Well, if you can't figure that out, then I think we will have to agree to disagree.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
FSD is a fraud
- Thread starter TessP100D
- Start date
Can't tell if serious or not. Internet is hard. Spike proteins however are very soft, and gentle. No harm other people. Much research.
Mmmm, sensing some real strong Facebook PhD scientist "do your research" vibes here.
I get a kick out of someone that would be worried about someone else getting a COVID vaccine because it might harm them, but are totally cool with the car driving down the street that your kids are playing near running FSD with a completely untrained driver behind the wheel. Humans are so weird at risk analysis.
Apparently Tesla does not agree that disclosing their proprietary data to you is required to get a lot of participation in the FSD beta program.
"Viral shedding can occur following vaccination if the vaccine contains live, weakened viruses because that type of vaccine causes immunity through viral reproduction. Examples of routinely used live, weakened vaccines include measles, mumps, and rubella (MMR); chickenpox; rotavirus; and the intranasal influenza vaccines. The oral polio vaccine (OPV) also contains live, weakened virus. While OPV is no longer used in the U.S., it is still used in some other countries. The yellow fever vaccine, while not routinely used in the U.S., is also a live, weakened vaccine. The new COVID-19 vaccines being used in the U.S. do NOT contain live virus, so they are not capable of causing shedding."
Feature Article: Viral Shedding and COVID-19 — What Can and Can’t Happen | Children's Hospital of Philadelphia
Viral shedding has been discussed in relation to masks and vaccines during the COVID-19 pandemic. Find out more about how these topics are related.
Daniel in SD
(supervised)
Yes, it was a joke!
I assume that what @drtimhill is eluding to is that people are "forced" to take the COVID vaccine through economic coercion. Not sure what any of this has to do with FSD Beta though!
no worries. Just wanted to post documented info because there actually are people who believe anything. Often its the same persons who believe that over 50+ cases (across multiple states, across state and federal level courts) that were dismissed or thrown out entirely, were all the result of a conspiracy between all 50+ judges.
My understanding of how FSD works is (overly simplified):
Cameras send their data to preprocessors coupled with hard-coded algorithms which convert the visual information into data points (currently referred to as a "bag of points").
That bag of points is sent to neural nets (NN) which are configured to analyze the data and return a decision or spatial awareness information.
NNs designed to handle spatial awareness send their data to hard-coded algorithms for visualizations (showing objects around you and the lane markers it sees on your screen), and also feed that data into other NNs that handle decision making.
NNs designed to handle decisions send their data to hard-coded algorithms to execute hardware controls, such as acceleration, deceleration, steering control, and turn signals.
The NNs do not learn - they are preconfigured matrices which Tesla tunes with new software updates. They process data multiple times per second (the cameras are 60fps, but I don't know if the NNs are processing data and sending decisions 60 times a second), and you see the outcome on your screen with a blue/grey line. Many times, that blue line is moving quite jerky around the lane, and sometimes you'll even see it move in a totally different direction for a split second before returning to a predicted route. What we're seeing is the output from the NNs at that moment. The fact that the blue line, which is consistently showing a left turn route while we're in a left turn lane, but then, for a split second, shows the blue line switch from a left curve to a sudden straight line into oncoming lanes then back to the left curve, tells us that the NN did not handle that data properly. The question is, was the data from the pre-processors (bag of points), fed to the decision NN, incorrect? Was the camera data itself flawed (poor quality camera) before being turned into a bag of points? Was the NN incorrect in its decision for that specific bag of points? Or was the hard-coded hardware control algorithm misinterpreting the output from the NN?
When you experience jerky turning (usually right turns), watch the screen and you'll likely see the blue line jerking around a bit. The control algorithms are processing the output from the decision NNs and adjusting the steering during the turn.
To Dan's point, those in FSD Beta can report when our cars do something incorrect, and the camera feeds, telemetry from the control systems, and NN input/outputs for the last several seconds are sent to Tesla for analysis. What Tesla does with this data is something we don't know. My guess is that someone reviews the video to see what the car did (jerky turn, run a red light, make an incorrect lane decision, etc), and if it meets some qualification it is sent to the FSD team for further analysis. They may take the NN data and feed it into their virtual system to see how the car reacted and then make adjustments to the NN to fix/smooth out issues. In this case, the more people on the Beta, the better. The reason it's limited to 100K right now could just be the influx of reporting data is as much as they can handle right now.
