P.S. If i was using google, I could probably look that up, but I don't care to. Editing a post to add clarity or detail doesn't have to go through google. That one happened to come from AGEN 265. All thanks to the late Dr. Lepori.
It absolutely does do that. Every post you make is just your ignorance being put on full display. Total lack of understanding.
Please find one example of a computer that has successfully accomplished a goal that it wasn't programmed to accomplish.
I won't wait, you'll be looking forever.
It is not programmed. It is trained. Again revealing your lack of understanding. They switched from programmed controls in 2022 to capturing video data and driver input married into their training system removing explicit instruction. The capabilities developed between 2017-2022 from an instruction set system was improved by double in 12 months (released in 2023) by removing human coded instruction from the system. Just a watch and learn process.
The rate of improvement of time between intervention accelerated from 2023 to 2024, 3xing the average period between safety interventions. From 2024 to 2025 it again accelerated, 4xing the period between safety interventions.
If the rate of improvement continues with the upcoming v15 release (late 2026) it will easily be good enough to deploy as a ride hailing service. The cause of the acceleration in the rate of improvement is the quality of edge case discovery improving due to larger data sets and more training capacity.
Number of miles driven on FSD is roughly doubling every year, with this year likely to see more than doubling due to recent approvals in Australia, full approval in China, and pending approval across Europe. Currently at a ~30 million miles/day usage rate. As they increase the period between nags (eventually removing them under some circumstances, potentially end of this year) the quantity of consumption will rise. Would expect ~70 million miles per day of data availability to train the model by next year.
If you want to actually understand the topic rather than use your shoot from the hip extremely flawed intuition watch this video.
Traction motors, maybe? Traction motor and hub assembly? Subframe assembly with two sets of each?
Regardless, the drivetrain starts at the battery on an EV.
You sure speak with a lot of authority on something you "aren't sure of".
You also never answered my question so I'll ask again. Are you suggesting that every part of the drive unit be replaced as a regular maintenance item at 120k miles or 8 years?
No it's a lot of words to explain to you that your framing does not align with reality. It is not programmed. It is trained. It responds to scenarios it has never specifically been exposed to through a probabilistic determination of the correct path to take. Same as you do. Does that guarantee the path is the correct one? No. But you don't have a guarantee your response is correct either and the decision rate of the computer is 36 times a second, which is much faster than yours. Meaning if it makes a prediction or an input and the prediction or input is incorrect it will correct the failure faster than you can.
The model is exposed to millions times more scenarios than you. Those scenarios are capable of being extrapolated to similar, but not the same scenarios. That extrapolation allows it to respond to scenarios not in its training set, and the quantity of scenarios not in its training set declines as the training data set population grows.
No it's a lot of words to explain to you that your framing does not align with reality. It is not programmed. It is trained. It responds to scenarios it has never specifically been exposed to through a probabilistic determination of the correct path to take. Same as you do. Does that guarantee the path is the correct one? No. But you don't have a guarantee your response is correct either and the decision rate of the computer is 36 times a second, which is much faster than yours. Meaning if it makes a prediction or an input and the prediction or input is incorrect it will correct the failure faster than you can.
The model is exposed to millions times more scenarios than you. Those scenarios are capable of being extrapolated to similar, but not the same scenarios. That extrapolation allows it to respond to scenarios not in its training set, and the quantity of scenarios not in its training set declines as the training data set population grows.
No. He says it can't respond to things it's not programmed to respond to. If you possessed reading comprehension capabilities you would have determined he is saying it is programmed to respond to some scenarios, but not all. I'm saying it's not programmed to respond to scenarios at all, it is trained to respond to scenarios. It learns through commonalities.
It is exposed to scenarios, recognizes commonalities between those scenarios and the inputs that successfully navigated in similar scenarios and then predicts the appropriate input. If the prediction is wrong it corrects the prediction.
That makes it capable of dealing with more scenarios than are in its training set. Things that it has never seen before. The probability of any specific scenario substantially deviating from its training set decreases as the quantity of data in the training set increases.
You are doing here what you just accused other people of doing. Acting snarky to get a rise out of others.
Merely pointing out that any and every statement made by the prime disagreer on this thread is attacked. He was indeed right on the programming. Your snark came in trying to create a gotcha by defining it as "learning."
No. He is not right. He is so fundamentally incorrect that he can't even appropriately frame a critique. It's like telling an EV owner they forgot to account for engine oil changes when discussing maintenance costs. It is that incorrect.
This is the pattern of these threads. Some fundamentally uninformed person believes they understand a topic (both of you being absolutely egregious offenders) claiming that they've made a valid critique and then when it's pointed out it's not even the in the realm of a correct understanding the claim in return is that the critique is being ignored. It's not, it's just being pointed out that you know so little that you don't even understand how to lay out a critique.
Somewhere my great-great grandfather is lying in his grave still arguing that automobiles will never replace the horse and because there are not enough places to get gas and because there are not roads that go everywhere that you need to go, whereas a horse will. This is an actual family story handed down over the generations of an argument between my great great grandfather and my great grandfather.
Other than having a vested financial interest in the perpetuation of oil extraction, there's not much in the way of a valid argument for having an ICE vehicle as your daily driver, unless you just like slow and noisy as opposed to fast and quiet. I got my first Tesla after watching soccer moms in Teslas blow Corvettes off the track in a standing quarter mile.
