Tesla’s vision-only approach in a downpour is a classic example of high-tech hubris prioritizing ideological minimalism over redundant safety. It is an arrogant bet against physics that treats human life as a data point for engineering ego.
Deep Dive
Prerequisite Knowledge
- No data available.
Where to go next
- No data available.
Deep Dive
Tesla Vision-Only FSD In A Downpour
Added:It is moving left. Look at that.
>> Oh yeah. Wow. There is a lot of water right here for sure.
>> Hey guys, welcome to another Out of Spec Bits episode. Um, it is pouring rain outside right now. So, I think what we should do is go see if FSD can work in torrential downpour. This is like red on the radar map. Pretty crazy rain. Um, it's just about to pass through, so we may catch the tail end of it. But a lot of people talk about how driver assistance systems need multiple sensors for redundancy. And um the reason they say that and a lot of people you'll hear will say, "Okay, FSD maybe works really well camera only, vision only on a beautiful sunny day with no bad weather." Well, we've got a pouring rain downpour situation happening at the track right now. And I'm gonna grab Alex. We're going to run out to the car and put it in FSD and see what the heck happens. Um, that's it's just a quick little extra video. We figured we'd go test it out. This is not us saying these are the worst possible conditions.
There's fog, there's nighttime, there's snow, there's other things that can happen. This is just a heavy rain scenario in the middle of the day.
Typical big North Carolina summer thunderstorm rolling through. So, let's just take the car out, go down the highway for an exit, and see what the heck happens.
Stop right there, folks, because some of you may fast forward through this part of the video, but stick with me for this one. We are just weeks away from the CCAN's annual EV raffle giveaway program ending. And that means that your chance to win an electric car is amazing. And you really don't get a chance unless you get in on the action and get a ticket.
So, we're getting close to the end here.
And this is why I really want you guys to watch this video. There are only 10,000 tickets available to be sold, which means your chances to win are amazing because every ticket gives you three chances. If you've watched this part of the video, you already know the spiel. There are some incredible prizes this year, including a Porsche electric for the first place option. You could also choose a Rivian or a Lucid if you happen to win that one. And that is just so cool. I mean, the prizes are amazing.
They cover all taxes and fees as well.
Second place option, Rivian R2, the freaking awesome vehicle, Volkswagen ID Buzz or Hyundai Ionic. And the third place opportunity is the new Chevy Bolt EV. So, we're getting right up on the edge of this program. If you've been delaying spending the $200 for a ticket because you just wanted to see how it would play out, well, guess what?
Tickets are slightly undersold, which means your chances of winning are absolutely amazing. So, get in on there.
All you have to do is head to evraffle.org.
evraffle.org.
$200. Enter in. Let me know when you win. Thanks to the CCAN Action Fund for sponsoring this video. Let's get back to the main portion of the content. The car is right there. Let's do [laughter] it.
>> All right.
>> Holy smokes, [laughter] dude.
>> It's a little bit wet out there.
>> It really is. All right. Okay. So, I'm going to hand that to you.
>> All right.
>> Thanks. Let's buckle up for safety. How convenient is it that Sam's car was parked right here.
>> Indeed.
>> And let's go.
I don't know. Let's go into town. We'll go Cracker Barrel back.
>> All right.
>> And just start self-driving. They're safe right now in his tractor. Maybe show everyone that.
>> There he is. And the wipers are going crazy. So, everyone says you need LAR, radar, and a bunch of other stuff. We are about to drive off a curb.
>> Right off the curb.
>> Right off the curb. But that's fine.
>> Yeah, [laughter] that's kind of a weird spot where we were parked. It's like on a sidewalk basically.
>> This is what the cameras can kind of see. So, I guess we're just jumping in and trying this out. I I might even go heated seat here, Alex. It's a little little chilly. If I go here to software, we are on FSD version 14.3.4.
But ultimately, I think the big question that we're trying to answer is, okay, everyone says FSD camera only Tesla's approach works really well on the sunny perfect condition day, >> right?
>> Uh, now I have driven Tesla FSD in poor weather conditions and I know it can handle some situations, maybe not all situations. Um, especially like my hardware 3 car will freak out sometimes in rain like this, >> right?
>> Poor poor visibility, >> right? Poor visibility. It'll like limit your top speed, which I was driving a hardware 4 Model Y to work today, and that limited my speed on the highway to 65, but it was like this, >> right?
>> So, I guess the question is, I mean, this is about as much rain as you could really expect. The camera always makes it look >> Yeah. less than you actually are getting, but it's almost white out rain conditions. Like right now, >> I mean the Yeah, I mean the amount of rain coming on the windshield, the wipers can't even keep up.
>> It It's a lot.
>> And FSD it says might be degraded or weather detected.
