This is inaccurate and this point is repeated. Waymo has a remote driver.
Waymo’s business is not viable because they have put several cost and practical constraints. Expensive operational burden, expensive car, expensive sensors. The more they expand, more money they lose.
Err, that article is about Autopilot, not FSD.
While FSD doesn’t pretend to Level 4 yet, Waymo shouldn’t either with the frequent use of the remote driver. Wasn’t it like every 5 to 20 miles?
Where do you get your data on the frequency of critical interventions?
Frankly, with hands off the steering wheel and attention who knows where, I would expect a very large percentage of critical interventions by the computer to result in accidents … which we are not hearing about.
No, Waymo does not use remote drivers:
- Autonomous mode
Waymo’s cars are fully autonomous and responsible for their own driving while in autonomous mode.
- Remote assistance
Waymo does have remote operators who can provide advice on strategic planning questions. For example, in complex situations, a remote operator can propose a path for the car to consider
AI synthesis of Waymo, asked the question “Does Waymo use a remote driver”
You can call me a curmudgeon if you like (actually, no need to because I call myself a curmudgeon), but I like holding a steering wheel when I’m in a car. Especially when my wife is driving.
I love all the safety features in my new car (adaptive cruise control, lane assist, blind spot alerts, rear camera, etc), but I like being in control. Even if it is just an illusion.
And can we please stop calling driverless taxis Robotaxis. Everyone knows they should be called Johnny Cabs.
Oh, and get off my damn lawn!
If the cars are really consulting the remote operator every 5 to 10 miles, that doesn’t sound very strategic.
I believe that was Cruise, not Waymo. I couldn’t find a figure for Waymo.
Again, the defining aspect of Level 4 is that the AV driver can take over the driving function without the need for a human (no, Waymo does not have humans drive the cars remotely). You don’t need a human in the driver’s seat. That’s a very hard thing to do. Waymo has done it, but for a very limited set of circumstances. Tesla has not done it.
That might very well mean that both companies are very far away from a profitable driverless taxi service. But certainly the optimism surrounding Tesla being able to flip that switch and turn on true driverless operation is unsupported. Regardless of whether the same is true of Waymo.
Disagree. Tesla is inching closer and closer as each version gets better. Meanwhile, supervised FSD is making money as adoption increases and now they are going global.
I expect that we will see no steering Tesla Robotaxis launch in back half of 2025.
But all of that can be true and Tesla can still be very far away from having true self-driving. That’s the point. “inching” can get you where you need to be, but only if your goal is some number of inches away. But if what you need to achieve is very far away from where you are, “inching” there won’t be enough to get their in the short- or even intermediate-term.
There is a huge gap between even a very good Level 2 ADAS system and actually being able to let the car drive without a human driver. The latter is necessary for a robotaxi service. There’s no indication yet that Tesla’s anywhere close to the incredibly low level of critical interventions necessary to let the car drive itself. So even if it’s very, very, very good at aiding a human driver, it can still be very far away from being able to drive itself safely enough to operate as a Level 4 or 5 system.
Everyone is speculating. No one know what the future will be.
You are speculating that Tesla could be far away.
I am speculating that they are close.
The basis for my speculation is my personal experience in the pace of acceleration in FSD capabilities over the last 8 years.
The other issue here is that the answer may vary according to the target. For FSD to be an added feature in an otherwise conventional car is straight forward because the human and the driving apparatus are already there for backup. A driverless taxi with remote drivers (probably used only after a pull to the side) seems more an economic challenge than a technical one, i.e., how many remote drivers does one need and what does that do to profitability. Fully independent taxis seems like a much more distant goal.
I don’t doubt that you’ve personally experienced an acceleration of FSD capabilities. I just think you’re vastly underestimating the relationship between that pace of acceleration and how far there is to go.
Upthread, you noted that during your 40 mile commute the number of interventions you experienced has fallen from 5-10 times a year ago to only 1-2 times daily. That may seem like a really fast acceleration of FSD capabilities. But it’s only half an order of magnitude within a year, and an intervention every 20-40 miles is nowhere near sufficient for a robotaxi system. And even if they’re noncritical interventions, that rate of noncritical interventions almost certainly denotes a rate of critical interventions that’s way too high, also.