No. He says it can't respond to things it's not programmed to respond to. If you possessed reading comprehension capabilities you would have determined he is saying it is programmed to respond to some scenarios, but not all. I'm saying it's not programmed to respond to scenarios at all, it is trained to respond to scenarios. It learns through commonalities.
It is exposed to scenarios, recognizes commonalities between those scenarios and the inputs that successfully navigated in similar scenarios and then predicts the appropriate input. If the prediction is wrong it corrects the prediction.
That makes it capable of dealing with more scenarios than are in its training set. Things that it has never seen before. The probability of any specific scenario substantially deviating from its training set decreases as the quantity of data in the training set increases.
You are doing here what you just accused other people of doing. Acting snarky to get a rise out of others.
I never said it was programmed to respond to anything. You've made that inference. If it is programmed to run a BFS spanning tree so be it. It's still subject to GIGO.
Somewhere my great-great grandfather is lying in his grave still arguing that automobiles will never replace the horse and because there are not enough places to get gas and because there are not roads that go everywhere that you need to go, whereas a horse will. This is an actual family story handed down over the generations of an argument between my great great grandfather and my great grandfather.
Other than having a vested financial interest in the perpetuation of oil extraction, there's not much in the way of a valid argument for having an ICE vehicle as your daily driver, unless you just like slow and noisy as opposed to fast and quiet. I got my first Tesla after watching soccer moms in Teslas blow Corvettes off the track in a standing quarter mile.
Electric is only "fast" when slow. Get the RPM up and the torque falls, unlike ICE or steam where the limit is expansion velocity. That flat HP curve of electric is a major hindrance. Want to expand on what happens if you compare them on a roll-start?
Somewhere my great-great grandfather is lying in his grave still arguing that automobiles will never replace the horse and because there are not enough places to get gas and because there are not roads that go everywhere that you need to go, whereas a horse will. This is an actual family story handed down over the generations of an argument between my great great grandfather and my great grandfather.
Other than having a vested financial interest in the perpetuation of oil extraction, there's not much in the way of a valid argument for having an ICE vehicle as your daily driver, unless you just like slow and noisy as opposed to fast and quiet. I got my first Tesla after watching soccer moms in Teslas blow Corvettes off the track in a standing quarter mile.
Electric is only "fast" when slow. Get the RPM up and the torque falls, unlike ICE or steam where the limit is expansion velocity. That flat HP curve of electric is a major hindrance. Want to expand on what happens if you compare them on a roll-start?
This is total nonsense. All of it. ICE torque falls at high RPM. The rest I'm not bothering with.
I'll take a Tesla at any roll speed and leave you in the dust in your ICE vehicle. The roll-start is the EV's strong suit, not its weakness.
Then yes. Expect to start replacing stuff when the warranty is up. Designing any one component to outlive the others is a waste of money.
You guys have to remember TexasRebel is coming from the viewpoint of one of those posters that brags about their truck with 230,000 miles on it. Which honestly, I never understood. It's like bragging online about how inexpensive your mobile home is. Sure. Nice flex. Who the hell worries about wheel bearing maintenance at hundreds of thousands of miles, lol.
You said a computer cannot respond to scenarios it is not programmed for. If I'm misinterpreting you then feel free to correct, but your claim seems to be that if you stuck a dinosaur in the road in front of a Tesla it would not stop because it is not programmed to respond to a dinosaur. Except it would. Because even though dinosaur is not in its training set the world model is sophisticated enough to understand it is an object in its path. If the dinosaur ran towards the car it would be smart enough to reverse.
Next car will likely be an EV. However, I'm happy to wait for the Toyota (or similar) batteries that are hitting production in 27/28, effectively eliminating the risk of leaks, thermal runaway, and battery fires. China is going to require these batteries, hopefully all EVs will make this change shortly.
Then yes. Expect to start replacing stuff when the warranty is up. Designing any one component to outlive the others is a waste of money.
So it's your contention that all Tesla owners, as a good and standard maintenance practice, should pre-emptively replace wheel hubs, the battery, and the motors at 120k?
This demonstrates a fundamental misunderstanding of motors and engines.
On your part yes, along with a fundamental misunderstanding of AI and other subject matter.
And engineering. The idea that you should replace any and every non-serviceable part in a piece of machinery as soon as the warranty expires as a maintenance practice is one of the most absurd things I've ever read.
No. A computer cannot do something it is t programmed to do.
Except it is programmed to do something different than what you think it is programmed to do. It is not programmed on specific scenarios. Unless you think Tesla programmed it to recognize a mural in the middle of the road designed to match the background. It can deal with novelty through commonalities.
If FSD saw a dinosaur in the middle of the road it would stop. If a dinosaur walked towards the Tesla it would reverse.
Then yes. Expect to start replacing stuff when the warranty is up. Designing any one component to outlive the others is a waste of money.
So it's your contention that all Tesla owners, as a good and standard maintenance practice, should pre-emptively replace wheel hubs, the battery, and the motors at 120k?
Is that correct?
There is reactive maintenance which tends to get expensive and happen all at once, and there is preventative maintenance.
I'm not telling anyone how to live. I just know that "dumbass" charges do exist in vehicle service.
Your claim is that it can't deal with novelty. That it can't exceed human capability. It does. Is it at this stage universally superior? No, but directionally it is on a path to exceed the capabilities of human drivers on the whole.
You have previously made the claim that the quantity of scenarios are too large for a computer to exceed human capability.
And you're being purposefully non-specific to avoid falsifiability.
