-to-scale map, or written turn-by-turn landmark & street sign based instructions. Tesla is expressly avoiding the HD map crutch in order to fully focus on the best possible vision AI as soon as possible, which is the best/fastest way to get to autonomy. The more and better it understands what it sees, the better it will be able to adapt to any location or situation and make the right decisions without any prior hd mapping being done.
Tesla is working off a crucial insight into AI: that intelligence is inextricably linked to data compression. Amongst other things, the human brain is an mind-bogglingly efficient data compressor. HD maps are ridiculously inefficient use of data storage, which in turn means wasted compute to process all the unneeded data. By focusing on vision AI first, Tesla will eventually be able to have their fleet of cars generate an efficient "neuromap", which will be vastly more efficient and more useful, roughly analogous to how humans remember places we've been, tricky intersections, and so forth, except this will be a cloud-based shared memory among all vehicles. But it also will not be *required* for the vehicle to function, as they must function reliably while disconnected from the internet and lacking an up-to-date neuromap. Ideally it will usually be able to navigate even with GPS disabled.
I'm not particularly interested in what is acceptable to a robotaxi passenger. I'm interested in what this situation tells me about the progress (and weaknesses) in the vision AI. FSD skips the turn and re-routes because it cannot clearly and accurately interpret the layout of the intersection until the car is almost to the middle of the intersection, showing that there is yet a lot of improvement to be made in the pure vision AI.
It definitely acts much more confident than previous versions and most left turns are unprotected, that's a big step.
Great drive. It was great to see FSD handle so many of the situations it had trouble with in past videos. Good test.
Very encouraging. It often feels like the program is stalled and then you wait a bit and don't pay attention and come back and bam. It feels like it has made a great improvement.
I’ve been watching your FSD for months now and your videos are the best! Keep up the great work, you really need more exposure!
At 6:43 there seems to be an error in the displayed position of that semi. That semi is well past the curb, but it's shown on the visualization as further back.
Yeah the red/blue threw it off :)
Semis usually don't line up right. They don't stretch the model of the semi to line up with the actual dimensions of the vehicle it detects, so the model just gets snapped to one side of the detected bounding box. Probably it should have chosen the front of the box instead of the back, though.
At least we know it saw the semi, given that it stopped weirdly.
@@DontThinkSo11 yes, It stopped because it didn't know the truck was in the road. the ultrasonics detected it when it got up close,
@@justinmallaiz4549 The car did NOT detect that semi with ultrasonics. You can clearly see on the UI when ultrasonics detect something, and they only go off as it's passing the semi. On the other hand, we know for a fact that the semi visualization is not stretched to match bounding box size, based on previous videos when they were actually rendering bounding boxes.
Very nice video!! FSD is getting really good. Let me say one thing: as a German I’m shocked at the state of your roads!! Your roads are BUSTED!!!! Our worst roads here are much better holy shit, what is going on?!!
City is broke... That's the US for you. Even in France we have better roads
That's a huge achievement. I hope the program gets better with further iterations, but it's already starting to look smooth.
Awesome performance ego! Great vid Frenchie, thanks as always!
I'm very pleased with the progress we're experiencing with Beta 10.5.
They need to add axxhole mode so the car can squeeze every opening and not yield to anyone to be a regular human driver.
Always looking forward to how a new version does at the road closed sign :-). Nice drive, thanks
That was a very good performance. Great choice of turns L and R.
7:42 It didn't stop for that stop sign on the left hand side this time :) Seems to be getting better with updates every time (except 10.4)
It's also handling rerouting much faster and easier now. Hopefully the next update will improve getting it through cross traffic faster, especially when it's already in the middle of the road.
It did great , and it’s loverly to just have a look around your city.
Its definitely learning, I used to do driving instructing in the UK, and watching the FSD getting better is almost like watching a learner improve lesson by lesson. Its learning pattern is eerily human like. Its in the phase now where a human is about halfway to test I feel. Its confident in its abilities, maybe overly so, its thinking more about what its doing in the moment and not really planning ahead.
For example left turning into narrow roads at a shallow angle meaning its cutting the lanes a little. 99 times out of 100 you get away with it, but 1 in a 100 your going to find someone who wasnt going to stop at the sign and isnt prepared for you to be blocking their lane.
Same with right turns, its making its turn often from the left side of the lane. 999 times out of a 1000 you get away with it, but 1 in a 1000 a cyclist, a motorcycle, skateboarder or similar doesnt spot your signal or assumes youve signalled wrong due to your position and shoots up the inside. FSD has an advantage that it can always see all around it, but even if it stops in time it doesnt mean that the other vehicle will be able to and might end up going over your bonnet.
You touched on it in the video too, that its not considering far enough ahead when its choosing lane positions that will ultimately bring it behind parked vehicles rather than staying in the outside lane. Its reached the point where it knows how to drive. and if there was no on else on the roads it could dive all day without issue. Its just not taking into account the "what if's" that other road users cause.
Its definitely improving though. I would say keep your wits about you. Most of the times i was closest to an incident with a learner was at the stage FSD is now as you think they have it all under control right up until they don't and it takes you by surprise the first few times!
