5 ms·

Elon Musk Says Autopilot Self-Driving Beta Will Start Next Tuesday

17 points by nathandaly 6y ago

mensetmanusman 6y ago
So hopeful that their ML algorithm converges with the data they have now. 2020 would be almost redeemed if driverless cars are released for real...
- Waterluvian 6y ago
  We are just nowhere close. If we were anywhere we would be seeing iterations of it everywhere. Self driving will be one of those things that’s so gradual that you barely notice it arriving because there is no concrete “it’s here!” moment. It doesn’t just get unveiled one day. For example, I bought a base line suv this year that drives itself on highways most of the time. I think it’s the coolest thing ever. And a better version will be around every year until one day I wake up and think, “Oh wow self driving snuck up on me!” I’ve had that very moment about my smartphone recently. My Star Trek data pad snuck up on me.
  - ChrisClark 6y ago
    I like that observation. We're focused on the end result, and don't see steps we're taking. For the car example, yeah. At first my car just stayed in the lane, then an update and it could change lanes, next it could choose to pass slower cars for me, then it could go highway to highway taking offramps by itself. And now it stops at red lights and stop signs. It's not everything or nothing, by the time full self driving is here we'll have already been used to 90% of it and it won't feel like such a big step.
  - fiddlerwoaroof 6y ago
    My Model 3 is very good, even on badly marked roads: I know it’s sort of fashionable to hate on Tesla’s Autopilot system, but—whatever limitations it does have—I miss it when I’m driving a car without it. And, for this sort of thing, Tesla’s OTA updates are a major competitive advantage.
  - cltsang 6y ago
    I've never been a Tesla owner, so I'm not speaking of experience. But from what I see on various YouTube reviews, in urban areas, it's pretty close. Are there anything specific that you feel are nowhere close?
    - Rebelgecko 6y ago
      Without doing any research into the state of the art, I'd be curious about how well it does in rain, snow, and parking lots
    - objclxt 6y ago
      It depends what you mean by “urban area” - somewhere like Palo Alto, where you’ve got nice big roads and a few turns before getting on the freeway and lots of sun? It’s not too bad, I’ve driven with colleagues there in their Teslas. My current school run in London (small streets, changeable weather, congestion, roadworks a plenty) - absolutely awful. I imagine it’s similar for New York. I have been driving my Model 3 in London a lot more recently due to the pandemic, and autopilot is nowhere near ready. Almost worse than useless.
  - ilaksh 6y ago
    What your car does and what technology is available are two totally different things. Waymo is exclusively driverless right now. So that's where we are occupying parallel universes. I see videos on YouTube of Waymos with no drivers delivering people that were recorded by riders and came out in the last few days. So in my universe, it already exists. In your universe, it's not close. Tesla does have amazing abilities and they are close. They will have their limited beta. And it may take a year or two in the end but eventually they will roll it out.
    - sumedh 6y ago
      > Waymo is exclusively driverless right now. So that's where we are occupying parallel universes. Can Waymo drive everywhere? No it cannot so the question of parallel universes is not even relevant.
jameslk 6y ago
> The CEO also noted that the system would be able to get commuters from home to work most of the time with no interventions. I'm so glad it will be that easy. I guess we'll know if there's still some bugs if it still drives into highway barriers or other stationary objects at full speed? But somehow it'll still probably be the users fault
- pfundstein 6y ago
  By making Autopilot available but requiring that you're paying attention and ready to take over, Tesla is getting you to perform the testing of the Autopilot feature for free while the vehicle collects data for improving the system. Very clever business model.
  - m463 6y ago
    This has been the case since day 1. Thing is, over time the tricky bits have gotten better (or even possible) An example of a tricky bit -- crossing a multilane intersection and getting back in your lane when the dotted lines disappear and start up again.
aphroz 6y ago
I am waiting for the comments saying "We are far from level 5 autonomous driving, we will not see it before at least 20 years". And Elon Musk keep doing "impossible" things. It really makes you think you should not listen to people who tells you it's impossible :) (I decline all responsibility)
- m463 6y ago
  Every single technology has the same pattern (overestimated in the short term, underestimated in the long term) for example, I remember when speech recognition would "take over". Well, it didn't and people forgot about saying that. And then gradually it started doing "yes" and "no" during phone calls, then doing phone trees, and then "siri/alexa/etc is so awful" and now people use it all the time.
- objclxt 6y ago
  We are. This isn’t Level 5, it’s nowhere close to level 5, and calling it “full self driving” is very disingenuous. I say this as someone who very much likes and owns a Tesla.
mola 6y ago
I don't understand how can you call something "Full self driving" car, and say that it always need a driver, because it can be autonomous most of the time. That not most of the time events are exactly what makes it not FSD. And all the promises FSD could bring to the table that don't amount to less driver leg work. Autonomous cars are a revolution, could actually make cars efficient, and introduce a different ownership model. Tesla's FSD is just a very capable driver assist, which can't deliver on the promises of real autonomous cars.
- birdyrooster 6y ago
  Elon duped everyone by selling hardware to enable FSD before such a thing was even proved possible. Those people were scammed, fair and square.
  - jfim 6y ago
    The thing is, the hardware is technically sufficient for full self driving, at least for some definition of "full self driving". Even in the late 90s, there was experimentation with self driving cars that were running a single Pentium with a handful of megabytes of RAM and inexpensive black and white cameras [0]. Now whether the software for that hardware is ready (or will be any time soon) is a different question. [0] https://www.semanticscholar.org/paper/ARGO-and-the-MilleMiglia-in-Automatico-Tour-Broggi-Bertozzi/8193f359528cc94e2bd3d883227f78f27b067325 https://www.semanticscholar.org/paper/ARGO-and-the-MilleMigl...

