For example, Waymo claims 700k fully autonomous rides in 2023, which was more than ~0 Tesla rides that were fully autonomous in 2023.
To define fully autonomous, I'll use the simplest definition which is that no human operated any direct controls inside the vehicle like the steering wheel or pedals. (UPDATE: And no human in the car is actively monitoring the driving being ready to intervene.) A ride must be of a non-trivial distance (e.g. crossing the parking lot does not count).
I will delay resolution until both companies report results or until the result is extremely clear, or until it's January 2027, where I'll make an informed guess.
Update 2024-13-12 (PST): Rides in the Las Vegas Loop will not be counted towards Tesla's total, as they do not represent general self-driving capability on public roads. (AI summary of creator comment)
1,000
3.00That’s a Wrap! Waymo’s 2024 Year in Review
We served over 4 million fully autonomous rides this year alone, bringing us to over 5 million rides total.
@HankyUSA different resolution criteria (US only different thresholds) might explain some of the discrepancy.
@HankyUSA Dang, it sure does keep improving. In case it's not obvious, there is a person in the driver's seat ready to intervene. I'm not updating much on that video but the following is bumping my probability up a little, though 2025 still seems pretty unrealistic: https://www.understandingai.org/p/teslas-robotaxi-strategy-looks-a
@JamesGrugett My thinking is that Tesla is not on track to hit SAE level 3 in 2025, let alone level 4. You're right that if it does, it can quickly shoot past Waymo. I just think that's a huge if.
(Review: Tesla is currently level 2, meaning a human has to be actively monitoring and ready to instantly intervene. Level 3 is roughly, human in the driver's seat, can read a book but not fall asleep and has to be ready to intervene quickly if the AI beeps at you. And level 4 is like Waymo, with no human in the driver's seat.)
@JamesGrugett Even if Tesla figures out level 4, they won't just enable it for every Tesla everywhere right away, especially since level 4 means taking legal responsibility as the driver. It will probably be limited to certain cities at first. It will probably be available for certain models before others. Tesla owners will likely need to to pay (perhaps repeatedly) to enable the feature, and not all Tesla owners will want to use the feature regardless of price. Even if Tesla is unwise enough to rush ahead, there's going to be an incident, backlash and legal trouble, triggering Tesla to backpedal.
Of course that's all assuming Tesla figures out level 4 this year.
Tesla has the most data and exponential progress in self driving.
Quantity of data is not the only thing to consider. There's also the quality of the data and how it's being used. I suspect Waymo has higher quality data and uses it better.
How are you measuring progress in self driving such that Tesla's progress is exponential?
2.5 million teslas vs 1k waymos driving around.
One Waymo taxi driving around completes far more rides than one Tesla that spends far more time parked.
Waymo owns and maintains their vehicles. They can maintain and manage their fleet in whatever way they need to make level 4 possible. The hardware on those vehicles was designed over many iterations for nothing less than level 4. Waymo can further adapt the hardware design if needed to solve challenges such as poor weather conditions.
Tesla designed the hardware in its vehicles for level 2 and maybe someday level 4 according to their historically-overoptimistic CEO. Tesla probably prioritized aesthetics, production cost, and low maintenance more than Waymo did because Tesla had to actually sell the vehicles to consumers. I expect level 4 with such hardware is a greater challenge. Tesla must either reach level 4 with the hardware they sold years ago without the ability to adapt the hardware or they must require vehicle owners to get their hardware upgraded.
Shouldn't this market be like ~50%?
If you believed that then wouldn't your limit orders for YES be greater than 20%?
@JamesGrugett that’s lame. It seems like such a low bar to say that Tesla doing 1 million rides in a curved tub for 1 mile with no traffic or lights beats waymo doing 999k rides on city streets.
@JDTurk Is it "lame," or is it a really great way of raising incremental difficulty and finding synergies between companies? If you're smart, then you will realize that there's a lot of machine learning that can be applied to predicting what passengers are going to do just before and just after they board or disembark a robo-taxi. The Las Vegas tunnels would be a great place to gather that data, without the complication of traffic!
@StCredZero Yeah, it's not lame for Tesla to do this, it would just be lame for us NO holders to lose on a technicality, which this could potentially be.
To be clear, I'm betting NO because Tesla has not actually figured out level 4 autonomy yet and I think it will take well into 2025 if not longer for them to serve even their first fully autonomous ride. Whereas Waymo is serving >150k per week and growing. I think it seems like Tesla is close based on how impressive FSD is, but it's all about the number of 9's. If there's even one time in a month or a year where you personally had to intervene to prevent FSD from crashing, that's a far cry from the reliability level needed to have no one in the driver's seat.
Unless maybe it were limited to something very controlled and predictable like the Las Vegas loop? Idk, people complain that Waymo is limited in that they only serve rides where they've thoroughly mapped the routes. But as we know from Google Street View, that doesn't make Waymo's approach unscalable.
@WrongoPhD I see a non-zero chances that they won't have level 4 figured out and still start offering full self-driving
@AlexanderTheGreater The whole smoke and mirror of FSD disappears once Tesla accepts legal liability for accidents. The stock value would plummet! I don't see FSD ever being offered unless Musk is first able to unload his stock