Keep up the content, its very interesting!
Wow, Outstanding Brother!
Cheers,
Eric
Best drive ever ! So exciting to see it handling Chicago so smoothly.
Thank you for sharing. Never seen Chicago with so few cars on the road???? Please do this same route with some cars on the road
@7:38 is something they really need to work on. The vehicle stays way too center at stop signs. On narrow roads there isn't even room for a car to pass on the left and I have to disengage.
Main thing for me is that the jump in capability proves the approach and capacity for Improvement. Improvement won't always be big but step changes will occasionally occur, with minor improvements the norm.
Congrats, Frenchie! xD
Gotta say, just a few months ago I used to watch your vids and think it was gonna be a long time before it could handle cities like Chicago. Did not expect it to be anywhere near this good so soon! Long way to go before it's good enough everywhere but it'll be sooner than I expected! Wonder how long until robos are approved for limited areas. Less than a year for the capability. Regulation, though... *shrug*
awesome! seems to be happening a lot like Elon said it would... less and less interventions needed... eventually it will be how many drives per intervention instead of how many interventions per drive... Looking like we are on the borderline of that now.
Good video, thanks! I ordered my Model 3 yesterday! 8-12 weeks is going to seem like years!
Congratulations! You'll love it!
Looks pretty smart and assertive with no intervention. Would like to see how FSD handle more traffic after it consistently perform this good in your regular test routine.
I am impressed. A long drive in Chicago and not one gunshot.
Hahaha
Imagine how many people love each other in Chicago!
Zero
Once the car can drive without disengagement’s, FSD will be able to learn exponentially via drivers reporting micro-errors
Need to incorporate at least two different unprotected left turns at stop lights for your route.
Wow. Bought all tsla I can afford to build the positions. Let’s go.
Thank you. Great video. A Milestone!
The reroute at the road closed sign was great actually. I'm more impressed by it apparently thinking "Ya, not going there. I'll take the next street." than just barging it's way through. That was a very impressive drive. It encountered lots of difficult situations and handled them very well.
Any news on the next update from Elon? I've got a few problem locations that 10.5 handles poorly so I'm very anxious to see if the next version (10.6? 11.0?) solves those issues. Like everyone, I'd dearly like waypoints too as I don't particularly like the way the navigation system routes me to and from my work (takes the scenic route).
I think it's called "Failing Gracefully" and is an excellent skill to have!
Nice drive! The only question that needs to be asked is: “would you have reached your destination safely if you were asleep in the back of the car?” The answer on this drive is clearly: “yes”.
You’ll need more challenging routes soon! 😁
I’m expecting some regressions when they introduce new systems like the generalised height detection, unless that was 10.4 already. But once that is in we should be close to beating the average driver in most circumstances.
Looking good!
Thank you once again, Frenchie. 🙏
Hey man. Just found your channel. I'm local to the Chicagoland area and Tesla investor. Haven't purchased a Tesla yet, but maybe sometime in the next year or so. Love the videos. If it can handle crazy Chicago streets and traffic, then it should handle anything :)
Gad you found me! When are you planning on reviewing Tesla Tech on your channel? :)
All this progress is awesome! Next.... Let's see how it goes when the snow falls and builds up.
Nice!! I'm so hopeful to finally get my FSD someday!
Can you do an updated one on 10.8 thanks!
FRENCHIE, THAT MIGHT'VE BEEN THE BEST FSD DRIVE I'VE EVER SEEN!!!!;)💋💋💋💋💋💥💥💥💥💥😭😭💕🤠
Great video and a great drive!
Here from xQc! Great video, subscribed.
I think barriers, signs and timing etc should be prioritized, manner can be addressed later. They can teach FSD 'tactics' like 16 wheelers are slower and more patient even later, those would be in the logic rule layer not muscle memories.
Loving your vids mate, keep it up 🙏
Compared to earlier versions FSD10.5 is an incremental improvement....but only incremental,. Real traffic in a busy urbain setting (say, Chicago loop, Manhattan midtown, Paris place de l'Etoile, etc.) it is still a work "in progress".
how do you look at that and dont think ev´s are just around the corner. that is super crazy awesome
Salut capitaine! Keep up the great work!
Can't wait to finally get the FSD download annnyy day now..
Great ride!
Thanks!
The number of crossroads with just a stop sign is insane in your city! So hard for an AI and yet it got to solve them.
Beyond impressed
It’s kind of crazy how well it does and mine freaks out constantly breaking ugh I don’t understand
Wow, really impressive!
7:18 why is almost every single car flashing its warning lights on that street 😄
5:36 FSD did not see the person standing at the back of the truck, it seems...
Yeah. It only detected him once we past the truck. That was strange. Maybe because he was wearing all black close? Still very unusual.
At this point if you are a CEO of any other auto company you will be hyperventilating watching this video
So good drive. Doe joe at work???
