Waymo

[mikes_fsd]

Of course. That's why about $100 billion has been invested in autonomous driving tech.
Unfortunately it turns out to be an extremely difficult problem to solve.

I sure am glad you are not the one trying to solve this problem.

 

[diplomat33]

The logical conclusion of seeing technology doing better in portions of the driving tasks is to replace all the driving tasks with technology (aka autonomous driving)!

That's why Waymo skipped L2 and went straight to autonomous driving. As Krafcik said, the Waymo Driver is a complete replacement for a licensed human driver.

 

[Daniel in SD]

I sure am glad you are not the one trying to solve this problem.

Why? I didn't say it was a hard problem for me to solve. I was only observing that it has proven to be a hard problem for all the engineers that have been working on it for a decade.

 

[mikes_fsd]

Why? I didn't say it was a hard problem for me to solve. I was only observing that it has proven to be a hard problem for all the engineers that have been working on it for a decade.

It is the mindset that is the problem.
Hard to overcome when you've convinced yourself that task at hand is "extremely difficult"

That's why Waymo skipped L2 and went straight to autonomous driving. As Krafcik said, the Waymo Driver is a complete replacement for a licensed human driver.

Waymo didn't skip $#!t they still don't have a viable product!

 

[mikes_fsd]

I was only observing that it has proven to be a hard problem

The only folks who "prove" that a problem -- within the boundaries of physics -- is "extremely difficult" are the ones that are too lazy to take on the problem to solve the problem.

They would rather create consortiums and regulatory bodies or some other crutch than work on the solution.
Oh, they use the "think of the children" or "who should AI chose to kill" click bait arguments to pave the way as well, but it is never to actually solve the problem.

 

[diplomat33]

It is the mindset that is the problem.
Hard to overcome when you've convinced yourself that task at hand is "extremely difficult".

The whole industry of thousands of top engineers have been working on FSD for years and say it is a super hard problem. But I guess they all just have the wrong mindset. Silly them for not realizing what a super easy problem autonomous driving really is.

Time and time again, you prove you don't know what you are talking about.

Waymo didn't skip $#!t they still don't have a viable product!

Wrong. They do have a viable product. They have a robotaxi ride-hailing service in Phoenix, AZ that anyone in the service area can use anytime. The service is public now. Anyone in the Phoenix service area can use the app anytime and get a ride.

 

[mikes_fsd]

The whole industry of thousands of top engineers have been working on FSD for years and say it is a super hard problem. But I guess they all just have the wrong mindset. Silly them for not realizing what a super easy problem autonomous driving really is.

Probably the same ones starting all the consortiums and screaming for regulations as well. Some probably are responsible for the UN/ECE r79 regulation steaming pile of poo.

Wrong. They do have a viable product. They have a robotaxi ride-hailing service in Phoenix, AZ that anyone in the service area can use anytime. The service is public now. Anyone in the Phoenix service area can use the app anytime and get a ride

Sorry, I know you are in academia where "viable product" can be viable in your imagination.
But even if Waymo was able to write off ALL R&D costs and start Waymo 2.0 from what they have today, but they had no external (Google) money, they would be out of business by the end of THIS year.

 

[diplomat33]

Probably the same ones starting all the consortiums and screaming for regulations as well. Some probably are responsible for the UN/ECE r79 regulation steaming pile of poo.

No. I am talking about ML engineers who are solving vision problems everyday. They say FSD is a hard problem.

 

[mark95476]

This is gold!

Time and time again, you prove you don't know what you are talking about.

 

[diplomat33]

This is gold!

Well, when people say dumb sh$t like "FSD is not a hard problem", "it's just a mindset problem" and "people who say it is hard are just consortiums looking to pass regulations for no reason." I am going to call them out for it.

 

[mikes_fsd]

Well, when people say dumb sh$t like "FSD is not a hard problem", "it's just a mindset problem" and "people who say it is hard are just consortiums looking to pass regulations for no reason." I am going to call it out for what it is.

You are correct. That would be a dumb thing to say, so far I've only found with your post #330 quoting it without actual reference.

Can you point to those ppl please -- with sources/quotes in context?

I'll wait.

 

[mikes_fsd]

For 50 years+ we were told that propulsive landing for rockets is "impossible", "extremely difficult", "non-starter", "not viable"
They ran some calculations on paper (note that no-one actually tried to engineer a solution - just theoretical exercises) and wrote it off.

2 generations pass by without anyone really giving it any thought, then you have some schmuck look at the calculations and say, wtf is going on here..? here is what I think we need to make propulsive landing possible.
And then actually start to engineer (design the solution) and build it into an actual product.