>> But so far there's no indication that anything is driving unnecessarily. I'll actually put us into hurry mode.
>> Yeah, I think we were in sloth, right?
>> We were. It was going a little gentle around there.
>> Yeah. So, we'll uh >> I actually had it in sloth because there was so much water on the road on the way here.
>> Oh, yeah.
>> I was like almost going to hydroplane.
It was going so fast.
>> Yeah.
>> I was like, you need to chill out.
>> So, there are some, you know, this is I wouldn't say it's an edge case. And I but I also wouldn't say the system is fully optimized for poor weather yet.
Um, in this exact car, my buddy Logan and I look at the water on the road. It actually did hydroplane a little bit there.
>> Yeah. Just a little.
>> Yeah. And it kind of corrected itself.
My buddy Logan and I um took this exact car on the track when it was covered in ice.
>> Oh yeah.
>> And it was handling it and doing some crazy stuff. But when it would get too crazy, when you would get ESP intervention, it would actually uh ask you to take over at that point.
>> Oh yeah.
>> So I'm putting us in slob because it's going way too fast for >> It's going pretty fast for the conditions right now.
>> Way way way too fast. then there's no way to slow it down, which is something I think Tesla really needs to bring back, which is max speed limits in FSD, >> right? Yeah. I mean, you can speed it up by pushing the accelerator pedal, but there's no way for you to slow it down.
>> It's kind of a real annoyance for situations like this. It's always going a little bit too slow or a little bit too fast on the highway, >> right?
>> And I just want to have it go the speed I tell it to go. Ultimately, I'm the one responsible for the ticket, so I don't really want to go more than like, you know, nine over the speed limit.
Hopefully, it doesn't pull out from this truck. It's not.
>> Yeah, it was creeping a little bit.
>> Rain did let up a little bit here, but >> yeah, it's not quite as heavy here for sure.
>> Right. It's definitely coming in like batches and in waves, but I think there's a lot of misinformation out there, which is like, okay, these systems will not work in poor weather.
And I think what I've learned at least is it will work in poor weather. But kind of like a human driver, you're just you just got to be a little bit easier, a little bit more gentle with it, >> right?
>> And as far as I can tell, the camera can always see better than I can see in extremely bad weather scenarios. Like there will be situations where sometimes I take a look. What's this guy doing?
>> I think he's going >> way over the line.
I was thinking he might try to >> Yeah, that's what I was thinking too because a lot of people do that at that roundabout, >> right? If I go here to service camera visibility, no camera visibility issues detected.
>> Okay, >> so that's good. This doesn't have the newest update that lets you pull up all the cameras that it's using for FSD at once, but we'll jump on the highway and see what it does.
>> Yeah, >> just go down one exit, up one exit.
Ultimately, this is just normal rain, but it is 11:30 in the morning and it's pretty dark outside, actually.
>> Yeah, I think we may have caught the tail end of the storm. We'll see.
>> Yeah.
>> So, I guess my question is whenever I've used LAR, maybe it's not LAR, but whenever I've used night vision, which is not LAR, that's infrared.
Infrared doesn't seem to work when it's raining. At least in the cars that I've owned and tested >> with IR sensors. So when you drive like a Escalade IQ or uh you know a TYON or Cayenne with night vision or whatever, everything just goes gray when it's raining, >> right?
>> So that's that particular technology. I don't really think many systems use IR.
It's just something to note. Liar does see through uh rain, I believe, >> right?
>> But I do know that in really sunny conditions, LAR struggles, right? It has to like up the power of the output device. We're going real slow off the highway now. See here >> stuck behind this uh slowpoke.
>> You see this indication here, Alex? It says maximum speed limited >> or road conditions detected.
>> Okay. So, it it knows um and it's not going to go faster than whatever it deems as necessary. Also, it will not let me put it in anything higher than standard drive mode. Oh, interesting.
>> So, no hurry, no mad max. That's locked out at the >> But I mean, look at human drivers.
Everyone's doing maybe 50 miles an hour.
We're going a bit slow, but >> you know, maybe 50 mph on the highway.
We'll see if it goes for a pass automatically or not.
>> Yeah. If it wants to pass this guy or >> clearly the cameras can see pretty well.
And this is just the bumper camera, >> right? Yeah.
>> This camera probably sees no issue whatsoever.
>> Yeah, that one's probably way better.
>> Mhm.
>> For sure. is that's the back bumper.
These are the sides, which are perfect.
And I spoke one time to the company that did the aerodynamics works to um determine where to place the cameras for those B-pillars.
>> Oh. So that they don't get locked with water. Okay.
>> Yeah. They're like perfectly done.
>> So like a lot of people put covers and weird stuff on them and it messes up the arrow.
>> Oh yeah.
>> The cameras massively. So you don't don't mess with that is what I've learned. after talking to those guys, they're like perfectly positioned. So, so far, no issues using it in the rain.