Also to Dan's point, you can see the NN processing even when you don't have FSD Beta engaged on city streets by watching the grey line. You can even press the report button when FSD Beta is not engaged, though I'm not sure how that would be processed. I guess if you see the grey line jerk suddenly into a dangerous place (like oncoming traffic), you can report it, but the camera data would show you in control and nothing wrong, but the NN data would show a potential problem. Could be valuable.
Cameras send their data to preprocessors coupled with hard-coded algorithms which convert the visual information into data points (currently referred to as a "bag of points").
That bag of points is sent to neural nets (NN) which are configured to analyze the data and return a decision or spatial awareness information.
NNs designed to handle spatial awareness send their data to hard-coded algorithms for visualizations (showing objects around you and the lane markers it sees on your screen), and also feed that data into other NNs that handle decision making.
NNs designed to handle decisions send their data to hard-coded algorithms to execute hardware controls, such as acceleration, deceleration, steering control, and turn signals.
The NNs do not learn - they are preconfigured matrices which Tesla tunes with new software updates. They process data multiple times per second (the cameras are 60fps, but I don't know if the NNs are processing data and sending decisions 60 times a second), and you see the outcome on your screen with a blue/grey line. Many times, that blue line is moving quite jerky around the lane, and sometimes you'll even see it move in a totally different direction for a split second before returning to a predicted route. What we're seeing is the output from the NNs at that moment. The fact that the blue line, which is consistently showing a left turn route while we're in a left turn lane, but then, for a split second, shows the blue line switch from a left curve to a sudden straight line into oncoming lanes then back to the left curve, tells us that the NN did not handle that data properly. The question is, was the data from the pre-processors (bag of points), fed to the decision NN, incorrect? Was the camera data itself flawed (poor quality camera) before being turned into a bag of points? Was the NN incorrect in its decision for that specific bag of points? Or was the hard-coded hardware control algorithm misinterpreting the output from the NN?
When you experience jerky turning (usually right turns), watch the screen and you'll likely see the blue line jerking around a bit. The control algorithms are processing the output from the decision NNs and adjusting the steering during the turn.
To Dan's point, those in FSD Beta can report when our cars do something incorrect, and the camera feeds, telemetry from the control systems, and NN input/outputs for the last several seconds are sent to Tesla for analysis. What Tesla does with this data is something we don't know. My guess is that someone reviews the video to see what the car did (jerky turn, run a red light, make an incorrect lane decision, etc), and if it meets some qualification it is sent to the FSD team for further analysis. They may take the NN data and feed it into their virtual system to see how the car reacted and then make adjustments to the NN to fix/smooth out issues. In this case, the more people on the Beta, the better. The reason it's limited to 100K right now could just be the influx of reporting data is as much as they can handle right now.
Also to Dan's point, you can see the NN processing even when you don't have FSD Beta engaged on city streets by watching the grey line. You can even press the report button when FSD Beta is not engaged, though I'm not sure how that would be processed. I guess if you see the grey line jerk suddenly into a dangerous place (like oncoming traffic), you can report it, but the camera data would show you in control and nothing wrong, but the NN data would show a potential problem. Could be valuable.
Mr. Jim
Member
The question is where in medicine we allow tests to be run that put non-consenting people at risk. Not coercing someone into "consent." Fully completely unaware, non-consenting people at risk just to benefit the consenting people (or companies with Trillion dollar valuations). Just like the pedestrians on the road around cars running FSD.
People are arguing that FSD being publicly tested is fine and ethical, but it's interesting that they would likely not feel the same if this was medical or aircraft.
Of course they don't! And because they don't, I'm also allowed to be skeptical that their deployment of FSD into a bunch of cars on the road is actually part of their development, and not just marketing.
My guess is that he was describing the vaccine during trials/testing. A person receives the vaccine during a trial, and is exposed to COVID. But if the vaccine didn't work well (which happens during a trial), the person could become infected. If they come into contact with others, they could spread the virus even if asymptomatic. I have no idea how vaccine trials are conducted for things like COVID - are the testers sequestered during the entire testing phase to eliminate exposure to others?