17 points by nathandaly 6y ago

So hopeful that their ML algorithm converges with the data they have now. 2020 would be almost redeemed if driverless cars are released for real...

Waterluvian 6y ago
We are just nowhere close. If we were anywhere we would be seeing iterations of it everywhere. Self driving will be one of those things that’s so gradual that you barely notice it arriving because there is no concrete “it’s here!” moment. It doesn’t just get unveiled one day. For example, I bought a base line suv this year that drives itself on highways most of the time. I think it’s the coolest thing ever. And a better version will be around every year until one day I wake up and think, “Oh wow self driving snuck up on me!” I’ve had that very moment about my smartphone recently. My Star Trek data pad snuck up on me.
- ChrisClark 6y ago
  I like that observation. We're focused on the end result, and don't see steps we're taking. For the car example, yeah. At first my car just stayed in the lane, then an update and it could change lanes, next it could choose to pass slower cars for me, then it could go highway to highway taking offramps by itself. And now it stops at red lights and stop signs. It's not everything or nothing, by the time full self driving is here we'll have already been used to 90% of it and it won't feel like such a big step.
- fiddlerwoaroof 6y ago
  My Model 3 is very good, even on badly marked roads: I know it’s sort of fashionable to hate on Tesla’s Autopilot system, but—whatever limitations it does have—I miss it when I’m driving a car without it. And, for this sort of thing, Tesla’s OTA updates are a major competitive advantage.
- cltsang 6y ago
  I've never been a Tesla owner, so I'm not speaking of experience. But from what I see on various YouTube reviews, in urban areas, it's pretty close. Are there anything specific that you feel are nowhere close?
  - Rebelgecko 6y ago
    Without doing any research into the state of the art, I'd be curious about how well it does in rain, snow, and parking lots
  - objclxt 6y ago
    It depends what you mean by “urban area” - somewhere like Palo Alto, where you’ve got nice big roads and a few turns before getting on the freeway and lots of sun? It’s not too bad, I’ve driven with colleagues there in their Teslas. My current school run in London (small streets, changeable weather, congestion, roadworks a plenty) - absolutely awful. I imagine it’s similar for New York. I have been driving my Model 3 in London a lot more recently due to the pandemic, and autopilot is nowhere near ready. Almost worse than useless.
- ilaksh 6y ago
  What your car does and what technology is available are two totally different things. Waymo is exclusively driverless right now. So that's where we are occupying parallel universes. I see videos on YouTube of Waymos with no drivers delivering people that were recorded by riders and came out in the last few days. So in my universe, it already exists. In your universe, it's not close. Tesla does have amazing abilities and they are close. They will have their limited beta. And it may take a year or two in the end but eventually they will roll it out.
  - sumedh 6y ago
    > Waymo is exclusively driverless right now. So that's where we are occupying parallel universes. Can Waymo drive everywhere? No it cannot so the question of parallel universes is not even relevant.

jameslk 6y ago

> The CEO also noted that the system would be able to get commuters from home to work most of the time with no interventions. I'm so glad it will be that easy. I guess we'll know if there's still some bugs if it still drives into highway barriers or other stationary objects at full speed? But somehow it'll still probably be the users fault

pfundstein 6y ago
By making Autopilot available but requiring that you're paying attention and ready to take over, Tesla is getting you to perform the testing of the Autopilot feature for free while the vehicle collects data for improving the system. Very clever business model.
- m463 6y ago
  This has been the case since day 1. Thing is, over time the tricky bits have gotten better (or even possible) An example of a tricky bit -- crossing a multilane intersection and getting back in your lane when the dotted lines disappear and start up again.

aphroz 6y ago

I am waiting for the comments saying "We are far from level 5 autonomous driving, we will not see it before at least 20 years". And Elon Musk keep doing "impossible" things. It really makes you think you should not listen to people who tells you it's impossible :) (I decline all responsibility)

m463 6y ago
Every single technology has the same pattern (overestimated in the short term, underestimated in the long term) for example, I remember when speech recognition would "take over". Well, it didn't and people forgot about saying that. And then gradually it started doing "yes" and "no" during phone calls, then doing phone trees, and then "siri/alexa/etc is so awful" and now people use it all the time.
objclxt 6y ago
We are. This isn’t Level 5, it’s nowhere close to level 5, and calling it “full self driving” is very disingenuous. I say this as someone who very much likes and owns a Tesla.

mola 6y ago

I don't understand how can you call something "Full self driving" car, and say that it always need a driver, because it can be autonomous most of the time. That not most of the time events are exactly what makes it not FSD. And all the promises FSD could bring to the table that don't amount to less driver leg work. Autonomous cars are a revolution, could actually make cars efficient, and introduce a different ownership model. Tesla's FSD is just a very capable driver assist, which can't deliver on the promises of real autonomous cars.

birdyrooster 6y ago
Elon duped everyone by selling hardware to enable FSD before such a thing was even proved possible. Those people were scammed, fair and square.
- jfim 6y ago
  The thing is, the hardware is technically sufficient for full self driving, at least for some definition of "full self driving". Even in the late 90s, there was experimentation with self driving cars that were running a single Pentium with a handful of megabytes of RAM and inexpensive black and white cameras [0]. Now whether the software for that hardware is ready (or will be any time soon) is a different question. [0] https://www.semanticscholar.org/paper/ARGO-and-the-MilleMiglia-in-Automatico-Tour-Broggi-Bertozzi/8193f359528cc94e2bd3d883227f78f27b067325 https://www.semanticscholar.org/paper/ARGO-and-the-MilleMigl...