Um, so Tesla uses a camera only approach, which I believe they get a lot of flack for. Um, I see both sides of the argument. Wouldn't it be nice to have more data, but then also what if you do what do you do when sensors disagree? You've got to do some fusion and >> right >> determine what's true and what's false.
On the other side um of the coin, you go okay um more sensors is better because you can break you can take the path of most caution if something were to pop up, but then you get phantom brakes and other weird, you know, >> maybe even potential rear ending situations cuz you're always slamming on the brakes. Um, and I think Tesla's camera only approach, at least in my experience, has worked in almost any condition I can think of that I needed to completely white out conditions. No problem. There are scenarios where it will get so bad it will be blinded, where it tries to return the control of the vehicle back to the driver. And in those situations, I have no idea what it will do when it's autonomous.
>> Right.
>> Yeah. don't know how they're going to handle that and I'm sure they will figure it out and I'm sure it's just going to do the best it can which is what some humans do in bad situations.
>> Uh but it definitely seems to be a bit um >> yeah bit bit not ready maybe for full autonomy in the worst possible weather conditions.
>> Right.
>> But thankfully the cars know what's coming. So like here's one idea. I'm just thinking out loud. You can go here to the weather map, right? and it will show. Okay, we've got another thread now. We're pretty much out of the bad storm. It was that.
>> So, it can know, oh my gosh, we have some bad weather coming. Let me just park. Let that bit of the storm go by, >> right?
>> And then get back to autonomous driving.
Just an idea.
>> Yeah, >> there there's ways to make anything work really.
>> So, now we want to go back to work and it's all they're going to get in the left lane. Very snappy response.
>> Sweet.
>> So, there we go. Does it work in extreme rain? The answer is yes.
>> Yes, >> it does. And with hardware 4, I can't really think of a visual situation in terms of rain or snow or ice that I've got it in that it won't work from a visual perspective. There are stability stuff where it's like, oh, if you get a big yawn moment, if you're sliding, if it loses traction, it'll actually do what it can. But if you trigger stability control for too long, then it auto puts the um puts the responsibility back on the driver to drive, >> right?
>> But pretty cool what they can do with just cameras. The whole industry was kind of against them.
>> Yeah.
>> I I don't really care one way or another how companies get this stuff to work, but like this works really well with uh just camera only, right? I I see no reason a lidar would have benefited us during this drive today.
Just my experience. So, Alex, any thoughts on sensors? I don't think I've ever talked to you about that. What do you What's your preference? Cuz you have a Tesla with radar, >> right? My car has radar. Um, I mean, I would say that redundancy is good, but as long as it works, then does it really matter?
>> I would agree with that, >> right?
>> Yeah. Yeah. Yeah. I mean, this has eight cameras. That's a lot of redundancy.
Right.
>> I've only got two.
>> Exactly.
>> But sometimes they're not pointed in the right way, you know.
>> So, I think it's a lot better than what humans could do most of the time at least. And it's so good. I mean, I just trust the car so much to get get me down the road and I'm just looking around for any critical safety issues. So, there we go. Just a fun test to see if FSD works in really bad conditions. And the answer is at least on this day at this time in this car, yes, >> it seems to be doing the trick.
>> Thanks so much for watching. See you on another Out of Spec Bits episode soon.
Bye-bye. We are just about to go through a major hydro planning spot which is on the second bridge. Um I saw the cars coming >> when we were coming down going. I'm really curious to see what this will >> what this does.
>> Yeah, cuz the truck in front of us is going to blind us, >> right? It's going to blind us. The car isn't really going to see it until we're right on top of it essentially.
>> Exactly. Okay. I can see brake lights up ahead from people. It's going to get a little sketchy through here.
This is always a situation.
>> Oh, but we're really slowing down.
Someone is losing it. Here we go.
Yeah, I bet someone really slowed down for this.
>> It is moving left. Look at that.
>> Oh yeah. Wow. There is a lot of water right here for sure.
>> Oh, we did get splashed pretty good.
Still everything FSD though.
>> Yeah. I wonder what it would have done if there hadn't been such a traffic backup right here. Would it have just plowed right in or would it have slowed down?
>> Uh, that's a good question. I don't I bet it would have just flown right through that.
>> Yeah, >> it does slow down for big puddles though and it knows to avoid them. Originally, it would just like phantom break and have odd reactions. So, it's learned puddles.
>> Yeah.
>> It's funny to watch it figure things out. That was a lot of water, though.
>> Yeah, that was that was deep.
>> Yeah. So, they got to work on engineering the drainage for that bridge.
>> Yeah. Apparently, the uh drainage in that spot isn't so good.
Related Videos