This is great! I wish my car did this well. I don't understand why all the RUclips videos seem to be so much better than my experience. My car is also on 10.5 and I can't seem to go long without disengaging. It phantom brakes all the time, almost every time I run it.. I almost don't want to use it because it's so stressful with it slamming on the brakes all the time.
I have some terrible drives off camera. It's not always like this. Maybe FSD is trying to show off on camera! :)
Great videos! thanks.
great drive. Its interesting how in future drives you may have to go out of your way to find disengagements... to "chase the 9's" as Elon says.... it shows how far its come!
Bravo! Big improvement 👍🏻
rerouting on road works is definitely ok,
but incorrectly detecting truck (car was shown instead and as consequence incorrect size of truck was detected,
so it was beeping on conflict between vision and ultrasonic sensors (sensors was correct) 0:45, so on high speed in such situation would be a crash.
5:31 pedestrian on the right (he closing truck back door) is not detected until car passes him.
so with such two mistakes that's pure luck that there was no interventions(required).
agreed. But from a glass half full perspective these are incredibly easy edge cases to fix. All 3 issues this drive
0:45 -open truck door
5:31 - truck lift gate
and 6:43 - abnormal red / blue Truck with replacement panels
were big truck classifications that were thrown off from edge cases.
The big truck classification needs to be more varied, loosened up and retrain.
-I'd bet Tesla could do that on a lunch break :)
Excellent! FSD in Chi-town is nearing graduation from Beta. What do you think, eight months till graduation (July 2022)?
I would be conservative and give it a end of 2022 if Tesla continues to improve at this rate
@@FrenchieFSD That's terrific! Thanks for the part you're playing in helping refine this amazing advancement. The impact that this technology will have on global infrastructure over the coming decade and our way of life going forward is under appreciated.
beautiful.
Incredible!!
Great video!
great! thanks for the video
Wow, 6:37 the rendering is so off. I guess NN for vision has some serious training gaps.
11:03 Turning from right side to the left. Seems like there's 2 lanes (no lines on the road) but you know if there was a car it would be right next to you. Ify
10:43 Didn't it cut across a lane to turn?
Great video! Be-dem Be-dem!!
Very impressive some humans cannot drive this good.
This was 7-8 out of perfect 10. I did not think in one year it would do this. Another order of magnitude improvement in a year and then it's as good as a very good driver? That would be really a big deal.
I feel like 10.5 has been an extra slow roll out, i'm still on 10.4......
Fantastic Frenchie!
On second crossroad Tesla doesn´t stop on STOP sign ... Not good. But overall it is GREAT an big step forward. FSD are actually driving similat like little bit experienced driver, so want to be safe and go slowly, don´t want to risk, which is good I think.
9:14 is a bop
I don’t find any noticeable difference between chill, average, or aggressive. I’m driving a model X and it seems more “jerky” with turns than your car. Yours seems very smooth. Enjoy your videos.
Same for me
@@Exalta101 2019
When you put in your destinations does the turn signals come on automatically
Depends if it requires to turn soon or not. Usually it doesn't
I saw more STOP-Signs in this video than in the last 40 years of my life... 🛑🤣
I wonder just why my Model S Plaid does not do half as good as yours is doing? After calibrating cameras, it really fails to get in turn lanes about 85% of the time. I'm finding it useless. Hope 10.6 will be better?
Hum. Give it some time, my first drives were horrendous. Where about are you? I wonder if Tesla has enough testers in your area.
I agree with Steven. I’m in a busy area in suburban Phoenix with 4 lanes plus a merge lane at 45-55 mph. I find right and left turns are very hesitant to get in position and jerky, particularly with oncoming traffic, while yours seemed faster and smooth.. It looked like you filmed in real time and it was much different than what I’m seeing. It’s certainly possible the higher traffic volume and number of interactions are a key factor in the hesitancy. I also recalibrated the cameras after the upgrade.
@@joby6462 Thanks Joby. Yes I also suspect multi lane is quite hard for FSD right now. Hopefully it will improve over time.
@@FrenchieFSD I'm in Indian Wells, CA. more or less well marked Country roads. Today I tried again, and had to take over every left turn in a 15 mile trip, also some Very hard phantom braking for no reason. Very late and hard Right turns and missed 4 out of 5 right turn lanes. It was completely confused near some construction cones and tried to go into oncoming opposite traffic. If 10.6 isn't much better, than I'm opting out. No sense in putting the car through such rough handling, braking and risk. Glad your having a far better experience than I.
3:24 Will someone please kidnap that Road Closed sign for the Museum of Autonomy? For the sake of posterity!
I'm in the UK or I'd do it myself.
Can I use the FSD for my driver's license test?
I think you need to continue that route. Short stops, slow acceleration, not getting completely back over after going around objects, cutting a few lefts off a few feet, temporarily getting "stuck" behind a truck not in the path.
Yes, it's a million times better but I watch for all of the details. It's a good benchmark and I'm happy you don't intervene with the go pedal because that throws most people's test off.
Zero intervention and Maximum comfort are usually not tied together with beta (yet)
To be honest I end up using it as a steering assist feature and basically using the accelerator to nudge it constantly and get it through a path - which is still "teaching" it