5 ms·

...so in what way do they have the lead?

by scotth 6y ago

fullshark 6y ago
It's not a certainty
- fullshark 6y ago
  I think I misunderstood your comment initially and you were merely curious what the argument is. The argument has been made elsewhere in the thread but basically they have a real product with real usage / data creating a positive data feedback loop that will lead to incremental improvements. The question is whether it will improve along an asymptote far from full self driving (what most people believe) or some leap occurs in the future or some technology can be added to their capabilities to achieve full self driving at some point. The latter is possible imo even if highly unlikely. The fact remains they are out there, and you can buy them, and they can do some sort of self-driving. Waymo is an impressive tech demo at this point.
jrockway 6y ago
I guess they have the lead in the sense that you can drive to a Tesla dealership today and buy a car that will kind of sort of drive you home. That's something!
- birdyrooster 6y ago
  Just don't take 101 S bound near the 85 S on ramp on your way home and you might live to do it again
- olyjohn 6y ago
  Big deal. You can do this in a Honda Fit for crying out loud.
  - colejohnson66 6y ago
    You can, yes, but marketing is very much a thing. The average person knows Tesla has self driving (to whatever extent), but not everyone knows a Honda Fit can drive itself. I sure didn’t.
    - rtx 6y ago
      Isnt it amazing we need billions of dollars of investment in AI for that. If they could have continued with knobs instead of touch screen.
  - fullshark 6y ago
    They are in 2nd.
  - QuercusMax 6y ago
    Nissan Leaf, too. It's not great, but is good enough that I'm not scared to look away for a moment to change stations.
lozaning 6y ago
They certainly have the largest corpus of data to train their models on. All their cars reporting back all that data is an extremely valuable moat.
- karpierz 6y ago
  Is unlabeled video data in limited supply? My impression was that the limiting factor was labeled data, especially labeled video + lidar data.
- agildehaus 6y ago
  Yet their cars still crash into huge stationary objects and pedestrians. https://www.youtube.com/watch?v=FVgkWii5JdM https://www.youtube.com/watch?v=FVgkWii5JdM
- nmstoker 6y ago
  Just a small nitpick point: data is collected in the order of a GB per minute so there's no way they're reporting back even a fraction of the data gathered.
jfoster 6y ago
I would argue that the most significant way in which Tesla have the lead is simple but a bit unintuitive. It's that their customers pay for the cars. Think about how much capital will be required to put a million Waymo vehicles on the road. How many zeros are there on that number? Who is going to be paying that bill? Given this, even if Waymo are first to driving 100% perfectly, I expect that Tesla will eventually leapfrog them.
- ra7 6y ago
  > Think about how much capital will be required to put a million Waymo vehicles on the road. Waymo can just license the tech and sensor suite to manufacturers and make money off that. They don't need to actually build and sell a million Waymo vehicles.
  - fullshark 6y ago
    Yeah this is ultimately the case Waymo are in the lead, if you imagine the only way this happens is if eventually one organization gets the hardware/software/dataset/patents that makes full self-driving possible and they will license it to everyone else, then Waymo is the most likely candidate.
- sidibe 6y ago
  Waymo is part of a company that is sitting on $100B+, if they thought rolling out their own cars everywhere was worth it they could afford it easier than Tesla, but also they don't need to do it in a capital intensive way, they can partner with car manufacturers or ride hailing businesses and just be responsible for the driver.
  - jfoster 6y ago
    They have $100B+, but do they want to spend it all in one go like that? Sure, there's also ways to split or reduce the cost, but those also require things to be done. How long do you think it takes Waymo to get their stuff ready for a new vehicle model? Last I heard, it was between 6 and 12 months. The beauty of the Tesla model is that they've already made tonnes of the cars that should eventually be able to do this. Furthermore, instead of each vehicles being a cost for them, the vehicles are a source of revenue. I acknowledge that it does all hinge upon Tesla reaching a point with their autonomy where it can be unsupervised. Some people say they will never get there, but I think that's pretty shortsighted. The data collected by their existing fleet is a massive advantage. Let's not forget that it's been more than 11 years since the first video of what became Waymo, and they can pretty much only drive a small area in Phoenix, with high likelihood of needing to train extensively for every new location as they're now doing in California. Tesla are doing most of the world in (more or less) one go.
imtringued 6y ago
You can buy it and get your hands on it today.

