Tesla FSD Beta 10.69.25 Struggles in Snowy Detroit MI: A Scary Video | Hacker News Reader

Tesla FSD Beta 10.69.25 Struggles in Snowy Detroit MI: A Scary Video

youtube.com

31 pointsbaptiste31326 comments

FSD will continue to be a joke outside of the Bay Area pretty much indefinitely, at least for camera-only Tesla, for a really simple reason that’s obvious to anyone who has driven in a snowy metro: Salt spray.

Five minutes on the road and those cameras will be 100% coated. It doesn’t matter how much neural net training you do even on no-lane snowy roads or getting the car to know how to adapt to conditions, if the cameras can’t see anything it’s over.

Every time there’s some breathless article about upcoming cars that don’t even have a steering wheel I just laugh because they’re only ever going to sell or even work in SF.

If only there was some kind of way to clear visual blockages on cars. Without that, any kind of driving (automated or not) would be pretty difficult ;)

You'd be going through a lot of washer fluid spraying off all eight cameras every couple of minutes!

Without functioning spray and wipers, even a human wouldn't be able to make it further than 30 minutes where I live.

Some of us live in areas where a car wash subscription is common and basically required.

Since the Bay Area is currently a mess of flooded roads and intersections, I’m curious how many self driving Teslas will drive right into floods.

I'd love to see FSD up here in Canada in the winter.

We play "guess where the lanes are" at over 100km/h because our major highway system (401) ranges from 2 to 8 lanes wide and remains fully open.

Sometimes Humans get it wrong and crash even having years experience driving in snow, how will a tesla handle road conditions when everything is white and any "lane indicators" are buried in snow?

https://ottawa.ctvnews.ca/opp-responds-to-numerous-crashes-o...

Will FSD ever advance beyond some "ideal weather and road conditions" toy? I guess time will tell but it has a LONG way to go to support all road conditions.

"I'm risking my life, and my car".....fails to mention the general public.

This is a common objection, but it’s been a couple years now. Are we aware of any incident where the general public was harmed?

Uh? Yes? I'd link them but a quick google will help you. Despite the counter, my point remains. You're doing something knowingly unsafe in public. At least be conscious of the public...

If you were interested, you could have googled it.

Does an 8 vehicle crash where 9 people are sent to the hospital count?

https://www.cnn.com/2022/12/21/business/tesla-fsd-8-car-cras...

How about fatalities?

https://electrek.co/2022/11/16/tesla-reports-two-fatal-crash...

FSD doesn't work on interstates. It's literally impossible for the CNN article to be true.

Maybe the first Google hit isn't enough to get a complete understanding of the situation in this case.

Hmm.. story was run by CNN, ABC, etc.

If you dont agree with CNN, what about the others?

Even Reuters ran it : https://www.reuters.com/business/autos-transportation/tesla-...

If you read the article, it is still under investigation as well.

"The Tesla driver told police the FSD malfunctioned but police were unable to determine if the software was in operation or if his statement was accurate, according to the report which was made public after a Reuters request."

It is possible it is FSD, also possible the driver is lying.

> Maybe the first Google hit isn't enough to get a complete understanding of the situation in this case.

Maybe no google hit will work because sometimes people consider things more of a "religion" and no evidence can convince someone to sway from their religious beleifs. What does it say when you wrote "It's literally impossible for the CNN article to be" when multiple respected agencies carried the story?

Do you have Developer access to Tesla and know for a fact this is not true? If so, have you contacted any authority to report this?

Edit: do you dispute the fatalities as well?

I have FSDbeta so I'm very aware that it doesn't work on highways.

As far as the fatalities, the article itself states clearly that the data doesn't indicate whether autopilot or FSD was active (both are ADAS). The article also indicates that one of the incidents happened while the car was not moving. An article with any real value would dig a little deeper and share the circumstances of the accidents rather than two rows of a spreadsheet.

As a Tesla owner I would say on a personal level that the automatic emergency features of my car which leverage the self-driving work the company is doing are spectacular and have likely been a net benefit to road users. I think the Euro NCAP safety assist score reinforces this opinion (not to mention the overall score). I'm extremely happy to drive my two young girls around in what testing (not clickbait) reveals to be the safest car ever built.

