I'm both excited and skeptical. There was a similar demo in 2016 and then repeated claims that full self driving ("coast to coast with no interventions") was less than a year away. How is this different? Apparently people at the event were taken on demo rides and it was very impressive. However, Tesla's current system has several notable issues like crashing into stationary vehicles at highway speeds. So what's the truth here? Very exciting if this is the breakthrough we've been waiting for, but I'm skeptical of all claims until released into the wild and shown to be generally true.
The new systems are vastly improved over the ones from the last few years. I have a 4 year old one. They were explicit that you have to keep your hands up and on the wheel, but they still chose that vastly too suggestive name "autopilot". I want to see more explicit public testing, even though I own (an older) one. I want to see them test it with people walking on the side of the road, walking across the road, riding a bike on the side. Someone weaving or suddenly stepping in the road. Those are the things I want to be reassured about. My car is great at following roads with lines on them or just cars, if the lines peter out.
I am not supporting Elon or Tesla, but they did mention that Autopilot in Tesla is akin to how it's used in flights. Pilots are required to be vigilant at all times. It's more of assist than FSD.
Pilots are highly trained at flying and will fully understand what Autopilot in planes can and cannot do. To apply this logic for an application that will be used by mass consumers is a bit reckless.
The video was impressive, but I'd like to see videos where the self-driving failed and the driver had to take over. Those would give me more confidence in the technology since I'd know what the limits are.
Now it seems very staged (probably isn't but still - i'd imagine they'd pick the most impressive video to post out of all their tests)
A CNN journalist experienced the new autopilot on a test drive, and he has a brief description of a situation in which the Tesla representative chooses to hit the brakes:
There's no dismissing how incredible this really is... there's no way I'd have imagined letting a machine control my car at 50mph in the city and highway just a few years ago. A mass-market video processor capable of 2000 frames/sec being piped to a machine learning system capable of >100 trillion instructions/sec is mind blowing tech.
Elon successfully uses the same 'distortion reality' shield that Jobs did. This forces his staff to keep trying to achieve his practically impossible demands.
Success requires that it keep working over and over while not burning out critical staff and investor's money.
It makes sense to be a lot more strict about it in locales where vehicles are allowed to operate at widely divergent speeds on the same set of lanes. In the US many major roads have both a maximum and minimum speed limit that's within 30mph and most folks are actually driving within 10mph of each other.
Also England and half of Australia (although it's 'passing on the left' that's illegal because cars drive on the left there.) I used to think it was a silly law but now realise that it's not actually about passing on one side being safer, but about creating social pressure to force slow drivers out of the fast lane (since they can't just expect people to go around them.)
Highways are relatively easy (always the first target for almost any type of driving assist). And highways account for large amounts of the distance people drive. Maybe not for the time spent driving though.
If you can do full self-driving on highways and similar roads, you can start shouting about full self-driving on high percentages of distance driven.
Strange. It's very impressive, but it feels less impressive than the demo that released in 2016 [0], because the 2016 version had significantly more time on non-freeway roads. I skimmed through the latest video and didn't see any pedestrians. The 2016 video does have interaction with pedestrians and cyclists [2]
The thing which most amazes me is how many people taking this one sped-up video of a country drive in perfect conditions as proof that Tesla has solved level 4+ autonomy.
Waymo shows a similar video and everyone's asking about whether it can correctly interpret a police officer's hand signals asking it to perform an otherwise-illegal maneuver at night in a thunderstorm, or change its own tyres, or come up with a 'right answer' to the trolley problem.
21 comments
[ 3.5 ms ] story [ 58.0 ms ] threadNow it seems very staged (probably isn't but still - i'd imagine they'd pick the most impressive video to post out of all their tests)
https://edition.cnn.com/2019/04/22/success/tesla-navigate-on...
Elon successfully uses the same 'distortion reality' shield that Jobs did. This forces his staff to keep trying to achieve his practically impossible demands.
Success requires that it keep working over and over while not burning out critical staff and investor's money.
If you can do full self-driving on highways and similar roads, you can start shouting about full self-driving on high percentages of distance driven.
[0] https://www.youtube.com/watch?v=VG68SKoG7vE
[1] https://youtu.be/VG68SKoG7vE?t=479
[2] https://youtu.be/VG68SKoG7vE?t=97
Waymo shows a similar video and everyone's asking about whether it can correctly interpret a police officer's hand signals asking it to perform an otherwise-illegal maneuver at night in a thunderstorm, or change its own tyres, or come up with a 'right answer' to the trolley problem.