In the words of Elon Musk "Physics is the law, everything else is a recommendation" https://twitter.com/elonmusk/status/1243076222596374528
Nothing in FSD is breaking the laws of physics, so to throw up your hands "it's an extremely difficult problem" at every turn just seems dumb.

 

[diplomat33]

You are correct. That would be a dumb thing to say, so far I've only found with your post #330 quoting it without actual reference.

Can you point to those ppl please -- with sources/quotes in context?

I'll wait.

You said it.

You gave Daniel SD two "disagrees" when he said that FSD was a hard problem.

You gave this a "disagree":

Why? I didn't say it was a hard problem for me to solve. I was only observing that it has proven to be a hard problem for all the engineers that have been working on it for a decade.

You gave this a "disagree":

Of course. That's why about $100 billion has been invested in autonomous driving tech. Unfortunately it turns out to be an extremely difficult problem to solve.

You said it was a just a mindset problem and people who think it is hard are just too lazy to solve the problem and they are consortiums looking to pass regulations here:

The only folks who "prove" that a problem -- within the boundaries of physics -- is "extremely difficult" are the ones that are too lazy to take on the problem to solve the problem.

They would rather create consortiums and regulatory bodies or some other crutch than work on the solution.
Oh, they use the "think of the children" or "who should AI chose to kill" click bait arguments to pave the way as well, but it is never to actually solve the problem.

It is the mindset that is the problem.
Hard to overcome when you've convinced yourself that task at hand is "extremely difficult"

Probably the same ones starting all the consortiums and screaming for regulations as well. Some probably are responsible for the UN/ECE r79 regulation steaming pile of poo.

 

[mikes_fsd]

lol, so me disagreeing on his blanket statements is somehow twisted into me making statements that I did not make? Makes total sense.

Not in a single one of those do I say that "FSD is not a hard problem"
And with the specific follow up that this sort of mindset does not help in solving the problem of FSD
"It is the mindset that is the problem.
Hard to overcome when you've convinced yourself that task at hand is "extremely difficult""

 

[diplomat33]

For 50 years+ we were told that propulsive landing for rockets is "impossible", "extremely difficult", "non-starter", "not viable"
They ran some calculations on paper (note that no-one actually tried to engineer a solution - just theoretical exercises) and wrote it off.

2 generations pass by without anyone really giving it any thought, then you have some schmuck look at the calculations and say, wtf is going on here..? here is what I think we need to make propulsive landing possible.
And then actually start to engineer (design the solution) and build it into an actual product.

In the words of Elon Musk "Physics is the law, everything else is a recommendation" https://twitter.com/elonmusk/status/1243076222596374528
Nothing in FSD is breaking the laws of physics, so to throw up your hands "it's an extremely difficult problem" at every turn just seems dumb.

People said propulsive landing for rockets was extremely difficult in the beginning because the technology was new and we had not solved it yet. Yes, we understood the laws of physics but designing the solution was difficult. A lot of mini problems had to be solved. A lot of technology had to be invented. A lot of engineering had to happen. Once we developed the technology and gained a lot of experience by doing it a lot, propulsive landing for rockets became easier. But it was not always easy.

And don't be fooled. SpaceX makes it look easy but propulsive landing for rockets is an extremely complex engineering feat. A lot can go wrong.

Nobody is saying that FSD breaks the laws of physics. That is not why it is hard. The reason FSD is hard is because "designing the solution" is hard. There are lot of mini problems to solve. There are engineering challenges that need to be solved. There are technologies that need to be developed. For example, the NN needed to do camera vision are extremely complex. Somebody needed to figure out how to develop these NN from scratch. It was not simple or obvious. Things like pseudo-lidar needed for camera vision to see things in 3D is not simple. It requires a lot of complex equations and algorithms. Somebody needed to develop those algorithms and equations. Somebody needed to figure out how to translate 2D images into a 3D map. Somebody needed to figure out how to build a NN that can predict paths or predict intent of a pedestrian based on hand gestures. etc... There is are a ton of problems that need to be solved. There are computer and engineering problems related to FSD that we have not solved yet. That's why it is hard.

It has taken AV companies years just to get to where we are now. So it is clearly not an easy problem. But nobody is throwing up their hands just because it is hard. That's why AV companies like Waymo, Cruise, Zoox, Mobileye, Tesla and so many others are working so hard on FSD.

