So the most revolutionary thing about self-driving cars is onboard computing? I guess that is the case if you happen to be a blog for a company who specializes in that, but for everyone else the most revolutionary thing is still the huge advances in computer vision and scene understanding via deep learning...
Self-driving cars don't really need to talk to "the cloud" much, if at all. They need some map data, and they might contribute to map data. They can use some traffic data, and can contribute to traffic data, which needs about as much data as Waze. None of this is real-time; it can be seconds or minutes behind.
Transmitting all the car's imagery to servers is an R&D activity. Google downloads all that from their vehicles to train their machine learning systems, but training is an offline process. Production vehicles might occasionally upload "interesting" imagery they didn't recognize, or data from situations where there was trouble, but there's no need to upload it all.
That's a bit different from talking to "the cloud" as the parent poster was talking about. Cars will be using a short range radio technology like DSRC (5.9ghz band) to communicate locally. The FCC and DOT are already pushing mandates to have cars equipped with it.
Can we get this clickbait title changed? I even skimmed the article and I still don't know what "the most revolutionary thing about self-driving cars" is supposed to be.
I've taken a crack at it, but if anyone suggests a better title (i.e. more accurate and neutral, and using language from the article itself), we can change it again.
No, the most revolutionary thing is the freedom from police harassment it will lead to. Self driving cars will obey all local traffic laws to reach certification and deny the police a power the abuse which is the right to use a traffic stop pretextually. As in, stop you for speeding or running a light to regardless if that was his real intent.
See https://www.washingtonpost.com/news/volokh-conspiracy/wp/201... for a better explanation than I can ever make. Think about it, the abuse that can happen to minorities and poorer elements of society will be much more difficult to execute. Even stealing from people under forfeiture laws will be much more difficult if the officer cannot find a valid reason to stop you for his otherwise invalid purpose
"(Cops) can follow self-driving cars all day but won’t be able to pull them over."
Yes, that's going to be interesting. I'm waiting for an event when some small-town sheriff pulls over a Waymo self-driving car without justification, and Waymo shows up in court with lawyers, full imagery, LIDAR scans, GPS info, a complete video reconstruction of the event, and the browsing history of the sheriff.
For now they still have trouble staying in their lane, I don't see any lack of reasons to pull them over, in fact I'd love to see them taken off the roads entirely until such bugs have been ironed out. Public roads are not the place to beta test software.
Yes, but 'where there are clear lane lines' is exactly the problem. I know plenty of places where there are no clear lane lines. If every one of those is an accident waiting to happen then we first should invest in some kind of minimal level road markings. And even then you'll have plenty of exceptions.
> Self driving cars will obey all local traffic laws
I doubt it. I don't think consumers would tolerate a car that always obeyed the speed limit. And if a self-driving car did, it could be pulled over for impeding traffic.
Is there an actual source for that number at all? Intel did a
marketing blog post [0] where they said 4TB per 1.5HR, but without any actual explanation of it, and since then a bunch of posts started using "per 8 hours".
I'm cautious of this number because the camera and sensor feeds are very often reduced heavily in resolution and then passed for processing, so if they are considering raw uncompressed sensor streams then that's not really representative of the theoretical stored size of the data or the size of the data actually processed.
In practice, far less data is saved. Tesla's production "Autopilot" saves and uploads 15 frames when there's an "interesting event", according to the NTSB report of a Tesla crash. (There's no imagery of the event where the vehicle plowed under a semitrailer, because Tesla's autopilot didn't detect anything that required braking before or during the crash.)
I think this post should be renamed to something like "The Self-Driving Car's Impact On Edge Computing." I thought that the most revolutionary thing about self-driving cars would be the economic impact and resulting cultural changes, which weren't really addressed in the article...
Previously all you had to worry about with your car was someone cutting your breaks or stuffing some prawns in your A/C, but with "driverless" cars you need to suddenly be concerned about Russian teenagers making you drive off a cliff.