What about Waymo? I took a half dozen cab rides in San Francisco last month and not a one had a driver in the car.
Well, in a limited way, we do have autonomous vehicles. But could it hop on the freeway and take you to to the next town? Can it operate in rain, fog and snow? I could do any of those.
Nope. Currently boxed into San Francisco proper (and less relevantly, LA and Phoenix) but within that area it could go anywhere on surface streets. And it was flat out astonishing. It dealt with gnarly SF streets and traffic flawlessly. It was certainly the most ideal set of taxi rides I’ve ever taken.
I watched about 20 minutes of the livestream waiting for something to happen, then I turned it off. This was basically like a very boring sci-fi movie that happened to have a big special-effects budget, or alternatively, like a big-budget display at a world fair pavilion…
I was astonished at how content-free this was. The only takeaway I got from it was “look at the all the toys you can build when you’re as rich as I am”. And that’s pretty much it.
Look, I have to admire someone dedicated to taking us into the future and giving us a glimpse of his vision of it, but it remains to be seen whether Elmo will contribute anything more substantial than the superficial glitz we saw tonight.
So, yes, a very limited autonomy. I do think that what they’re doing is more impressive than what Musk has accomplished, but I am skeptical about their ability to generalize it beyond its current limited abilities.
I would characterize it differently. I’d call it near-complete autonomy but in a limited area. In any case, I felt safer it the cars at every movement than I ever did in our Tesla even with it only attempting monitored autopilot (not FSD) driving.
Speaking of Waymo, I thought I would provide an update on my experience from earlier. I’m now up to 201 rides, 677 miles, 3568 minutes (data helpfully tracked by the app). Rides continue to be mostly straightforward and unremarkable, except for the time it ably avoided an accident from another car pulling out right in front of it. The route choice is occasionally questionable, but it gets there in the end. You can also sometimes tell the software must have been updated and the car gets “skittish” avoiding even the vaguest hints of problem situations, but it tends to go back to normal driving within a few days.
Can’t argue with how you felt. But Wikipedia sez:
- By July 2021, the NHTSA had found 150 crashes by Waymo.
- A Waymo robotaxi killed a dog in San Francisco while in “autonomous mode” in May 2023.
- In February 2024, a driverless Waymo robotaxi struck a cyclist in San Francisco.
- Later that same month, Waymo issued recalls for 444 of its vehicles after two hit the same truck being towed on a highway.
Waymo claims it’s safer than a human driver. Maybe safer than some. But I’ve been driving for half a century and (a) never struck a cyclist, (b) never killed a dog, and (c) never had an accident, period. Waymo is clearly not yet where it needs to be, even within its current extremely constrained operating conditions.
The raw numbers there aren’t that interesting. The real statistic would be to put those side-by-side with human drivers on a per-trip and/or per-mile basis. Bad shit will happen in cars, no matter what — and the meaningful comparison is whether the computer drivers are net safer than the human equivalent.
The relevant question then is, “which human?” The average driver? Just like we see with AI in other contexts, the AI of full self-driving may be better than average human drivers in some respects, but will also be different, and have different failure modes, perhaps unexpected ones. For that reason alone, standards should be much higher than for human drivers.
I mean, that just means you’ve driven about ~700K miles vs the 22,000K miles driven by Waymo thus far. When you’ve driven for another 1500 years, report back on your accident stats then.
Sure, but the point is, a raw list of anecdotal incidents is meaningless out of context. Like, let’s say that, every year, one Waymo taxicab drives off a bridge and plunges to its fiery doom and kills its passengers — but that’s the only incident every year, and the taxis otherwise have a perfect annual safety record. Given the human-caused carnage on our highways, one high-profile spectacular fatal accident by a robot driver is an easy tradeoff to accept if it eliminates hundreds of other crashes. Obviously this is an absurd extreme example solely to illustrate, but the point is, “there are 150 crashes on the record” is a nothing fact all by itself.
I think your math is a little off, but I can report back to you right now. When Waymo has successfully driven in some of the conditions I have – howling blizzards with whiteout conditions, roads covered in black ice, unlit rural country roads in blinding rainstorms, unexpected detours in the middle of a dark night – then I’ll have more confidence in it than I have now.
You have a valid point, in the same sense as airline travel: may feel scary for some, is actually very safe, and on a mile-for-mile basis on average it’s much safer than driving. But that took a very long time to achieve, and the evidence is nowhere near there yet for proving the safety of self-driving vehicles in all the conditions that they might encounter even in a protected environment.
For your plan to work, I’m pretty sure the Waymo needs to plunge into a volcano. And the passengers need to be virgins. I guess they could track that in the app profile.
Hey! You noisy bastards take it to the autonomous vehicle thread!
Huh, I’ve already got access to a large local transit network of vehicles on a very specific, very limited set of routes so carefully mapped out that the vehicles are, in effect, on virtual tracks.
And each of them also comes equipped with a fully functioning uniformed human driver so I never have to worry about control of the vehicle at all.
Has Musk basically spent all this time, money and hype just to come up with an inferior alternative to city buses?
(Yes, I get that the ultimate goal for autonomous vehicles is to provide much more capacity and flexibility than that, but it doesn’t look like the tech is really going to get out of “not significantly better than the bus system” zone for quite a while, at best.)
‘No fair! All of my ideas for improving mass transit are inferior alternatives to what has existed since the 19th century!’ ~ Elmo
Is it me, or has usage of “Elmo” actually ticked up since a Musk fan complained about it? 
Musk fan or “not a Musk fan”?