fka ftc said:

---

Teslag said:

---

fka ftc said:

---

What's going to be great are the guys that figure out how to cover cars and other items with laser absorbing / reflecting materials to confuse the computers.

What happens on a Tesla if a gravel truck passes you and dust / gravel fly off the truck confusing the sensors and taking out the sensors. Does the car swerve, stop, keep going? I have yet to see a computer that can process at the speed it requires to observe and react. Most humans are capable of successfully navigating incredibly complex inputs in essentially real time.

---

If anything obstructs the autopilot cameras an alarm sounds and instructs the driver to take over.

---

That's probably the smartest line of code in the whole car then.

---

I'm sure that Elon Musk and some of the greatest engineering minds in the world at Tesla appreciate your validation.

As usual, your rants are missing the point. You're on tangents to the point while you think you're down the middle. And because of that, you say things that give the impression that you actually believe humans have a faster reaction time than computers do. It's not a reaction time issue.

It's a neural net comprehension issue. And no, the neural nets don't understand, nor have an answer for everything at this point.

aggievaulter07 said:

---

As usual, your rants are missing the point. You're on tangents to the point while you think you're down the middle. And because of that, you say things that give the impression that you actually believe humans have a faster reaction time than computers do. It's not a reaction time issue.

It's a neural net comprehension issue. And no, the neural nets don't understand, nor have an answer for everything at this point.

---

Thanks, whew I needed that clarification. Thanks!

Its always helpful for people to tell me how to think or explain to me what I know and do not know.

Ability to do a mathematical calculation at a speed faster than our brain is only tangentially related to reaction time.

Reaction time involves understanding an undefined set of inputs and circumstances combined with an immeasurable and incalculable amount of experience in same and similar situations, go subconsciously through those prior experiences and reaction times and translate that to muscle or motor movement to result in the most desirable or at the least an acceptably desirable outcome.

So no, a computer does not have a faster reaction time. It can calculate math faster than a human, that is all.

Thanks for trying though.

On a board absolutely piled to the brim full of people who have zero clue what they are talking about (myself included, at times), you're absolutely the most confident in the things that you say that are just objectively wrong.

Congrats on that particular trophy.



Computers are thousands of times better and faster at this than humans. I don't know what else to tell you.

To be fair, computers are faster at A reaction, not necessarily THE BEST reaction.

For something as straight forward as shifting a transmission, computers are clearly react faster AND better than humans.

For collision avoidance it's not so clear, right now.

cecil77 said:

---

To be fair, computers are faster at A reaction, not necessarily THE BEST reaction.

For something as straight forward as shifting a transmission, computers are clearly react faster AND better than humans.

For collision avoidance it's not so clear, right now.

---

No, no. Other posters have their peculiar definitions on reaction times and it only applies to computers and math transactions. But you have to be willing to substitute "input" for "stimulus", even though they are not the same thing.

The problem with a lot of your debate is you keep saying other people have their own definitions of things when it's actually you that has your own definition. Everybody else seems to genuinely try to use broadly accepted definitions like the one I cited above.

It interesting are there are studies out there looking at human reaction time vs computer "reaction" time.

Some of those even show that instances where the car is self driving and the situation requires human intervention, the overall reaction time to the situation was SLOWER than if driver was operating the vehicle.

Physical limitations of what a car can and cannot do make an actual advantage in reaction time by computers v humans moot. For instance, if in order to avoid the accident a 20 millisecond reaction time is required and a computer can deliver a solution in 100 milliseconds and a human at 300 milliseconds, then both result in the accident occurring. Now maybe it helps overall in the outcome and severity, but conceivably it can be made worse.

This information is out there and requires no special understanding or definitions. Computer processing time and reaction time are related but not the same. And even when comparing apples to apples, it can still be completely irrelevant.

And we could come up with plenty of situations where a human will determine a better outcome than a computer. The exponential variables are a huge problem for even the best computers. But they will always fall short, just like humans.

The problem is no lawyer nor lawmaker is going to put their neck out and say "well, computer may have gotten it wrong here but gee willikers its getting better and in a few years less people will be run over".

I fully understand what you are saying in processing time, but you fail to consider a broader concept and set of circumstances. Fortunately there are those that are paid to consider all of these things rather than just computer power.

cecil77 said:

---

To be fair, computers are faster at A reaction, not necessarily THE BEST reaction.