If your experience were representative (which you are assuming it is), and Tesla could keep u; that pace, then Tesla would be 3-4 orders of magnitude away from being ready to run a robotaxi - which would put them past 2030 before they could turn on that kind of system.
Now, I’m not sure either of those things is accurate. You’re experiencing more interventions than most crowdsourced/third party folks report, but also reporting faster improvement than they are. Also, I’m not sure that Tesla’s rate of improvement is sustainable, given that they’ve acknowledged declining utility in the data they’re getting (more data =/= more useful data on edge cases).
Perhaps we’ll know more after 10/10, but I kind of doubt it. They’ll certainly be showing off the vehicle, but honestly the car itself is perhaps the least significant part of the challenges of self-driving. I doubt that Tesla will delve much into the weeds of their disengagement stats, and their forward-looking statements have always been aggressively (and inaccurately) optimistic about when the car will actually be able to drive itself. So we’ll see the car, maybe the app, see some folks do short driverless trips around the WB lot like Waymo’s been doing for more than a year or two…but probably not anything that would help us assess how close (or far) Tesla is from an actual self-driving system.
That is highly doubtful, IMO.
Tesla needs to pass a few checkpoints BEFORE they configure a factory to make such a car, such as:
Mike
Many of these questions will be answered in 10 days. Wait for it.
I don’t think any of these questions will be answered in 10 days. I expect they’ll show the car, talk about the Tesla Network (or whatever they call the app), demonstrate a driverless ride or two, and then issue some non-specific forward-looking statements about when they think all this might be ready. Since they’ve been projecting that they would have a car that can function without a driver within a year or two since (checks notes) 2016, projections about when they’ll have self-driving won’t actually answer the questions about when it will be ready.
This is how innovation works. Dream of solving hard problems.
Try, fail, getup try again. Rinse and repeat. Eventually success comes.
There are always salty naysayers on the sidelines that point to failure of others. In the long run, they are irrelevant.
Exceptionally unlikely. It’s nowhere close now, even as it drowns under reams of data.
How bad is Tesla's full self driving feature, actually? Third-party testing bodes ill5:11 am EDT Sep. 30, 2024
According to testing firm AMCI, Tesla’s FSD software can’t drive more than 13 miles without needing intervention.
We’re Just weeks out from Tesla’s big RoboTaxi presentation, where the automaker’s self-driving shuttle will be revealed, and third-party independent research firm AMCI Testing has some bad news that could hang over the event like a cloud. AMCI just completed what it claims is “the most extensive real world test” of Tesla’s Full Self Driving (FSD) software, ostensibly the technology that would underpin the RoboTaxi’s driverless tech, and the results are not confidence inspiring.
AMCI says its test covered over 1,000 miles of use and, in short, showed that the performance of Tesla’s FSD software is “suspect.”
If you think 13-miles intervals between instances where a driver must grab the wheel or tap the brakes is pretty good, it’s not just the number of interventions required, but the way those situations unfold. AMCI’s final point is the most eyebrow-raising (emphasis theirs): “When errors occur, they are occasionally sudden, dramatic, and dangerous; in those circumstances, it is unlikely that a driver without their hands on the wheel will be able to intervene in time to prevent an accident—or possibly a fatality.”
We’ve been promised this since 2016. For eight years now for crying out loud. This is not a solvable problem. This is not a problem with data anymore. It’s not a problem with training. Does it make human driving better, easier, less tiresome? Yes. Will it ever drive completely on its own? No. And we’re wasting tons of dollars, compute and people’s time trying to do so.
Wait and watch as AGI comes alive. Between autonomous driving and humanoid robots it is tens of $Trillions.
I mean, sure - at a completely different level of generality than what we’re discussing on this thread. Will mankind eventually develop a self-driving vehicle? That seems likely, sure.
But that’s a very different question than whether Tesla specifically will have a viable self-driving service within a commercially relevant time frame. Or to borrow the thread title, whether FSD is ready for prime time.
It’s an important question for Tesla, since they appear to have eschewed development of any other conventional mass-market vehicle models (a second mass-market sedan or SUV, for example) in favor of concentrating on autonomy. [It’s possible that’s not the case, of course - there are those who think that the “robotaxi” is a dual-purpose vehicle design that will also be developed with a steering-wheel version.] But if that’s where Tesla is going, it will be tough sledding for them if the “eventually” in your statement ends up being 2035, rather than 2025.