Watch GM’s self-driving car navigate the streets of San Francisco
Looks like GM made a pretty wise investment in acquiring Cruise Automation.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Video of fully autonomous Chevy Bolt driving around San Francisco
- Thread starter bro1999
- Start date
Watch GM’s self-driving car navigate the streets of San Francisco
Looks like GM made a pretty wise investment in acquiring Cruise Automation.
Chopr147
Active Member
Impressive. With so many objects,wires.lights etc....even a computer can have difficulty finding a red light. It stopped though.
JeffK
Well-Known Member
I wonder if some of it is for unnaturally long delays for obstacles... I'd like both sped up and normal speed videos. One for simple brevity and one for critique.
Waiting4M3
Active Member
Do they have a slower speed version? The camera seems to be weaving left/right a lot and making me dizzy, even when the car is going straight, I wonder if at slower speed (closer to actual speed), we can see what is the reason for the weaving motion.
thegruf
Active Member
playback speed- open it in youtube you can slow it down to 0.25 speed
... at which point all manner of issues are apparent.
As has been said, apparently good attempt by GM/Cruise but far too many unknowns to assess really.
A full framerate realtime video from inside the vehicle with view of the steering/controls is a bare minimum, plus we have no idea how much detail the vehicle was pre "optimised" for the route shown.
... at which point all manner of issues are apparent.
As has been said, apparently good attempt by GM/Cruise but far too many unknowns to assess really.
A full framerate realtime video from inside the vehicle with view of the steering/controls is a bare minimum, plus we have no idea how much detail the vehicle was pre "optimised" for the route shown.
In November, 2284 autonomous miles with 6 disengagements. Not too bad. I would not have thought they were that far yet.
Their latest demo is quite impressive. GM releases impressive real-world self-driving demo of Cruise’s Bolt EV prototype They need a whole circus tree on the roof for now but reducing that down to a reasonable size is the easier part to solve. The quality of a solution will be the quality of the software as sensor get commoditized.
Tesla report is consistent as well, as was explained several times by several people. You just do not want to accept that Tesla development cycle is different than others.
Nick S
Flash - MX
This is good news for consumers. Finally other manufactures are getting serious about providing AP features. So far the AP demos have been chartware and on their EV products lines. Wonder how many will port this functionality to their ICE fleet. AP is definitely a market disrupting feature. It could accelerate EV adoption.With more manuf. introducing AP (Tesla, GM, Volvo, MB ... ) next few years are going to be very interesting. Will be great to see how Tesla's team maintains their lead when the big boys start chasing them. It is entirely possible one of the big boys will just buy instead of build their AP functionality and try to acquire Tesla. Just speculation based upon history not backed by any concrete facts.
No, because it would be an extraordinary claim that Tesla develops like no development team does (car or other) And for such an extraordinary claim there is not the smallest shred of evidence except that it could be possible in the same way that it could be possible that tomorrow aliens land on our planet.
Again here are the facts. In October, Elon had to postpone a highly anticipated public demo because their team struggled to put together the video. This was explicitly acknowleged by Elon and the 168 disengagements officially reported clearly support that claim. In November they ran 20 miles uninterrupted in basically exactly the same environment. Why the sudden progress?
That's just not consistent with how you expect normal development to go. You claim Tesla's development model is different? Well, let's compare with Tesla's own development on the exact same hardware and software platform. EAP. We are not going to get closer than that. In December the 'release' EAP is not able to do several things that you'd necessarily need for autonomous drive. Ie, not freak out over hills, keeping a stable model of where line markings are, blind spot detection, react to very basic merging information, interpreting signs, ... Over the following 6 weeks they release a few versions that only marginally improve said functionality. May I remind you that progress is WITH the benefit of millions of miles in shadow mode.
Progress is slow and gradual. That is just the nature of software development. And no, deep learning and AI don't change that. Going from 168 disengagements to zero in the span of 4 weeks and then suddenly losing all progress when they release EAP has never been explained except through some handwaving that 'Tesla is different'.
JeffK
Well-Known Member
When data collection is exponential progress is not necessarily slow and gradual.
It is you who wave hands a lot here. As was pointed out more than once, this is similar to the progress that was done with AP1, including that it was also late.
Similar threads
- Article