For something as straight forward as shifting a transmission, computers are clearly react faster AND better than humans.

For collision avoidance it's not so clear, right now.

---

If you read my posts, you'd see that I'm not claiming computers always reach better conclusions than humans. I specifically said with self driving it's a neural net issue, not a speed (reaction time) issue.

It's a "reaching the right or wrong conclusion" issue. A computer can reach either the right, or wrong conclusion infinitely faster than a human can do either.

Reaching the wrong conclusion is absolutely not the same thing as reacting slower. fka can't seem to separate the two.

aggievaulter07 said:

---

cecil77 said:

---

To be fair, computers are faster at A reaction, not necessarily THE BEST reaction.

For something as straight forward as shifting a transmission, computers are clearly react faster AND better than humans.

For collision avoidance it's not so clear, right now.

---

If you read my posts, you'd see that I'm not claiming computers always reach better conclusions than humans. I specifically said with self driving it's a neural net issue, not a speed (reaction time) issue.

It's a "reaching the right or wrong conclusion" issue. A computer can reach either the right, or wrong conclusion infinitely faster than a human can do either.

Reaching the wrong conclusion is absolutely not the same thing as reacting slower. fka can't seem to separate the two.

---

I am certain I can give you a faster wrong answer than any computer 11 times out of ten.

Sounds like my wife claiming victory because she yelled first regardless of whether she is right or wrong.

Did you really mean to say the bolded part? Because that makes no practical sense whatsoever in relation to automating human processes.

At 0:50 in the video, Tesla with FSD Beta 11.3.1 hesitates on a left hand turn, then almost turns in front of vehicle approaching faster than anticipated. Has a few other times the driver has to override the system.

I am not sure of all the technology Tesla uses but Ouster Lidar's sensors can pick up around 5.2 million points per second. Compare that to the average human reaction of 150-300ms and we have nothing on what the sensors are capable of. What is holding the vehicles back is the programming and not having the ability to see the little details like a biker looking over their shoulder since they want to turn

bmks270 said:

---

Friend got a Mach E and says its driver assist is very advanced. Maybe not Tesla level, but he says others are catching up, he was impressed.

---

The Ford driver assist is so far behind it can't hold a 30% curve on the highway going 65. It's barely above cruise control. They are not catching up. They are getting further behind. Tesla FSD stops at red lights and stop signs, turns left and right. Turns left after cars have passed on an unprotected left turn. V11 just released to the public merging their highway and city driving stacks so now they are coding and testing a single version of the software making improvements scale at a much more rapid pace.

tk for tu juan said:

---

At 0:50 in the video, Tesla with FSD Beta 11.3.1 hesitates on a left hand turn, then almost turns in front of vehicle approaching faster than anticipated. Has a few other times the driver has to override the system.

---

It's driving 40 mph down a two lane highway where the speed limit should be much higher. Some of the speed changes would annoy the hell out of me if I were following that guy. Not very impressed, looks like self driving vehicles still have a LONG way to go.

P.U.T.U said:

---

I am not sure of all the technology Tesla uses but Ouster Lidar's sensors can pick up around 5.2 million points per second. Compare that to the average human reaction of 150-300ms and we have nothing on what the sensors are capable of. What is holding the vehicles back is the programming and not having the ability to see the little details like a biker looking over their shoulder since they want to turn

---

This is where humans still win, predicting behavior of other drivers or knowing when to be cautious from blind turns, traffic congestion, weather, etc.

bmks270 said:

---

P.U.T.U said:

---

I am not sure of all the technology Tesla uses but Ouster Lidar's sensors can pick up around 5.2 million points per second. Compare that to the average human reaction of 150-300ms and we have nothing on what the sensors are capable of. What is holding the vehicles back is the programming and not having the ability to see the little details like a biker looking over their shoulder since they want to turn

---

This is where humans still win, predicting behavior of other drivers or knowing when to be cautious from blind turns, traffic congestion, weather, etc.