Setting up a curved screen with Immersive Calibration Pro 4 and multiple cameras (P3D v4)
FlyerOneZero
23K views•2019-07-21

Robot Learning with Sparsity and Scarcity
allenai
379 views•2025-10-14

Jorge Mendez-Mendez: Unlocking Lifelong Robot Learning With Modularity (2023-10-05)
umassmlfl
237 views•2024-01-06

Northwestern’s MS in Robotics: Student Robotics Projects, 2023
NorthwesternEngineering
1K views•2024-05-31

"Perfect" Turns: Turning by the Gyro - FIRST LEGO League (FLL) SPIKE Prime + EV3 RePlay Programming
ZacharyTrautwein
94K views•2020-10-02

Gorkem Secer: TSLIP-based Deadbeat Running Control of Bipedal Robot ATRIAS
DynamicWalking-wv6qm
298 views•2018-06-22

Self-Driving Cars Need Lessons On Human Drivers | Maddie About Science
skunkbear
26K views•2018-08-21

Milrem Robotics’ THeMIS UGVs used in a live-fire manned-unmanned teaming exercise
MilremRobotics
99K views•2021-05-20
Trending

NOLAN WELLS AUTOPSY: DEEP TISSUE DISCOLORATION BACK OF HEAD, NECK BONE MISSING
nancygrace
561K views•2026-07-22

Big Tech's Biggest Gamble Is Finally Falling Apart
houseofel-ai
79K views•2026-07-22

we're almost finished the house (ep.125)
JennaPhipps
347K views•2026-07-22

We flooded a field - the results blew our minds
MossyEarth
73K views•2026-07-22