7 ms·

No carmaker has a city-driving autonomous vehicle that doesn't require super-precise, cm-scale, unrealistically always-up-to-date maps for the AI. The autonomo

No carmaker has a city-driving autonomous vehicle that doesn't require super-precise, cm-scale, unrealistically always-up-to-date maps for the AI. The autonomous super-size dump trucks drive along carefully pre-programmed routes using GPS. We're still a long ways away from practical and safe self-driving cars. AI and computer vision just isn't there yet.

swombat 11y ago
"A long way away" - only true in terms of linear extrapolations of current progress. Think in terms of exponential progress, and we're about 5 years from the majority of trips being done by self-driving car, IMHO. The benefits of self-driving cars are so powerful that they will take over the world even faster than the electric vehicles described in this article...
CountSessine 11y ago
Think in terms of exponential progress, and we're about 5 years from the majority of trips being done by self-driving car, IMHO 5 years is a fantasy. Continued exponential increases in device computing power is by no means guaranteed now that chips are at the 14nm feature size. And the recent progress in AI can only be partially attributed to improved computing power - much of it was from improved approaches to training neural networks. We could stumble into another AI winter and have to wait another 30 years for the next round of big algorithmic AI breakthroughs. I guarantee you that self-driving cars will not be the mainstream even in 15 years. Probably more.
- toomuchtodo 11y ago
  > I guarantee you that self-driving cars will not be the mainstream even in 15 years. Probably more. Want to make a Long Bet on that? I'm willing to bet up to $1000 USD, loser pays winner's choice of 501(c)(3) US charity org, that self-driving cars will be mainstream in under 10 years.
  - ghaff 11y ago
    What do you mean by self-driving cars? As in highly assistive driving on highways so that you can not really pay attention most of the time whether that's legal or not? That I could (maybe) buy though it's a stretch given replacement cycles. But something I can summon with an app and take a nap? 50 years may be too soon.
    - toomuchtodo 11y ago
      "Level 4: The vehicle performs all safety-critical functions for the entire trip, with the driver not expected to control the vehicle at any time. As this vehicle would control all functions from start to stop, including all parking functions, it could include unoccupied cars." https://en.wikipedia.org/wiki/Autonomous_car#Definition https://en.wikipedia.org/wiki/Autonomous_car#Definition
      - ghaff 11y ago
        That's many decades away.
        toomuchtodo 11y ago
        My bet calls for under 10 years.
      - hvidgaard 11y ago
        Now you just need to define "mainstream".
  - CountSessine 11y ago
    How gauche! No, I don't want to make a Long Bet on that. I don't live in the United States where every CS graduate gets a Ferrari as a signing bonus (maybe a self-driving Ferrari one day). $USD1000 is clearly a lot more money to me than it is to you. But if you'd like to take your own side of the bet and choose your own charity, please feel free. Moreover, I'd like to know what the terms of that bet that you'd be making with yourself. Is it for a generally available car that can drive reasonably anywhere? In San Fransisco but also in small towns without super-accurate cm-scale mapping data (which will never be both everywhere and up-to-date)? Do we consider it a success if self-driving cars blindly drive into construction sites that haven't been mapped? Or crash into newly erected barricades? Because I don't, and that's exactly where we'll be in 15 years. Autonomous vehicles are quickly approaching what I will acknowledge is probably an 80% solution. In nice weather, when their mapping is perfect in a perfectly-mapped area, and they have a good GPS signal, autonomous cars will work well (and currently are working well in testing). That's not the standard that's expected. No one lives in a city without road work or construction going on. No one lives in a city where temporary or make-shift signs aren't being erected, or where there are never detours, or where there isn't ad-hoc communication going on between pedestrians and drivers, or where weather doesn't block LIDAR, or, ...whatever. We will be nowhere near level 4 autonomous vehicles in 10 years time. And not in 15 years, either. No one in this field is confident of that anymore. Not even Google is making the same bold claims that they were 3 years ago. Their own researchers are publicly hedging now (http://gizmodo.com/how-to-teach-an-autonomous-car-to-drive-1694725874 http://gizmodo.com/how-to-teach-an-autonomous-car-to-drive-1...) and backing away from their most ambitious claims. The only other group that I'm aware of that's still claiming that we're anywhere near having generally-useful self-driving cars are the established car makers, none of whom have traditionally been serious AI and computer vision research contributors.
- Tossrock 11y ago
  I don't think there will be any more AI winters. Previous AI winters were largely the result of a crash in federal grant writing (DARPA/the Strategic Computing Initiative) to labs at research universities, which depend on them for cash. This time, the ongoing AI/ML renaissance is funded by private industry, and the value they are extracting from it far outstrips the costs of keeping their in-house R&D labs operating. Even if there's a contraction in the market, I can't envision any possible world where Google/Facebook/etc stop working on AI, and they have the money to keep it up as long as they want.
mavhc 11y ago
I assume each car will record its data and upload it. Handily google fiber has lots of upload capacity. Google will then aggregate changes in the maps.

swombat 11y ago

"A long way away" - only true in terms of linear extrapolations of current progress. Think in terms of exponential progress, and we're about 5 years from the majority of trips being done by self-driving car, IMHO. The benefits of self-driving cars are so powerful that they will take over the world even faster than the electric vehicles described in this article...