---

Yeah Rogan had an automated car expert on years ago and he said it would take a while for predictive behavior and at that time only Tesla had the technology and proper sensors to do predictive behavior.

bmks270 said:

---

P.U.T.U said:

---

I am not sure of all the technology Tesla uses but Ouster Lidar's sensors can pick up around 5.2 million points per second. Compare that to the average human reaction of 150-300ms and we have nothing on what the sensors are capable of. What is holding the vehicles back is the programming and not having the ability to see the little details like a biker looking over their shoulder since they want to turn

---

This is where humans still win, predicting behavior of other drivers or knowing when to be cautious from blind turns, traffic congestion, weather, etc.

---

This is where spotting the lexus/buick driven by an old person/asian driver, or a teen male driver with a girl in the passenger seat etc. is something AI has not accounted for-yet.

Or the guy ahead of you on Friday evening who is bouncing from one side of the lane to the other, sometimes a bit over.

Homeboy sitting in his basement coding does not even have that situation on his "radar", pun intended.

What's funny is the police will likely have a AI powered device on their cruisers to identify drunk drivers but the EV companies cannot use it because of muh privacy.

Very true/good point.

bmks270 said:

---

P.U.T.U said:

---

I am not sure of all the technology Tesla uses but Ouster Lidar's sensors can pick up around 5.2 million points per second. Compare that to the average human reaction of 150-300ms and we have nothing on what the sensors are capable of. What is holding the vehicles back is the programming and not having the ability to see the little details like a biker looking over their shoulder since they want to turn

---

This is where humans still win, predicting behavior of other drivers or knowing when to be cautious from blind turns, traffic congestion, weather, etc.

---

From a completely outside perspective, the bridge here is a standardization of networking between individual vehicles that also retains a level of anonymity. If (when) the cars can communicate, sort, and coordinate with each over short periods of time, that should reduce the need for predictability. But again I'm just spitballing.

fka ftc said:

---

Or the guy ahead of you on Friday evening who is bouncing from one side of the lane to the other, sometimes a bit over.

Homeboy sitting in his basement coding does not even have that situation on his "radar", pun intended.

What's funny is the police will likely have a AI powered device on their cruisers to identify drunk drivers but the EV companies cannot use it because of muh privacy.

---

Yes they do... And they're not coding for that specific instance, they use a training algorithm to optimize the way the vehicle drives. They query their fleet of vehicles for instances that match their parameters, feed those instances through their algorithm, and teach the system how to respond to those circumstances. It is not a guy driving down the road that is moving from side to side in their lane that is the challenging part of developing self-driving software. The vehicle already shifts its location in the lane to account for individuals riding the lane lines. Of the videos I've watched for FSD beta, the primary failure points are intersections for unprotected lefts, rights on red, roundabouts and non-standard road design situations. A vehicle not maintaining its lane is not a complicated scenario to account for.

Their system has a 30 petabyte cache of video, expanding to 200 petabytes soon, and their fleet of vehicles generates 123 million miles of queriable information every single day. That could increase to more than one billion daily miles by the end of the decade. Their system is going to experience driving scenarios that you can't even dream of. Fail states get reported to their system either due to impacts, interventions, disengagements or manual reporting by users.

Which one you is this:

Tesla Model 3 Hacked in Less Than 2 Minutes at Pwn2Own Contest

https://www.darkreading.com/vulnerabilities-threats/tesla-model-3-hacked-2-minutes-pwn2own-contest

Quote:

---

The attacks gave them deep access into subsystems controlling the vehicle's safety and other components.

One of the exploits involved executing what is known as a time-of-check-to-time-of-use (TOCTTOU) attack on Tesla's Gateway energy management system. They showed how they could then - among other things - open the front trunk or door of a Tesla Model 3 while the car was in motion. The less than two-minute attack fetched the researchers a new Tesla Model 3 and a cash reward of $100,000.

---

Model 3 RWD price dropped by $2,000. Now starts at $39,990. They are doomed

tk for tu juan said:

---

Model 3 RWD price dropped by $2,000. Now starts at $39,990. They are doomed

---

That's cheaper than the new Nissan Ariya. Interesting.

bmks270 said:

---

Tesla is losing market share, and 60 million shares of Tesla stock was put up as loan collateral for Twitter purchase.

With the value of both tanking, and SpaceX struggles, Musk's equity driven networth may have peaked.

---

This is a great piece of TA analysis that has really stood the test of time so far.