Let us know if you also disagree with the nhtsa studies, and how your anecdotal evidence overrules them.

https://www.nhtsa.gov/laws-regulations/standing-general-orde...

They list 18 fatalities involving FSD, but i'm sure their inaccurate per your understanding?

I'd say a FSD, like we see in the youtube video, or that kills people as the NHTSA shows isnt a benefit to road users.

IN the youtube video, can the car go over 15 seconds without attempting to run into the curb?

I think if you check your facts you'll notice that the NHTSA doesn't distinguish between autopilot and FSD.

You should definitely not engage FSDbeta in environments that the computer vision system hasn't been trained on. If you do you should be sure to post it to YouTube.

Conditional Self Driving would be a more accurate name, then.

I noticed that too - closer to the end he mentions the public

I don't understand the approach that Tesla and others are taking to self driving. Why are they not building a model of the world and then basing decisions off of that? Roads do not magically change, so why are they constantly re-evaluating things and creating a new model all the time? There are static objects in the world (curbs) and then there are dynamic ones (other vehicles, people, traffic cones when a lane is shut down, etc). Every time I see tesla self driving stuff it looks like the internal concept of the world is wiggling and vibrating and changing alll the time. It seems to me that with enough sensors and fusion you can determine that a car is in fact parked and not wiggling. You can use known maps of the land to infer roads or at least general shapes of roads that could serve as a basis to stack realtime analysis on top of it. I can't help but think going all-in on image analysis is the worst approach possible for this sort of thing. Just because we have eyes and use them to see doesn't mean that computers should too. We have rules hard coded in our minds that help us do most of driving and our eyes are just giving the extra realtime data on top. I know that the road I am on is not going to magically shift/move. You can see very very far ahead in a car and immediately make concrete decisions about your journey (I know at least for the next 500 feet I am going to be on this same road that isn't changing, all I need to do is be mindful of vehicles in front of me).

I am not articulating this well but it has been on my mind for a long time. I can't help but think the current approach of throwing GPUs and OpenCV at driving is a fools errand. Tesla eliminated a lot of their other sensors like lidar and whatnot too, to go all-in on image analysis. Foolish.

The visualizations are separate from what the self driving sees

Aside from the FSD, is it legal to interact with the touchscreen like that while driving? I think it would be in the UK. https://www.gov.uk/using-mobile-phones-when-driving-the-law

Worst part is there doesn't even seem to be a significant amount of snow on the roads.

The cognitive dissonance is palpable.

I'm not a Tesla fan but to me the car seemed to handle it in a reasonable manner i.e. by disengaging whenever the FSD sensed slippery conditions (although the traction control system when on autopilot certainly seems to need some tweaking). I'm not sure what sort of tires were fitted but it seemed to struggle more when FSD was engaged vs. the human driver.

It's also trying to drive much too fast for the conditions and its veering all over the road until the driver takes over.

> by disengaging whenever the FSD sensed slippery conditions

If it needs to disengage so frequently in these conditions, its not FSD.

> I'm not sure what sort of tires were fitted but it seemed to struggle more when FSD was engaged vs. the human driver.

As a Canadian who has driven in snow for 33 years "blaming the tires" is odd. Most of my life i drove on all seasons and only got "snow tires" 4 years ago because I get a discount on my insurance for having them.

The reason it is having issues is because it is attempting to accelerate too quickly for the road conditions. Perhaps snow tires might help a bit, but if it is unable to handle "all seasons" and a bit of snow.. that itself is a problem and snows may not fix it because they are not some sort of magic that suddenly gives you amazing traction on something slippery like snow.

If there is over a few CM of snow on the ground, i can spin my 2 yer old snow tires as fast as i want... Speedo showing 90km/h while i am hardly moving... All i need to do is disconnect the TC and press hard on the gas, the snows quickly lose traction and spin.

The human driver is obviously lighter on the acceleration due to road conditions and so less TC is needed.

Reply on news.ycombinator.com