Yes, Tesla may be getting more data in the field, but Waymo is still much further ahead with actual fully self-driving car without anyone at the wheel (within a small region + good weather). Also worth pointing out that the quality of the data also matters, not only the quantity.
Tesla is ahead where it counts, sales. it feels like waymo was already 'successful' in small regions with good weather not requiring complex thought situations in 2018. meanwhile tesla is improving designs, expanding capabilities, and selling cars.
I'm not a tesla fan boy, but I definitely prefer their approach in terms of iteration (not sure if I agree with the tech choices).
Where it counts for what? Tesla has demonstrated it can sell non-autonomous cars, Waymo has not, but the GP's point is it's not clear what bearing that has on making an autonomous car work.
I'll give you a hint, the technology doesn't define the product.
Product here is "L4 self-driving technology", not "car" as you seem to be implying. Yes, Tesla is shipping cars, but they are nowhere close to L4 self-driving (aka no one at the steering wheel).
> they actually have a product in the market
No they don't. They have a car on the market, and an L2 driving assistant. As I was implying above and again repeated by the post you replied to, there's no indication that you can transition from L2 to L4 in a gradual manner, so "L2 product in the market" means absolutely nothing.
Here's an analogy for you, I'm saying that Google is ahead of Intel in making a Quantum chip, and you tell me that Intel is right now shipping millions of classical chips, therefore they are closer to making quantum chips. My argument is that the jump from L2 to L4 is similar to classical vs quantum chips, and shipping millions of L2 cars does not mean you're any closer to L4 technology.
Why is this where it counts? And if it were, wouldn't this mean that Ford/GM are way ahead of Tesla?
Source?
I suspect we will see a replay here of the ULA versus SpaceX history of the last ten years, and at the end, Google is going to look up and wonder "what happened" while Tesla continues to dominate the actual market.
My expectation is that Tesla gets better and better but doesn't move beyond a driver assistance package for a very long time. None of the hardware sold today ends up being FSD.
Waymo remains limited in scope by its business model. It rolls out driverless taxis to more and more cities, but you can't buy a Waymo FSD car, and it remains geofenced for a decade or more.
Which company wins the race to being able to buy a car without a steering wheel that will take you anywhere? I could see Tesla winning that.
But Waymo is in a position to capture significant economic benefit by being an Uber competitor and providing logistics in the meantime, if they can scale up quickly enough.
Regardless of how it goes, it's going to be fun to watch.
Waymo started with the iterative approach, and threw it out of the window after they observed what Tesla is observing now: Humans suck at monitoring. Their testers started doing all kinds of crap while the car they were supposed to be supervising was driving, including texting, make-up and taking a nap.
Waymo identified the risks, and had the luxury to take a step back, and re-aim the moonshot. Tesla made the same discovery, but can't (or at least didn't) go back and say "we need to re-evaluate our approach to FSD" - and instead keep trying to get people to stay attentive at the monotone task of monitoring the vehicle, while continue to keep the blame on the human when they crash and die.
Now Waymo is in the position of being not only the current leader in self-driving tech (having cars without drivers on the road), but also having the perception of being the responsible player in the field, putting safety over profit&sales. Having the first cars without drivers on the road allows them to help inform legislation in that space, while Tesla is still struggling to convince their test drivers (=customers) to pay attention and not crash.
There is a lot more that Tesla is lacking before self-driving. I'm baffled that people here are suggesting otherwise?
Recognizing traffic lane "darts" as a bifurcation of the current lane and not as a new lane to travel in.
Recognizing stopped fire trucks and other vehicles partially or fully obstructing the lane the vehicle is traveling in.
Comparing the amazing capabilities of Waymo cars to the Tesla toy is very misleading.
Every accident, i hear them talk about how autopilot is not driverless, as fully self driving does not imply driverless. As it is a driver assistance system that requires human oversight and control.
On the other hand, there are constant claims of Tesla being the only one with driverless cars on the road.
It’s way too early to know which approach will be the “winner”.
Personally, if i were invested in these companies my money would be on the ones developing autonomous car tech over the ones who aren't. Tesla and Waymo can both be leaders here.
What am I missing?
Here's the summary via 2 examples.
1. Example of gathering data that needs further labeling
To implement neural network (NN) to recognize stop signs they program the cars to recognize things that look like stop signs and deploy that to the fleet of over 600 thousand cars on the road. The cars send those "might or might not be a stop sign" images back to tesla and they get manually labelled and added to a training set. They gather enough images and the "recognize stop signs" feature is done. Apply the same logic to other recognition tasks: recognize cars, pedestrians, animals, speed signs, traffic lights etc.
Yes, labeling is expensive but everyone else (including Waymo) has the same cost.
But Waymo has even bigger cost of driving around to collect those images.
Tesla makes gross profit on every car they sell and they got 600 thousand people driving for them for free. Waymo has 600 cars, each of them reportedly costing over $200k and they have to pay drivers at least minimum wage for each hour of driving.
That's why Google is reportedly is spending $1 billion a year on Waymo and why they need $3 billion of additional investment to keep going.
2. Example of gathering data that doesn't need labeling
Consider implementing NN to recognize cut in i.e. other cars entering your lane in front of you.
They deploy a first version trained on small sample, running in shadow mode i.e. it makes predictions for cut ins but doesn't act on it.
When it makes wrong prediction, it sends the clip of the cut in back to Tesla. This doesn't need manual labeling. They know whether the car did cut in or not so they can rewind time and automatically label the past car action.
So in failure cases, they record what a human would have done.
If I were a investor that invested in individual stocks (which I am not), my money would be on Tesla over the long run here.