 

[Daniel in SD]

For 50 years+ we were told that propulsive landing for rockets is "impossible", "extremely difficult", "non-starter", "not viable"
They ran some calculations on paper (note that no-one actually tried to engineer a solution - just theoretical exercises) and wrote it off.

2 generations pass by without anyone really giving it any thought, then you have some schmuck look at the calculations and say, wtf is going on here..? here is what I think we need to make propulsive landing possible.
And then actually start to engineer (design the solution) and build it into an actual product.

In the words of Elon Musk "Physics is the law, everything else is a recommendation" https://twitter.com/elonmusk/status/1243076222596374528
Nothing in FSD is breaking the laws of physics, so to throw up your hands "it's an extremely difficult problem" at every turn just seems dumb.

We landed on the moon with propulsive landing.
Also:

 

[mikes_fsd]

We landed on the moon with propulsive landing.

lol, yet, in order to leave the moon we had to leave that lander on the moon.
I know you are trying hard, but try harder.

Propulsive landing for the purpose of reuse of the vehicle and thus lowering the cost to access space and/or other heavenly bodies.


By that same train of thought... FSD was "solved" in 1989 with the military prototype ALVINN (Autonomous Land Vehicle In a Neural Network) Meet ALVINN, the Self-Driving Car From 1989

There used to be YouTube videos of this as a demo, but I cannot find one that is not taken down...

 

[diplomat33]

I know people see me as a "waymo fanboy" and I am. Guilty as charged! But I do wish Waymo would ditch the Chrysler Pacifica. I don't really care for it. It's too conventional, it's ugly, it's gas and it has the sensors tacked on. The Jaguar I-Pace looks better IMO. And I know Waymo says they are committed to 100% EV fleet but I wish they would speed that up. I'd also love it if Waymo had a custom built driverless robotaxi like Cruise and Zoox have done. I know Waymo had the Firefly. I wish they would commit to like a new, bigger version of the Firefly or something like it. I would just love a "real robotaxi" instead the ugly Chrysler Pacifica. And I also wish Waymo would expand faster. They clearly got FSD that is good enough to be driverless in some conditions like good weather. So start deploying them in more cities. Even a limited deployment in other cities would go a long way IMO to show that their business model can work. Don't stay in Chandler until the FSD is 100% perfect. So there are some things about Waymo that I am critical of.

 

[Doggydogworld]

I know people see me as a "waymo fanboy" and I am. Guilty as charged! But I do wish Waymo would ditch the Chrysler Pacifica. I don't really care for it. It's too conventional, it's ugly, it's gas and it has the sensors tacked on. The Jaguar I-Pace looks better IMO. And I know Waymo says they are committed to 100% EV fleet but I wish they would speed that up. I'd also love it if Waymo had a custom built driverless robotaxi like Cruise and Zoox have done. I know Waymo had the Firefly. I wish they would commit to like a new, bigger version of the Firefly or something like it. I would just love a "real robotaxi" instead the ugly Chrysler Pacifica. And I also wish Waymo would expand faster. They clearly got FSD that is good enough to be driverless in some conditions like good weather. So start deploying them in more cities. Even a limited deployment in other cities would go a long way IMO to show that their business model can work. Don't stay in Chandler until the FSD is 100% perfect. So there are some things about Waymo that I am critical of.

Pacifica is a much better platform for early deployment than i-Pace (or Model 3/Y). Easier ingress/egress, enough room for a permanent child seat, etc. A custom design makes sense after 100k units, and will incorporate lessons they learn from early rollouts. Zoox was dumb to spend early resource on their pod instead of focusing 100% on the real problem.

I'm very critical of Waymo's extreme timidity. But they're the one risking billions, and their intense focus on safety may win out in the end.

 

[diplomat33]

Pacifica is a much better platform for early deployment than i-Pace (or Model 3/Y). Easier ingress/egress, enough room for a permanent child seat, etc. A custom design makes sense after 100k units, and will incorporate lessons they learn from early rollouts. Zoox was dumb to spend early resource on their pod instead of focusing 100% on the real problem.

I'm very critical of Waymo's extreme timidity. But they're the one risking billions, and their intense focus on safety may win out in the end.

Good points. And yes, I get that that the Pacifica as a minivan is probably a great platform. Lots of space for passengers. It's just not my kind of vehicle.

And yes, Zoox may run into trouble if their autonomous driving runs into problems when they've already invested in a bunch of robotaxis.

Yes, the safety first approach is probably very smart. It's probably better to go slow but steady and make it to the end than to go too fast and get shut down because of a serious accident. But obviously, it is a bit frustrating for fans like me who are impatient to ride in a driverless car.