Only idiots buy black-box based car insurance (whether Tesla or any other).
I mean, come on, insurance companies are already renowned for slippery policy wordings and the desire to get out of paying-out.
Do you seriously think an insurance company having access to your driving data is really being done for your benefit ?
I would be willing to bet its so that Mr Risk Adjustor at the insurance company can say "computer AI algorithm says it thinks you were driving 'aggressively' that morning".
And once "computer says no" you'll have a hell of a time trying to fight it.
Well yes, there's marginal value for both you and Tesla in excluding aggressive drivers from the program. Some company with worse telemetry eats the cost of the accidents from those drivers.
It's fine to argue that this is a privacy tradeoff you don't think should be legal but there's pretty obviously value to be had for both insurance and customer.
Monitored policies are almost always significantly cheaper. That's why people buy them. I won't engage on who the "idiot" is in that analysis except to say that Adam Smith probably has an opinion.
The reason this works is that it forces the consumers to self-partition into those who think they drive safely and those who know they don't. That doesn't correlate perfectly with actual safety, but it's clear that it's a good proxy signal. Insurance companies may be renowned for "slippery policy wordings", but they're even better known for doing good statistics.
Why would a source of signal that has aligned incentives for all three major parties (individual, manufacturer, and insurance company) not be selected for by the market?
It's not as if people are buying meteor insurance here. Automotive accidents are frequent enough to where if a policy provider fails to cover damage that really is in-policy they'll be sued, regulated, and shamed.
Even more so, since Tesla has a financial incentive to ensure that these benefits really do accrue to the driver.
Cars have been equipped with event data recorders that capture information moments before the crash for at least 20 years. It is part of the air bag controller. Insurance companies have equipment to read that data and use it against customers all the time.
You don't need a Tesla or other black box insurance to get cheated.
https://www.theverge.com/2023/1/27/23572942/mercedes-drive-p...
Mercedes version is extremely limited and is only available on $100k+ cars with tons of additional hardware.
I have Tesla FSD. It is certainly overpriced, but it has genuinely been getting a lot better in the last year. When I first got it I was disappointed because it was so useless, now I use it probably 70%+ of my drives.
IMO Tesla is the closest to a mass-market version of the technology.
Unfortunately the bulk of the internet seems incapable of recognising this and all too often gives hateboys a free pass.
I personally prefer more of a “let’s judge each product on its own” sort of person. I love my iPhone and AirPods, but I don’t personally have any desire to own a Mac or the new Vision Pro.
Tesla is stuck at level 2 because the level of reliability required at level 3 would mean they'd have to remove features to meet it.
MB is beating Tesla in the self driving game because they've released a car that is >0% self driving. Tesla is beating everyone at the driver assistance game.
Time will tell, but I suspect we'll see Tesla adopt some of the same strategies that other automakers are taking before this race is over.
That's the problem, though. The scope is always changing. Roads change, barriers get put up, roads are closed, potholes form, detours signs go up. As soon as any of this happens, the extremely accuracy mapping that systems like MB rely on stops working. This has been proven when this sort of tech caused literally a dozen self-driving cars to just randomly stop and block a single intersection a year ago.
> Tesla's approach doesn't have nearly as clear of a path to advancement.
The path is very clear and they're doing it.
> They need technology that doesn't yet exist to get there.
This isn't really the case. Self driving can be done with cameras and sufficiently advanced software alone. The software isn't 100% there yet, which is why they're still requiring human supervision. But the technology is definitely there.
> Some have posed some serious concerns about whether Tesla's feature-first maturity-second approach is an engineering dead end and will ever be able to mature to the point of being able to operate unsupervised
Sure, possibly, but MB's approach is by far the more certain dead-end due to my first point.
> Time will tell, but I suspect we'll see Tesla adopt some of the same strategies that other automakers are taking before this race is over.
I doubt it. The other automakers are using dated methods that have been attempted for decades. Tesla is the first to go balls-out on camera-only automation.
MB's approach will work when cars literally drive on rails.
MB is focused on highlines and providing little value for a very high price.
MB is simply no a software company I would trust. Company like Waymo have spend billions and billions to get to where they are and their software engineers are a hell of a lot better then those at MB if I know anything about German software engineering.
Tesla has deployed end to end neural networks that can handle a huge amount of situation and is getting better far faster then MB is improving its system. With MB approach it would take decades to get there and that if you assume really good execution.
It also implies that higher level == better, which is certainly not a given. Is a system that can drive well on 10% of roads 99% of the time better than a system that can drive on 90% of roads 80% of the time? Not objectively. The former is "level 3" and the latter is "level 2".
A solution that can get you to 3 quickly might never get you to 4. And just because you beat someone else to market with "level 3" doesn't mean their solution won't skip over yours straight to 4/5.
So who do you think is winning it?
At some point there will be a research breakthrough in ML with models that can extrapolate physics which is the current thing that is missing from self driving being actually good.
At that point Tesla is going to be the winner without a doubt because they have the hardware and data to train en masse and deploy to cars.
When I need to buy a new car in a few years (I honestly have no idea what kind of lifespan to expect), I'm looking forward to more options.