Self-Driving Tesla Fails School Bus Test, Hitting Child-Size Dummies… Meanwhile, Robo-Taxis Hit the Road in 2 Weeks.
Self-Driving Tesla Fails School Bus Test, Hitting Child-Size Dummies… Meanwhile, Robo-Taxis Hit the Road in 2 Weeks.
This test of Tesla’s self-driving tech is tough viewing: what Mark Rober did the fun way with his Wile E Coyote test of a fake wall in the middle of the road, the team over at The Dawn Project are doing the hard way by testing self-driving Teslas’ ability to stop for pedestrians at flashing […]
I have no confidence that Tesla will fix this before the planned Robo-Taxi rollout in Austin in 2 weeks.
After all, they haven't fixed it in the last 9 years that self-driving Teslas have been on the road.
You're viewing a single thread.
The article doesn’t mention anything about where the cars will be driving in Austin, but I reluctantly took a Waymo from the airport in Phoenix to a hotel and it did really well. Even slowed down for a guy who was jaywalking.
It ended up taking longer than expected because apparently they’re either require or trained on different roads than you would normally take (e.g. no highways). It did such a good job, though, that I ended up taking a Waymo a second time while there.
All this to say that it may not be ready for all roads, but is ready for some. Definitely still scary though.
I believe Waymo has a better set of sensors (Lidar + Radar+ Cameras instead of just cameras), more processing power, and more research / time / resources spent on it compared to Tesla.
So it's not that we aren't ready for self driving taxis, but rather about which cars are ready to provide that service
I think Waymo is also trying to prioritize safety. I was in San Francisco recently and took one, just out of curiosity, from my hotel to a Giants game. It seemed to stop when pedestrian traffic got heavy instead of going all the way to the stadium. So, like three blocks from the stadium. No biggie. I might have told a human taxi driver I could walk from there.
I’m not sure if it’s a California regulation or Waymo trying to play it safe but I will never get in a self-driving car regulated by Texas and designed to the specifications of one of history’s biggest dumbasses.
Working with cities to regulate self-driving and plan out specific routes/infrastructure was always going to be the only path to widespread adoption but Elon was too busy grifting off bullshit claims like everyone's Teslas moonlighting as self-driving taxis and paying for themselves.
Makes sense. There were sensors alll over that thing.
That's part of the reason Teslas are not well-suited for this. One camera, each direction, with no other sensors to help make decisions, is a really bad way to ensure safety.
Humans normally have two "front facing cameras" (i.e. two eyes) so we have depth perception. We also process light differently than cameras do so infrared light (for one) doesn't affect our decisions. We also have ears so the sound of a loud motorcycle engine tips us off if we just see a spec in the distance. We also use context clues to help our decisions, like if other drivers change lanes quickly we are extra observant of road obstacles.
Not that technology can never be as good as a human at driving, but we use a lot more than a single "moving picture" to decide what we should do.
To be fair, the Tesla vision system has 3 cameras facing forward. One in the center above the front bumper grille and two behind the rear view mirror. Those two provide some level of stereoscopic vision to help judge distances.
But yeah, the lack of other sensors is a huge issue. Anything from bug splatter to mud to snow etc. can easily obscure one or more cameras and render the whole vision system unreliable.
We also process light differently than cameras do
To expand on this a little further, human vision has also developed the ability to filter out unnecessary information in order to avoid overloading the brain. When tracking moving objects the eyes mostly send deltas of the movement to the brain. Computers, however, are the exact opposite. The cameras essentially send a series of still images, and it’s up to the computer to compare them to look for any movement.
Waymo is really interesting - you probably wouldn't guess it, I'm a cautiously optimistic autonomy person! Waymo is already 12x safer than human drivers, that's brilliant, I love that.
Teslas will (allegedly) start on a small, low-complexity street grid in Austin. exact size TBA. Presumably, they're mapping the shit out of it and throwing compute power at analyzing their existing data for that postage stamp.
The rub... that all points out the obvious danger of rolling out the wild-west FSD that Tesla drivers are currently employing everywhere else. If it's safe enough to trust to drive your car for you, why does it need a ton of additional guard-rails to operate without a safety driver?
Yeah it’s scary to think about. There should be laws though that you’re still 100% at fault if you were not driving during an accident. I imagine another issue with FSD is government having a backdoor into your car to immobilize you or whatever they want. Part of me is in favor of that, but of course that a huuuge responsibility that can be abused.
The fun part is tesla FSD shuts off just before accidents, so you're always the one at fault.
You would be the one at fault in most states anyway as long as you're technically operating the car. They do that mostly for potential lawsuits from their customers.
So you’re saying it’s good at anticipating accidents haha
Teslas will (allegedly) start on a small, low-complexity street grid in Austin. exact size TBA. Presumably, they're mapping the shit out of it and throwing compute power at analyzing their existing data for that postage stamp.
Lol where are the Tesla fanboys insisting that geofencing isn't useful for developing self driving tech?
Can Waymo sometimes use a remote human driver?
I am optimistic about self driving, just not Teslas. Unlike other self driving cars, Tesla is exlusively reliant on cameras. Others, like Waymo, have cameras, lidarr sensors and radar.
It's weird--Tesla at least projected some sort of 'higher-end' quality back when they first started coming around but as time has gone on it's proven itself to be a very cheaply-made vehicle.
The fact they insist on using their proven-inferior tech for FSD tells you all you need to know: they aren't looking to innovate or even bring themselves up to modern standards. They are determined to make their cars with the cheapest, shittiest tech and they'll just grease the right palms in order to proceed as necessary.