I always assumed that was the case. We are still experimenting with human drivers.
This is the best summary I could come up with:
"In late 2021, Lukasz realised that—even as a service technician—he had access to a shockingly wide range of internal data at Tesla," the group's prize announcement said.
Krupski was also featured last month in a New York Times article titled, "Man vs. Musk: A Whistleblower Creates Headaches for Tesla."
But Krupski now says that "he was harassed, threatened and eventually fired after complaining about what he considered grave safety problems at his workplace near Oslo," the NYT report said.
Krupski "was part of a crew that helped prepare Teslas for buyers but became so frustrated with the company that last year he handed over reams of data from the carmaker's computer system to Handelsblatt, a German business newspaper," the report said.
The data Krupski leaked included lists of employees and personal information, as well as "thousands of accident reports and other internal Tesla communications."
Krupski told the NYT that he was interviewed by the NHTSA several times, and has provided information to the US Securities and Exchange Commission about Tesla's accounting practices.
The original article contains 705 words, the summary contains 172 words. Saved 76%. I'm a bot and I'm open source!
It’s funny how some of Elongated Muskrat’s testing and experiments involve the subjects dying.
Monkeys dying with the Neuralink experiments, and humans are dying with these autopilot tests!
Lemmy as a whole appears to irrationally hate Tesla because of their stupid CEO. I think his penchant for calling what is essentially "advanced autopilot" FULL SELF DRIVING should be illegal. But he's a car salesman and for some reason the government is letting him call it that. Be mad at our lawmakers for that. He's just a sheister and our lawmakers suck at reining him in.
Tesla cars themselves are actually really good. Very safe cars that don't roll over because of the heavy battery located so low, very responsive acceleration, and some nice quality of life low hanging fruit in the technology department, like my phone being a key. I was told by my Tesla rep when I bought the car to not buy FSD. It's experimental and will not ever probably be driving you to your destination safely. The fact that they sell it with a name that implies it will is the problem. And people believe it. That's incredibly dangerous.
On the same note of blame the lawmakers
There's a lot of hate about Teslas cars not reaching EPA estimates on highways.
The EPA test is the problem. The test doesn't include real world driving such as at 70mph and for whatever reasons, a Tesla often takes a bigger hit at 70mph than some other cars.
I don't doubt Tesla did some ratio optimization on the motors to get better EPA numbers, that's just playing the game, but please lobby the EPA to change the testing methodology.
Tests need to better include faster driving. Manufacturers should be required to show both numbers not a combined number in their advertising materials, and they really need to add some sort of cold weather test.
Edit: also the whole 2 different test cycles they can choose between is ridiculous. Make it all the same.
This infuriates me to no end. The EPA could just mandate multiple numbers!
I want a graph of the car going every speed between 55, 65, 75, and 85 on a treadmill at 0, 25, 50, 75, and 100°F while maintaining a cabin temp of 72°F.
I want to know how much battery it used at those temps, simulating catching every red light in a downtown setting, in an hour.
I want discharge rates for all those temps with it just sitting there for a week, same for a month.
"Combined blah" is horseshit.
Hey, EPA, hire this person!!
Non-consentual Human Experimentation is a war crime.
It's peace time though so it doesn't qualify
/s
It's consentual if you buy it though.
Calling it a war crime is slightly extreme.
Except the other drivers on the road aren't all in Teslas, yet they are non-consentually and possibly even unknowingly a part of this experiment.
If you hit another motorist or pedestrian, it’s no longer consensual.
War crime is a tad much sure. Let’s just make it a felony.
Random question I've always wondered about in case anyone is more familiar with these systems than me. My understanding is that autopilot relies on optical sensors exclusively. And image recognition tends to rely on getting loads of data to recognize particular objects. But what if there's an object not in the training data, like a boulder in a weird shape? Can autopilot tell anything is there at all?
Yeah obstructions can be generalized to a road being blocked. Object recognition includes recognizing the shape of an object via curves, shadows, depth, etc. You don't need to know it's a boulder to know a large object is in the road.
Something like snow? The vehicle ignores what to do or where to go.
FSD, maybe. But autopilot operates fine and is no different than what most major manufacturers offer.
Edit: Lots of people that have never used Tesla or other manufacturers lane keeping systems I see.
