I bought a Model 3 back in 2020 and got the self-driving option mostly on a whim. Until recently, it was a handy feature on freeways but a gimmick on surface streets. Ever since the update a couple months ago, it actually does drive itself. I can't remember the last time I intervened. If I did, it was probably to make the car do something dangerous or technically in violation of traffic laws, as you have to be aggressive sometimes in city driving.
You also have people like John Carmack saying, "I have made multiple two hour drives without touching the wheel." His main complaint is that the car makes him touch the steering wheel if he's wearing sunglasses (because it can't track his eyes to determine whether he's paying attention).
I don't think owners will rent out their cars as taxis any time soon, but the technology really does seem to be there.
The technology really isn't there. Individuals' subjective experience of safety is totally irrelevant when you're talking about a system as huge as "automobile travel."
Even 99.999% reliable miles is still dangerous at 3.2 trillion miles per year
Would the comparison not be human drivers? If it's 99% percent reliable I would say that's far safer than humans and makes far more sense already? What am I missing
You think humans are only 99% reliable on a per-mile basis?! That would mean your average person crashes every ~100 miles, or about once every two days.
You can not just go around asserting comparative advantage when your baseline is so far off it is farcical.
I did not have the same experience with the free trial. It tried to crash in to curbs, and really drives in ways that make you think it isn't really planning very far ahead: driving towards parked cars and swerving at the last second instead of making an early correction to move over in the lane slightly, driving too fast for conditions, hard braking at stop signs, stopping very early, or very late for stop signs. This is in addition to all the other bad behavior: getting confused about where lanes are and blaring the alarm sound, taking no regard for other cars when merging, trouble following curves smoothly and respecting limit lines.
It was a couple months ago, whenever the free trial for everyone was. Bay Area peninsula suburbs. The neighborhoods that Tesla engineers live and drive in frequently.
We must have the old version in Australia because we bought extended autopilot for our 2024 M3 Highland and it drives for about 10 seconds before drifting into the left or right adjacent lanes and making everyone nervous.
And it tried to kill me on the freeway when it pretended it was going to merge and then just kept driving straight into the gutter.
In fact I have been thinking about asking for a refund because it is LESS capable than our 2020 Corolla’s adaptive cruise.
Really crappy situation to be waiting for someone to fix software that you paid for. If EAP is actually OK overseas, why don’t we have it here??
In Australia, self-driving features beyond lane keeping are banned, and lane keeping is limited to imparting a maximum lateral acceleration of 3 meters per second squared. Because of these restrictions, all Teslas in Australia are using a gimped version of autopilot from 2-3 years ago.
Well if two people say it on the Internet without providing any documentation, evidence or references to larger-scale reports, that sure proves the claim, doesn't it?
I've commented on a few Waymo threads about how I'm a total convert to self-driving cars (or at least to Waymos) so it does seem like it's possible for autonomous cars to work at some level.
On this subject, though, I happened to ride a Waymo to a destination only 1-mile away (hurt my knee last week! Not that lazy! Cost ~$8.00 btw) to meet a friend yesterday and it's the first time I got into the Waymo and didn't even watch the road because I guess I'm over the hump of worrying it's going to do something dumb (probably not sensible to give over so quickly, but just sharing how my behavior has changed over maybe 15-20 rides). The friend I met also happened to have a Model 3 with FSD and he offered to give me a ride home in it with FSD enabled. Aside from this blue prompt on the display that forces the driver to interact with the wheel to say "I'm still here" it didn't have any material difficulties with the route. Very short drive of course. But he did have to intervene because of the prompt at least 3 or 4 times, and even one time when we were stopped at a light FSD started turning the wheel even though we were going straight. You could see the car to the left of us get nervous because it noticed this wheel turn and so they moved over a little fearful that the Model 3 was going to go into their lane either right there at the red light or when it turned green. However, once the light turned green the wheel went straight and FSD just proceeded us through the intersection. An oddity for sure. We couldn't quite figure out why it thought it needed to turn the wheel.