CountSessine 11y ago

Think in terms of exponential progress, and we're about 5 years from the majority of trips being done by self-driving car, IMHO 5 years is a fantasy. Continued exponential increases in device computing power is by no means guaranteed now that chips are at the 14nm feature size. And the recent progress in AI can only be partially attributed to improved computing power - much of it was from improved approaches to training neural networks. We could stumble into another AI winter and have to wait another 30 years for the next round of big algorithmic AI breakthroughs. I guarantee you that self-driving cars will not be the mainstream even in 15 years. Probably more.

toomuchtodo 11y ago
> I guarantee you that self-driving cars will not be the mainstream even in 15 years. Probably more. Want to make a Long Bet on that? I'm willing to bet up to $1000 USD, loser pays winner's choice of 501(c)(3) US charity org, that self-driving cars will be mainstream in under 10 years.
- ghaff 11y ago
  What do you mean by self-driving cars? As in highly assistive driving on highways so that you can not really pay attention most of the time whether that's legal or not? That I could (maybe) buy though it's a stretch given replacement cycles. But something I can summon with an app and take a nap? 50 years may be too soon.
  - toomuchtodo 11y ago
    "Level 4: The vehicle performs all safety-critical functions for the entire trip, with the driver not expected to control the vehicle at any time. As this vehicle would control all functions from start to stop, including all parking functions, it could include unoccupied cars." https://en.wikipedia.org/wiki/Autonomous_car#Definition https://en.wikipedia.org/wiki/Autonomous_car#Definition
    - ghaff 11y ago
      That's many decades away.
      - toomuchtodo 11y ago
        My bet calls for under 10 years.
    - hvidgaard 11y ago
      Now you just need to define "mainstream".
- CountSessine 11y ago
  How gauche! No, I don't want to make a Long Bet on that. I don't live in the United States where every CS graduate gets a Ferrari as a signing bonus (maybe a self-driving Ferrari one day). $USD1000 is clearly a lot more money to me than it is to you. But if you'd like to take your own side of the bet and choose your own charity, please feel free. Moreover, I'd like to know what the terms of that bet that you'd be making with yourself. Is it for a generally available car that can drive reasonably anywhere? In San Fransisco but also in small towns without super-accurate cm-scale mapping data (which will never be both everywhere and up-to-date)? Do we consider it a success if self-driving cars blindly drive into construction sites that haven't been mapped? Or crash into newly erected barricades? Because I don't, and that's exactly where we'll be in 15 years. Autonomous vehicles are quickly approaching what I will acknowledge is probably an 80% solution. In nice weather, when their mapping is perfect in a perfectly-mapped area, and they have a good GPS signal, autonomous cars will work well (and currently are working well in testing). That's not the standard that's expected. No one lives in a city without road work or construction going on. No one lives in a city where temporary or make-shift signs aren't being erected, or where there are never detours, or where there isn't ad-hoc communication going on between pedestrians and drivers, or where weather doesn't block LIDAR, or, ...whatever. We will be nowhere near level 4 autonomous vehicles in 10 years time. And not in 15 years, either. No one in this field is confident of that anymore. Not even Google is making the same bold claims that they were 3 years ago. Their own researchers are publicly hedging now (http://gizmodo.com/how-to-teach-an-autonomous-car-to-drive-1694725874 http://gizmodo.com/how-to-teach-an-autonomous-car-to-drive-1...) and backing away from their most ambitious claims. The only other group that I'm aware of that's still claiming that we're anywhere near having generally-useful self-driving cars are the established car makers, none of whom have traditionally been serious AI and computer vision research contributors.
Tossrock 11y ago
I don't think there will be any more AI winters. Previous AI winters were largely the result of a crash in federal grant writing (DARPA/the Strategic Computing Initiative) to labs at research universities, which depend on them for cash. This time, the ongoing AI/ML renaissance is funded by private industry, and the value they are extracting from it far outstrips the costs of keeping their in-house R&D labs operating. Even if there's a contraction in the market, I can't envision any possible world where Google/Facebook/etc stop working on AI, and they have the money to keep it up as long as they want.

mavhc 11y ago

I assume each car will record its data and upload it. Handily google fiber has lots of upload capacity. Google will then aggregate changes in the maps.