Last time I tried autopilot was 4 years ago, so I imagine things have become better. That said, on a test drive, on a rainy day, auto lane change did some fighting stuff. Thought lanes were clear, learned they weren’t, then violently ripped the car back to the origin lane in conditions that were prime for hydroplaning.
My wife and I were scared shitless, and the woman from Telsa, who was also in the car, tried to reassure us by saying “it’s ok, this is normal.”
Then we return the car to the parking lot and auto park almost took out a kid in an enclosed parking structure.
I imagine it’s become better in 4 years, but how that was street legal baffled my mind.
None of what you mentioned is in basic autopilot. Autopilot is lane keep and traffic aware cruise control only.
Let’s not get pedantic. They are part of the “enhanced autopilot” package.
If these were called "cruise control", "adaptive cruise control", and "Blue Cruise" would it matter if the article said "cruise control" but was referring to "Blue Cruise"?
Tesla's names for these things are "Autopilot", "Enhanced Autopilot", and "FSD Beta".
At the very least, the names matter so that we can all agree we're talking about the same things.
Which is not included with the base vehicle. It’s an extra purchase.
Well in that case, the advanced autopilot features that almost killed me were totally safe.
Sure, which I consider part of FSD, which almost killed me like 3 times when I had a loaner with it active.
But that’s not basic autopilot. AP is fine assuming people pay attention.
which I consider part of FSD
Well when Telsa, this former employee / whistleblower, and these journalists refer to “autopilot,” they’re specifically talking about the software and hardware marketed under the “____ Autopilot” banner that Telsa uses for those features.
Some of these more advanced autopilot features clearly have issues, and it probably stems from the fact that they’re only using cameras and ultra sonic, not lidar.
In my experience with a Model 3 and AAP, when those cameras and sensors were wet, it was pretty clear that they were getting dangerous. It started raining during our test drive, so we had a before / after experience on the same roads. Once everything got obstructed with water, you could see the car’s collision detection struggle to detect other objects. Objects on the center display would erratically pop in out of view. And this was a showroom car, it wasn’t the first rain of the year, and it was behaving “normally” according to staff.
Even if basic autopilot was fine, this left such a sour taste in my mouth that I had no appetite to give that company my money. Almost dying and almost killing a kid were a big “fuck this company” for me.
My (non-tesla) vehicle can tell when the sensors are impaired by frost or mud or whatever. It flashes a warning on my dash and disables the lane-keeping and/or collision detection until next startup. Does Tesla not do that?
It was supposed to
AP is fine assuming people pay attention.
There’s a human tendency to become complacent after a while, which presents a risk.
Can’t wait for safer-than-human self-driving technology, and know we’ll need to take certain risks of some sort to get there, but there are good arguments against “PLEASE remain fully attentive 100% of the time for this technology that will in fact only require full attentiveness in edge cases”. You might be an exception of course! But Average Meat Driver is going to slip into complacency after many, many miles of perfect autopiloting.
It’s the same as cruise control, but it’s supposed to eliminate human error. I’d argue most of the people having issues from not paying attention probably weren’t paying attention in the first place and were dangerous to begin with
My vehicle can do almost all the same stuff as "autopilot" but it turns the autosteering and cruise off if I dont touch the wheel every 30 seconds. Its all the same types of sensors,etc. And mine isn't even a luxury brand. Just the higher end trim package of a budget vehicle.
edit: actually, it's just 10 seconds before the warning and another 5 or so before it disables lane-keeping
Autopilot also shuts off with no driver input. Faster than 30 seconds too.
I own a model 3 and a 2022 palisade with Lane assist and used to own a Subaru with Lane assist.
The model 3 auto steer, exit to exit EAP, and auto lane change are very different than the simple lane assist that either other cars offer and honestly after using EAP for five years, while I do use AP under specific circumstances, I have come to the opinion that it is not ready for prime time and has some major issues, especially the auto lane changing, that should have been worked out before release and I still never use that feature.
Given my background in embedded software, I honestly think the way they rolled out and advertised these features was reckless.
EAP is not based autopilot and closer to FSD. Base autopilot is on par with most manufacturers. I’d argue it’s safer than some in regards to capabilities with less common lane setups or lack of clear road lines.
