NerdUno
Member
It's worth mentioning that the Prius technology was actually based upon a GM patent that GM decided wasn't worth pursuing. Every sofa quarterback always looks like a genius.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Elon Musk is 'full of crap'
- Thread starter ishareit
- Start date
B
Bingo. The human mind has processing capacity that no computer system will match for a long while, and with a lifetime of training. I do believe FSD is possible, but we will need to make up for lack of processing with additional sensors, infrastructure changes and vehicle/vehicle, vehicle/infrastructure communications.
Bingo. The human mind has processing capacity that no computer system will match for a long while, and with a lifetime of training. I do believe FSD is possible, but we will need to make up for lack of processing with additional sensors, infrastructure changes and vehicle/vehicle, vehicle/infrastructure communications.
While it’s true that GM filed for and received a patent that describes the Prius hybrid design, the evidence is that Toyota independently developed and decided on their design before GM’s patent was approved and published where they could have taken notice of it.
GM’s and Toyota’s independent but overlapping hybrid work was derived from an earlier design patented and described publicly by TRW back around 1970 before digital controllers and battery tech allowed for it to succeed in a consumer production vehicle. TRW’s earlier design was not identical but it laid out the basic concept of a battery pack together with an ICE, 2 electric motors, and the drive shaft attached to a planetary gear set.
Revenge of the Two-Mode Hybrid - HybridCars.com
This is a very misleading statement, since the human mind operates in an entirely different manner than a computer does. This can be seen in the very different kinds of tasks that each excels at. Humans are much better at pattern recognition, while computers can analyze a decision tree much faster than a human can. Further, in a high-stress or high-fear situation, or when split-second decisions are needed, humans do a very poor job unless they've spent thousands of hours developing the muscle memory for exactly those kinds of situations. When an accident is imminent, a human is likely to step on the wrong pedal, or utterly fail to notice that in swerving to avoid the car approaching from his left, he's turning directly into the child on his right.
A computer will of course need sensors all around the car, but with those, it will never fail to take all obstacles into account, and its reaction time will be a hundredth or less that of a normal human driver.
And with the driver free to pursue other activities, there will be much less incentive for speeding. We have sufficient computing power today for full self driving. Right now it's just a matter of designing adequate software. Remember, a self-driving car does not have to be perfect. It just has to have fewer accidents than a human driver. This is not a trivial task, but it is not insurmountable.
So that’s all true and speaks to the potential of FSD systems in simple situations. It’s why autopilots are so successful in aircraft. But, even in aircraft, humans are still involved because there are situations that arise that the computer can't easily deal with... weather, interacting with ATC, aircraft issues, emergency situations. And aircraft operate in an environment much easier for automation to deal with, than what you find on typical urban road roads.
The human mind is still vastly better at dealing with novel and complex situations, that demand a wider contextual understanding of the situations. And it will be so for a long while, I'm quite convinced.
There are ways to work around these limitations... redesigning the infrastructure to be automation centric rather than human centric. And multiple redundant sensors will certainly help.
The argument "we have two eyes, so clearly cameras alone will work" is correct - with sufficient processing power. We just ain't there yet.
True. We're not there yet. Which is why Teslas in 2017 don't have FSD. Only time will tell how long it will be. We've got different estimates. I still think we have sufficient computing power now, and only need the software. Not a trivial task, but some pretty smart people are working on it.
The human mind is capable of remarkable adaptation to novel situations, but also far too susceptible to distraction and panic. Computers will, as I've said before, make mistakes that humans never would. But they'll make far fewer of the mistakes that humans make all the time. The result, I'm convinced, will be a safer roadway.
When? I think not too far away. Maybe a decade. Either way, I'm looking forward to the limited amount of additional safety that I'll have with AP on my Model 3. I won't be taking my hands off the wheel or my eyes off the road, but I'll feel a bit safer knowing that in addition to my eyes, the computer will also be ready to assist.
Right... and aircraft autopilots are vastly better at IFR flying than humans. They don't get distracted or fatigued and they don't panic. And they've made the skies safer for it. But that's a simple situation - very analogous to what the current generation of Tesla autopilots do.
Sometimes people don't realize how vast the difference is between what the current AP does and what FSD is supposed to do. Both may be based on neural networks (to some degree), but that's where the similarity ends. They're as different as a 1960s transistor radio and the most powerful supercomputer. Yes, they're both based on semiconductors. But they're completely different in what they're trying to accomplish and in the way they function.
The media has been drawn into the impending automated vehicle world by automated lane-following - something which was held in the realm of sci-fi for many years. But that success is entirely driven by the sudden success of convolutional neural networks in image processing. While impressive, it's quite unrelated to what is required to perform the more complex planning and decision making required for FSD.
I fear that we're going to be stuck in the "almost there" trap for a long time. But, I do agree that there are safety benefits to be gained from the simpler and more achievable systems - L3 perhaps - that are coming.
The point of IFR flying being programmable yet FSD being a far more difficult computer science problem is a great illustration of the current situation.
It is why 12 year olds should learn programming. It avoids later b.s. discussions about Tesla Network and paying for the model 3 through hiring the car out.
It is why 12 year olds should learn programming. It avoids later b.s. discussions about Tesla Network and paying for the model 3 through hiring the car out.
NerdUno
Member
You need to read some more threads on this forum if you're counting on AP2 to assist and make your Model 3 drive safer. Nothing could be further from the truth at this juncture.
I'm aware that AP is far from perfect, which is why I'm not going to be taking my hands off the wheel or my eyes off the road. But AP does have a suite of features that help to enhance safety. We've seen the tragedy that results when someone treats AP as though it was FSD. And there is indeed the risk of people getting overconfident and taking their hands off the wheel because they have AP. But I think the various warnings and emergency responses will help when used appropriately.
When you say we are going to be "... almost there... for a long time," what kind of time frame are you thinking? Because for me, my estimate of ten years is a long time. We might actually be agreeing.
When you get out past a decade, it’s all pretty speculative. True L5 - I struggle to see it in the next 20 years. You will likely have it in very restricted environments sometime before that.
Aren't you cute - you need attention from me - any kind of attention - I don't converse with you for 24 hours and so you actually resurrect a post several days old (and you do nothing except post here - so i know you saw it days ago) and resurrect it to try to provoke me. And funny if it isn't about @Bladerskb. As for my compliments to you? That's a one time deal - the first ones are free - print 'em out and put 'em under your pillow to have nice dreams. But I'll give you my paypal address and if you show me some appreciation I'll post more. That's how this game works - but then, you know that already.This really bothers you doesn't it?
Actually, @BobtheV resurrected that post. I returned to this thread to see it as a new post and responded to his quote of you (for whatever reason I had missed it at the time). It is right there above my reply.
Barklikeadog
Active Member
We are there. The new Tesla offerings have full self driving capability
I'm not sure "a car you can drive yourself" is the same as "a car that drives itself".
They have the hardware that Tesla believes will be sufficient to implement FSD when the software is ready. I think it's overly optimistic to think that a 2017 Tesla will be FSD capable any time soon.
Similar threads