The idea of testing FSD Beta on public roads, and how that presents a risk to others who did not agree to test, is just a fact of life. No matter how careful people are, how isolated the testing environment, or how confident the engineers are in the results, the final product has to be released to the world. It eventually has to go on public roads with real people around. And mistakes/accidents can happen. The goal is to minimize those as best as reasonably possible. That's a key word in our society - reasonable. As long as reasonable measures are being taken to protect others, the program should continue. Where regulators step in and force changes are when unreasonable measures are taken.
We all take a risk every time we leave the house. We assume that car companies have taken reasonable steps to ensure the safety of their cars while on the road. We assume parts companies have taken reasonable steps as well. But sometimes things happen out of our control. A small piece of metal could shred a tire at high speed, or a small defect in that one tire causes a failure, resulting in the loss of life (either another driver, a passenger, or a pedestrian). A control system in the car could malfunction (there was a rash of uncontrolled acceleration in some cars a few years ago - Toyota had this issue if I recall), resulting in loss of life. Boing 737 Max airplanes had design flaws that caused problems and significant loss of life. It's tragic and we hopefully learn as we go, but we can't stop progress. We just try to make reasonable decisions and choices to help progress move forward as smoothly and safely as we can.
The question is: Is Tesla being reasonable with their testing? Regulators, like NHTSA, will tell us soon as they continue to investigate not just Tesla, but other ADS and ADAS systems.
That's funny. First you state that Tesla 'needs' to disclose their trade secrets. Now you that they don't. Make up your mind.
I'm skeptical that your stated concern for the 'unsuspecting public' might have an undisclosed motive.
Daniel in SD
(supervised)
Vaccinated people show symptoms earlier because their immune systems react faster to the infection. I think this was an expected result but I'm not sure. If the people in the trial were sequestered they wouldn't get COVID...
I think the standard for FSD Beta should be whether it makes the roads safer. It's very hard to determine that though. Comparing it to other risks is silly because you can eliminate the risk of FSD Beta with the flip of switch. I believe it's possible to test autonomous vehicles without putting anyone at additional risk.
Daniel in SD
(supervised)
He's saying that Tesla doesn't want to disclose trade secrets...
PACEMD
Active Member
So we should immunize the general public against FSD-beta? Is that the consensus? I'm cool with that.......
I think most do. Tesla tests on internal tracks, in virtual environments, and with limited releases (engineers and employees first). Waymo and Cruise do as well. Once they are confident in their test environment, they release onto public roads for additional testing. Waymo in San Francisco is still testing - you cannot just book a ride like you can in Arizona. If it's still in testing, and not available to the public, why is Waymo running them on city streets? Because that's the next step.
Daniel in SD
(supervised)
If Waymo test drivers were caught doing what some FSD Beta users do they would be fired immediately.
It blows my mind how many FSD Beta testers publish videos of themselves breaking the law on YouTube. If they get in a collision later that can be used as evidence against them in a civil or criminal trial...
The 5th amendment does not protect you if you choose to incriminate yourself!
So, first off let's skip the "vibes". The point I was making, to be very clear, was that there are always unanticipated risks that may involve bystanders. Every vaccine carries with it a (vanishingly small) risk that the vaccine may itself cause side effects that are transmittable. It's VERY unlikely, but possible (think prions e.g. mad cow disease). At what point are these risks acceptable to the population at large? Partly, when they reach a very low absolute threshold, but partly when the overall good far exceeds the possible bad. (The anti-vaxxers messed up the good vs bad argument to the point of absurdity of course.)
Which is really what the argument about FSD beta, and automonouns car testing in general, is about in the first place. There is, of course, a risk that an innocent bystander is hurt by FSD beta (either financially or physically). Is that risk low enough to warrant the public beta being pursued by Tesla? You clearly think the answer is NO, but you have not quantified what risk is acceptable (zero is not a realistic answer, NOTHING has zero risk). Your argument boils down to "the risk is unacceptable", but you have not quantified (or even qualified) the risk, and until you do so, you are simply stating a random opinion, rather than making a considered argument.
My opinion (perhaps random too) is that autonomous cars will end up significantly safer than human drivers .. not because they are super-smart but simply because they wont do all the stupid things humans do that cause most accidents (drinking, texting, falling asleep etc). So is that ultimate goal worth some level of risk today? THAT is what is being debated here, and that is a good debate to have, but can we move on from the name-calling? It's not constructive, and doesnt contribute anything except feeding the trolls.
Preach it, brother Dan! I take FSD Beta very seriously. I wish everyone else did too.