I always assumed that was the case. We are still experimenting with human drivers.
This is the best summary I could come up with:
"In late 2021, Lukasz realised that—even as a service technician—he had access to a shockingly wide range of internal data at Tesla," the group's prize announcement said.
Krupski was also featured last month in a New York Times article titled, "Man vs. Musk: A Whistleblower Creates Headaches for Tesla."
But Krupski now says that "he was harassed, threatened and eventually fired after complaining about what he considered grave safety problems at his workplace near Oslo," the NYT report said.
Krupski "was part of a crew that helped prepare Teslas for buyers but became so frustrated with the company that last year he handed over reams of data from the carmaker's computer system to Handelsblatt, a German business newspaper," the report said.
The data Krupski leaked included lists of employees and personal information, as well as "thousands of accident reports and other internal Tesla communications."
Krupski told the NYT that he was interviewed by the NHTSA several times, and has provided information to the US Securities and Exchange Commission about Tesla's accounting practices.
The original article contains 705 words, the summary contains 172 words. Saved 76%. I'm a bot and I'm open source!
It’s funny how some of Elongated Muskrat’s testing and experiments involve the subjects dying.
Monkeys dying with the Neuralink experiments, and humans are dying with these autopilot tests!
Lemmy as a whole appears to irrationally hate Tesla because of their stupid CEO. I think his penchant for calling what is essentially "advanced autopilot" FULL SELF DRIVING should be illegal. But he's a car salesman and for some reason the government is letting him call it that. Be mad at our lawmakers for that. He's just a sheister and our lawmakers suck at reining him in. Tesla cars themselves are actually really good. Very safe cars that don't roll over because of the heavy battery located so low, very responsive acceleration, and some nice quality of life low hanging fruit in the technology department, like my phone being a key. I was told by my Tesla rep when I bought the car to not buy FSD. It's experimental and will not ever probably be driving you to your destination safely. The fact that they sell it with a name that implies it will is the problem. And people believe it. That's incredibly dangerous.
On the same note of blame the lawmakers
There's a lot of hate about Teslas cars not reaching EPA estimates on highways.
The EPA test is the problem. The test doesn't include real world driving such as at 70mph and for whatever reasons, a Tesla often takes a bigger hit at 70mph than some other cars.
I don't doubt Tesla did some ratio optimization on the motors to get better EPA numbers, that's just playing the game, but please lobby the EPA to change the testing methodology.
Tests need to better include faster driving. Manufacturers should be required to show both numbers not a combined number in their advertising materials, and they really need to add some sort of cold weather test.
Edit: also the whole 2 different test cycles they can choose between is ridiculous. Make it all the same.
This infuriates me to no end. The EPA could just mandate multiple numbers!
I want a graph of the car going every speed between 55, 65, 75, and 85 on a treadmill at 0, 25, 50, 75, and 100°F while maintaining a cabin temp of 72°F.
I want to know how much battery it used at those temps, simulating catching every red light in a downtown setting, in an hour.
I want discharge rates for all those temps with it just sitting there for a week, same for a month.
"Combined blah" is horseshit.
Hey, EPA, hire this person!!
Non-consentual Human Experimentation is a war crime.
It's peace time though so it doesn't qualify
/s
It's consentual if you buy it though.
Calling it a war crime is slightly extreme.
Except the other drivers on the road aren't all in Teslas, yet they are non-consentually and possibly even unknowingly a part of this experiment.
If you hit another motorist or pedestrian, it’s no longer consensual.
War crime is a tad much sure. Let’s just make it a felony.
Random question I've always wondered about in case anyone is more familiar with these systems than me. My understanding is that autopilot relies on optical sensors exclusively. And image recognition tends to rely on getting loads of data to recognize particular objects. But what if there's an object not in the training data, like a boulder in a weird shape? Can autopilot tell anything is there at all?
Yeah obstructions can be generalized to a road being blocked. Object recognition includes recognizing the shape of an object via curves, shadows, depth, etc. You don't need to know it's a boulder to know a large object is in the road.
Something like snow? The vehicle ignores what to do or where to go.
FSD, maybe. But autopilot operates fine and is no different than what most major manufacturers offer.
Edit: Lots of people that have never used Tesla or other manufacturers lane keeping systems I see.