...so in what way do they have the lead?

by scotth 6y ago

...so in what way do they have the lead?

fullshark 6y ago

It's not a certainty

fullshark 6y ago
I think I misunderstood your comment initially and you were merely curious what the argument is. The argument has been made elsewhere in the thread but basically they have a real product with real usage / data creating a positive data feedback loop that will lead to incremental improvements. The question is whether it will improve along an asymptote far from full self driving (what most people believe) or some leap occurs in the future or some technology can be added to their capabilities to achieve full self driving at some point. The latter is possible imo even if highly unlikely. The fact remains they are out there, and you can buy them, and they can do some sort of self-driving. Waymo is an impressive tech demo at this point.

jrockway 6y ago

I guess they have the lead in the sense that you can drive to a Tesla dealership today and buy a car that will kind of sort of drive you home. That's something!

birdyrooster 6y ago
Just don't take 101 S bound near the 85 S on ramp on your way home and you might live to do it again
olyjohn 6y ago
Big deal. You can do this in a Honda Fit for crying out loud.
- colejohnson66 6y ago
  You can, yes, but marketing is very much a thing. The average person knows Tesla has self driving (to whatever extent), but not everyone knows a Honda Fit can drive itself. I sure didn’t.
  - rtx 6y ago
    Isnt it amazing we need billions of dollars of investment in AI for that. If they could have continued with knobs instead of touch screen.
- fullshark 6y ago
  They are in 2nd.
- QuercusMax 6y ago
  Nissan Leaf, too. It's not great, but is good enough that I'm not scared to look away for a moment to change stations.

lozaning 6y ago

They certainly have the largest corpus of data to train their models on. All their cars reporting back all that data is an extremely valuable moat.

karpierz 6y ago
Is unlabeled video data in limited supply? My impression was that the limiting factor was labeled data, especially labeled video + lidar data.
agildehaus 6y ago
Yet their cars still crash into huge stationary objects and pedestrians. https://www.youtube.com/watch?v=FVgkWii5JdM https://www.youtube.com/watch?v=FVgkWii5JdM
nmstoker 6y ago
Just a small nitpick point: data is collected in the order of a GB per minute so there's no way they're reporting back even a fraction of the data gathered.

jfoster 6y ago

I would argue that the most significant way in which Tesla have the lead is simple but a bit unintuitive. It's that their customers pay for the cars. Think about how much capital will be required to put a million Waymo vehicles on the road. How many zeros are there on that number? Who is going to be paying that bill? Given this, even if Waymo are first to driving 100% perfectly, I expect that Tesla will eventually leapfrog them.

ra7 6y ago
> Think about how much capital will be required to put a million Waymo vehicles on the road. Waymo can just license the tech and sensor suite to manufacturers and make money off that. They don't need to actually build and sell a million Waymo vehicles.
- fullshark 6y ago
  Yeah this is ultimately the case Waymo are in the lead, if you imagine the only way this happens is if eventually one organization gets the hardware/software/dataset/patents that makes full self-driving possible and they will license it to everyone else, then Waymo is the most likely candidate.
sidibe 6y ago
Waymo is part of a company that is sitting on $100B+, if they thought rolling out their own cars everywhere was worth it they could afford it easier than Tesla, but also they don't need to do it in a capital intensive way, they can partner with car manufacturers or ride hailing businesses and just be responsible for the driver.
- jfoster 6y ago
  They have $100B+, but do they want to spend it all in one go like that? Sure, there's also ways to split or reduce the cost, but those also require things to be done. How long do you think it takes Waymo to get their stuff ready for a new vehicle model? Last I heard, it was between 6 and 12 months. The beauty of the Tesla model is that they've already made tonnes of the cars that should eventually be able to do this. Furthermore, instead of each vehicles being a cost for them, the vehicles are a source of revenue. I acknowledge that it does all hinge upon Tesla reaching a point with their autonomy where it can be unsupervised. Some people say they will never get there, but I think that's pretty shortsighted. The data collected by their existing fleet is a massive advantage. Let's not forget that it's been more than 11 years since the first video of what became Waymo, and they can pretty much only drive a small area in Phoenix, with high likelihood of needing to train extensively for every new location as they're now doing in California. Tesla are doing most of the world in (more or less) one go.

imtringued 6y ago

You can buy it and get your hands on it today.