Demos are presumably on-going. No live feed that we know of and no reports yet?
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Autonomy Investor Day - April 22 at 2pm ET
- Thread starter diplomat33
- Start date
Demos are presumably on-going. No live feed that we know of and no reports yet?
boonedocks
MS LR Blk/Blk 19”
just checked my Tesla account and I see that FSD is on sale for $5k.
I thought you had pre-paid for FSD.
A few things that still are unclear for me:
At some point depending on the neuron counts in your network, adding more data to the network will average out other data and cause edge scenarios to blur our. It becomes underfitted. How confident are they they got the capacity for a large enough network?
Also I didn't like Elon's input of an end-to-end video in - steering wheel/pedals out network. They said themselves a lot of situations are easier to solve with a heuristic approach, while vision and special judgmental cases are best solved with NN. I'd leave it like that.
An approach like that would also imply different datasets for each country/state, or country labels (which implies a much bigger dataset, NN and processing power in car).
Some places you can turn right on red light, some places you can't. In a heuristic approach you just turn of a flag and that if-sentence stops you from breaking the law. Your neural network tells you you're in an intersection staring at a red light waiting to turn. Sew it together and you got a good solution.
At some point depending on the neuron counts in your network, adding more data to the network will average out other data and cause edge scenarios to blur our. It becomes underfitted. How confident are they they got the capacity for a large enough network?
Also I didn't like Elon's input of an end-to-end video in - steering wheel/pedals out network. They said themselves a lot of situations are easier to solve with a heuristic approach, while vision and special judgmental cases are best solved with NN. I'd leave it like that.
An approach like that would also imply different datasets for each country/state, or country labels (which implies a much bigger dataset, NN and processing power in car).
Some places you can turn right on red light, some places you can't. In a heuristic approach you just turn of a flag and that if-sentence stops you from breaking the law. Your neural network tells you you're in an intersection staring at a red light waiting to turn. Sew it together and you got a good solution.
This here of course is the $50 billion dollar question: Is what Tesla says the same as what they know... and is what they know true.
It is all a bit moot. Failure is not really an option after this.
Yeah I didn't hear that, explicitly, at all. There was discussion of geofencing - and that if you have to geofence then you don't have L5. And that Tesla would have the capability for L5 with no geofencing in 2019. But not that it would be in the hands of customers, or have regulatory approval.
Reading your other posts - i think we're saying the same thing. Having the capability this year is amazing. The real question is when will we be able to use it.
I'm not sure why you say that. Elon has gotten away with this sort of thing over and over again.
Fair point. I wonder if even his luck would run out eventually if this is not delivered upon.
nah, not me. Must be the other guy. I only purchased EAP. (I am still hoping for the Line Waiter's discount.)
Indeed, they really laid the head on the plate now. Either they are extremely confident due to recent milestones, or just bald stupid. Karpathy got balls now (or he's planning an escape/new job
).Pathetic lie from Tesla. Per for course. Live off of myths and hype continues.
Compares the HW3 board that contains 2 NN chip, multiple GPUs and CPUs to a single Xavier chip instead of the Pegasus Board.
And the fans lap it up. Never change Tesla.
Nvidia says Tesla 'inaccurate' in self-driving comparison
Compares the HW3 board that contains 2 NN chip, multiple GPUs and CPUs to a single Xavier chip instead of the Pegasus Board.
And the fans lap it up. Never change Tesla.
Nvidia says Tesla 'inaccurate' in self-driving comparison
Don't forget that Tesla will be taking some portion of the revenue for any privately held car that is operating as a Robotaxi. So they get a fleet of cars generating revenue without additional capitol costs for the car. So they can sell the car, make some modest profit on the sale, and then receive a follow on revenue stream from the use of the car in their network without messing with any vehicle operating costs.
Of course they will also have their own cars, being returned from leasing plans, and these might be useful/necessary in booting up and validating the whole model so people are willing to provide their own cars. And in a way since these are cars coming off lease, much of the capitol expenditure for the vehicle doesn't need to come from Tesla for those either.
Eno Deb
Active Member
Karpathy sounded a lot more rational to me. Among other things, he said (paraphrasing form memory) that he considers using purely heuristics for driving policy "intractable" and that he expects end-to-end neural networks to be a better solution (as opposed to outright claiming that the problem is basically solved). But he didn't make any outlandish promises from what I remember.Indeed, they really laid the head on the plate now. Either they are extremely confident due to recent milestones, or just bald stupid. Karpathy got balls now (or he's planning an escape/new job
I noticed that too but what does it matter. Tesla said feature complete for Level 5 no geofence at the end of 2019. That is the big news here.
Comments?
That is a fair point, which I knew. Elon suggested it being like the app store and you'd be crazy not to buy a Tesla, which suggests you get to keep >50% of the profits. Of course, if they were the only game in town with FSD it would be reasonable for them to keep most of the profits and the owner to get just a modest return on investment. This would be reasonable and make more sense, but wasn't what they made it sound like.
My main point is that if this is so close to happening and being a big value generator, Elon was a lot less clear about it than they were about technical capabilities.
What are you afraid of? How many times does Elon need to say that HW2.0 can have FSD?
HW2.0 was mentioned for dash cams too.
I certainly hope it will happen. But, there’s my concern.
Karpathy did indeed sound rational.
But he's standing next to Elon talking about feature completion this year. It would be very awkward if Elon made those promises without discussing it with Karpathy and his team first. So he must be kinda in on it.
Similar threads
