Software's alleged inability to handle cross traffic central to court battle after two road deaths
Tesla knew Autopilot caused death, but didn't fix it::Software's alleged inability to handle cross traffic central to court battle after two road deaths
Didn't, or couldn't? Tesla uses a vastly inferior technology to run their "automated" driving protocols. It's a hardware problem first and foremost.
It's like trying to drive a car with a 720p resolution camera mounted on the license plate holder versus a 4k monitor on the top of the car. That's not a perfect analogy, but it's close enough for those not aware of how cheap these cars and their tech really is.
It remains to be seen what hardware is required for autonomous driving as no company has a fully functioning system, so there is no baseline to compare to. Cruise (the "4k monitor" in your anaology) just had to cut their fleet of geofenced vehicles after back to back crashes involving emergency vehicles along with blocking traffic and attempting to run over things like fire hoses.
A times B times C equals X… I am jacks something something something
Narrator: A new car built by my company leaves somewhere traveling at 60 mph. The rear differential locks up. The car crashes and burns with everyone trapped inside. Now, should we initiate a recall? Take the number of vehicles in the field, A, multiply by the probable rate of failure, B, multiply by the average out-of-court settlement, C. A times B times C equals X. If X is less than the cost of a recall, we don't do one.
Woman on Plane: Are there a lot of these kinds of accidents?
Narrator: You wouldn't believe.
Woman on Plane: Which car company do you work for?
All of the major ones. On the other hand, the Pinto's gas tank exploded less ofteb than competing models in the era, and wasn't the only design with the lowered gas tank.
Look up the You're Wrong about podcast on the Ford Pinto which is a great deep dive on car development. and product investigative reporting.
Two examples of the media creating a frenzy that wound up being proven completely false later.
In OP's case, both of these drivers failed to see a semi crossing the road right in front of them even though they were sitting in the driver's seat with their hands on the wheel. This technology certainly needs improvement, but this is like blaming every auto manufacturer when someone crashes their car while texting on their phone.
Here's how to do self driving cars in a reliable way. First, instead of cameras that try to use road markings designed for human eyes, use specially designed roads with guide rails on them to ensure it follows a safe path. Second, for added convenience, these roads could also power the cars so you don't need to stop to charge. Then we could even connect those cars together to increase efficiency. To mitigate the cost, no individual has to own them, they can stop at fixed points to pick up and drop off passengers, charging an affordable rate for each trip, or monthly/annual passes for frequent users. Maybe we could call them trains.
Probably because calling something "not real" is infuriatingly vague.
Feel free to expand on your position, I actually do want to know what "not real" means in this context.
If you mean, from a semantics perspective, that FULL means it should be a completely independent and autonomous system, bravo, you've made and won the most uninteresting form of that argument.
I mean, don't call your service something it's not? Words should have meaning? Tesla's Autopilot is very impressive, but it's not fully independent, and that's okay. Honestly if it had an accurate name people wouldn't attack it so much. Other manufacturers are gaining similar capabilities but no one is complaining that their cars aren't perfect either.
[comment clarification: I confused Autopilot and FSD]
Yeah they should have called it level 2 autonomous driving, like most other mass market car makers do (except Mercedes which have level 3 on the roads). People could then compare the different limits and clearly see what brands are or are not at the forefront of the tech.
I have nearly 20k miles on tesla's FSD platform, it works amazingly well for something thats "not real". There are countless youtube channels out there where people will mount a gopro in their car and go for a drive. Some of them like AIDRIVR and Whole Mars Catalog pretty much never take over control of the car without any drama. Especially in the past ~6 months or so of development it has been amazing.
"A bit misleading" is, I think, a bit of a misleading way to describe their marketing. It's literally called Autopilot, and their marketing material has very aggressively pitched it as a 'full self driving' feature since the beginning, even without mentioning Musk's own constant and ridiculous hyperbole when advertising it.
It's software that should never have been tested outside of vehicles run by company employees under controlled conditions, but Tesla chose to push it to the public as a paid feature and significantly downplay the fact that it is a poorly tested, unreliable beta, specifically to profit from the data generated by its widespread use, not to mention the price they charge for it as if it were a normal, ready to use consumer feature.
Everything about their deployment of the system has been reckless, careless, and actively disdainful of their customers' safety.
Everybody who has a bit of an idea what an autopilot in a plane actually does is not mislead. Do people really think that commercial airline pilots just hit the “autopilot” button in their cockpit after disengaging the boarding ramp and then lean back until the boarding ramp at the destination is attached?
I remember reading about the ethical question about the hypothetical self driving car that loses control and can choose to either turn left and kill a child, turn right and kill a crowd of old people, or do nothing and hit a wall, killing the driver. It's a question that doesn't have a right answer, but it must be answered by anybody implementing a self driving car.
I non-sarcastically feel like Tesla would implement this system by trying to see which option kills the least number of paying Xitter subscribers.
Meanwhile hundreds of people are killed in auto accidents every single day in the US. Even if a self driving car is 1000x safer than a human driver there will still be accidents as long as other humans are also sharing the same road.
When a human is found to be at fault, you can punish them.
With automated driving, who's to punish? The company? Great. They pay a small fine and keep making millions while your loved one is gone and you get no justice.
I think the whole premise is flawed because the car would have had to have had numerous failures before ever reaching a point where it would need to make this decision. This applies to humans as we have free will. A computer does not.
The driver was also not even paying attention to the road so the blame should be on him not the car. People need to learn that Tesla's version of autopilot has a specific use case and regular streets is not that.
TBH I dont watch ads ever so would you be able to show me one where they show people driving on regular streets with full self driving? I dont doubt that they advertise it im just curious how they present it and how someone might perceive.
Calling it Autopilot was always a marketing decision. It's a driver assistance feature, nothing more. When used "as intended", it works great. I drove for 14 hours during a road trip using AP and arrived not dead tired and still alert. That's awesome, and would never have happened in a conventional car.
I have the "FSD" beta right now. It has potential, but I still always keep a hand on the wheel and am in control of my car.
At the end of the day, if the car makes a poor choice because of the automation, I'm still responsible as the driver, and I don't want an accident, injury, or death on my conscience.
Tesla is producing advertisement videos (already in 2016) that said "our cars drive themselves, you don't need a driver". What you said is worthless when Tesla itself is marketing their cars like that.
Since when has autopilot, especially in 2019, ever had the ability to deal with “cross-traffic” situations? It always has been a glorified adaptive cruise with lanekeeping and has always been advertised as such. Literally the same as any other car with LKAS. Tesla’s self-driving software wasn’t released to the public until 2021/2022.
Meanwhile about 120 people died in traffic related accidents today in the US.
Yet Phoney Stark keeps on whinging about the risks of AI but at the same time slags off humans who actually know their stuff especially regarding safety.
Which uses computer vision, which is a form of AI. It doesn't have to be complex, or even work well, to be considered AI. All you need is a computer that makes decisions based on dynamic inputs and a set of rules on how to handle them.
It has a lot to do with AI. Their systems use a lot of deep learning etc to recognize agents/obstacles on the road (perception), to infer how the agents will move in the future (prediction), and to generate trajectories for their car (motion planning).
It definitely isn't Artificial General Intelligence, but it is most certainly AI.
There's like three comments in here talking about the technology, everyone else is arguing about names like people are magically absolved of personal responsibilities when they believe advertising over common sense.