The reason I bring this up is to share a datapoint of 1, but also because it got me thinking about Tesla's rollout of robotaxis. It seems like before they actually let these cars drive without a driver they're going to have to remove that blue, take-the-wheel prompt while drivers are still in the driver's seat. And then they'll have to have a bunch (hundreds of thousands, millions?) of miles under their belt where there's zero/near-zero intervention from the driver to confidently deploy an actual robotaxi. So I guess the point at which you hear that Tesla no longer requires the driver to take the wheel is the point at which the public will know that Tesla now thinks it's getting close to having a robotaxi. That assumes Tesla does the sensible thing of course and takes that intermediate step. Would be odd to skip it for sure while you have a "safety driver" in the driver's seat.
Assuming Waymo is able to drive safely, it seems like a matter of time until Tesla is able to achieve the same, no? I can’t see why Waymo should be able to do it but Tesla should not given enough time.
This could be outdated info, but don't they have quite a bit of hardware bolted to the car that Tesla doesn't have? Tesla seems to be betting that they can work some software magic with fairly basic hardware, whereas Waymo (I think) has taken an approach that relies on input from quite a few more sensors.
That is not how any technology works, let alone literal bleeding edge technology, unless you mean Tesla will just scrap their technology and then just buy it from Waymo.
Assuming SpaceX can make reusable rockets, it seems like a matter of time until Blue Origin is able to achieve the same, no? I can’t see why SpaceX should be able to do it but Blue Origin should not given enough time.
Assuming Microsoft can make a desktop operating system, it seems like a matter of time until IBM is able to achieve the same, no? I can’t see why Microsoft should be able to do it but IBM should not given enough time.
See how ridiculous that sounds. Serious technological R&D is not something you can just pick up off the ground with money and time.
I would be surprised in 20 years if nobody is making rockets similar to Space X.
For OS - which is software, more similar to EV - Linux devs are able to make a much better OS than Windows, commercial success on the consumer front be damned. OS is pretty much commoditized but everybody uses Windows for enterprise because of vendor lock in, not because it’s hard to make an OS. Arguably, windows is the worst of the three popular OS’s despite having the largest commercial success.
OpenAI was able to make great GPT models and everybody else followed suit pretty quickly.
I think bleeding edge tech with larger moats takes longer than bleeding edge tech with smaller moats. The moat for self driving is low enough that there are many companies pursuing it with only VC funding. Compared to self landing rockets, which is being pursued by two companies backed by the richest men on earth.
The argument you are making here is that somebody else will figure it out. However, the argument you made before depends on that any arbitrary entity will figure it out with time and money. Those are categorically different.
If, in the first place, you actually meant that somebody else will figure it out, then it is logically unsound to argue that a arbitrarily chosen entity, Tesla in this case, is a shoe-in to figure it out as well. You need to present actual arguments for why they in particular will be a somebody.
If you actually thought the argument you presented initially was logically sound, then you meant everybody and the arguments you have just made are irrelevant as they only argue “there exists”, not “for all” which is what your initial argument rested on.
tl;dr You are mixing up “there exists” and “for all”.
Because of underlying decisions about their respective approaches, it's not certain that Tesla will actually be able to achieve the same as Waymo. Specifically, Waymo does pre-mapping, and uses LIDAR, probably other things, that Tesla won't do, so with those two differences, it's not clear that the computing power is available for Tesla to achieve the same level of self driving as Waymo.
Seems like quite the blunder - maybe Tesla thinks they can add LIDAR to their own fleet of robo-taxis. Otherwise they bet the future of the company pretty foolishly.
Hardware stack. While Waymo stuffed their cars with all possible radars and lidars, Tesla sticked to cameras. I didn’t wash my model Y for a while and it is continuously complaining about not working cameras. Looks like that camera based autonomous driving isn’t low hanging fruit.