Last time I tried autopilot was 4 years ago, so I imagine things have become better. That said, on a test drive, on a rainy day, auto lane change did some fighting stuff. Thought lanes were clear, learned they weren’t, then violently ripped the car back to the origin lane in conditions that were prime for hydroplaning.
My wife and I were scared shitless, and the woman from Telsa, who was also in the car, tried to reassure us by saying “it’s ok, this is normal.”
Then we return the car to the parking lot and auto park almost took out a kid in an enclosed parking structure.
I imagine it’s become better in 4 years, but how that was street legal baffled my mind.
None of what you mentioned is in basic autopilot. Autopilot is lane keep and traffic aware cruise control only.
Let’s not get pedantic. They are part of the “enhanced autopilot” package.
https://www.tesla.com/support/autopilot
If these were called "cruise control", "adaptive cruise control", and "Blue Cruise" would it matter if the article said "cruise control" but was referring to "Blue Cruise"?
Tesla's names for these things are "Autopilot", "Enhanced Autopilot", and "FSD Beta".
At the very least, the names matter so that we can all agree we're talking about the same things.
Which is not included with the base vehicle. It’s an extra purchase.
Well in that case, the advanced autopilot features that almost killed me were totally safe.
Sure, which I consider part of FSD, which almost killed me like 3 times when I had a loaner with it active.
But that’s not basic autopilot. AP is fine assuming people pay attention.
Well when Telsa, this former employee / whistleblower, and these journalists refer to “autopilot,” they’re specifically talking about the software and hardware marketed under the “____ Autopilot” banner that Telsa uses for those features.
Some of these more advanced autopilot features clearly have issues, and it probably stems from the fact that they’re only using cameras and ultra sonic, not lidar.
In my experience with a Model 3 and AAP, when those cameras and sensors were wet, it was pretty clear that they were getting dangerous. It started raining during our test drive, so we had a before / after experience on the same roads. Once everything got obstructed with water, you could see the car’s collision detection struggle to detect other objects. Objects on the center display would erratically pop in out of view. And this was a showroom car, it wasn’t the first rain of the year, and it was behaving “normally” according to staff.
Even if basic autopilot was fine, this left such a sour taste in my mouth that I had no appetite to give that company my money. Almost dying and almost killing a kid were a big “fuck this company” for me.
My (non-tesla) vehicle can tell when the sensors are impaired by frost or mud or whatever. It flashes a warning on my dash and disables the lane-keeping and/or collision detection until next startup. Does Tesla not do that?
It was supposed to
There’s a human tendency to become complacent after a while, which presents a risk.
Can’t wait for safer-than-human self-driving technology, and know we’ll need to take certain risks of some sort to get there, but there are good arguments against “PLEASE remain fully attentive 100% of the time for this technology that will in fact only require full attentiveness in edge cases”. You might be an exception of course! But Average Meat Driver is going to slip into complacency after many, many miles of perfect autopiloting.
It’s the same as cruise control, but it’s supposed to eliminate human error. I’d argue most of the people having issues from not paying attention probably weren’t paying attention in the first place and were dangerous to begin with
My vehicle can do almost all the same stuff as "autopilot" but it turns the autosteering and cruise off if I dont touch the wheel every 30 seconds. Its all the same types of sensors,etc. And mine isn't even a luxury brand. Just the higher end trim package of a budget vehicle.
edit: actually, it's just 10 seconds before the warning and another 5 or so before it disables lane-keeping
Autopilot also shuts off with no driver input. Faster than 30 seconds too.
No.
I own a model 3 and a 2022 palisade with Lane assist and used to own a Subaru with Lane assist.
The model 3 auto steer, exit to exit EAP, and auto lane change are very different than the simple lane assist that either other cars offer and honestly after using EAP for five years, while I do use AP under specific circumstances, I have come to the opinion that it is not ready for prime time and has some major issues, especially the auto lane changing, that should have been worked out before release and I still never use that feature.
Given my background in embedded software, I honestly think the way they rolled out and advertised these features was reckless.
EAP is not based autopilot and closer to FSD. Base autopilot is on par with most manufacturers. I’d argue it’s safer than some in regards to capabilities with less common lane setups or lack of clear road lines.