Hacker News Reader

Comment by crancher | Hacker News Reader

From page nine:

Services and Other - Insurance

In Q3, we rolled out our “Safety Score” functionality, which will also be used for our telematics insurance product. We actively monitor braking, turning, tailgating (unsafe following), forward collision warnings and forced autopilot disengagements in order to predict the probability of a collision. This system will continue to be fine- tuned as we receive more data. We also launched our telematics insurance product in our first state – Texas – in early October. We believe our insurance premiums will be able to more accurately reflect chances of a collision than any other insurance product on the market. Additionally, we will proactively communicate to the user what driving adjustments need to be made to decrease probability of a collision.

View on HN

This could backfire on Tesla. If Tesla have better data than every other insurance company, then they'll be cheaper for low-risk drivers. But now other insurance companies know the only drivers that will not use Tesla insurance are high risk, so Teslas become harder to insure with all other insurance companies for a large fraction of drivers (and you don't know in advance if that's you), which in turn makes them less desirable to buy in the first place.

Yup, this is "adverse selection"

Or maybe people will now have an incentive to drive more safely if they know it will save them money

This has always been the case, regardless of any Tesla involvement. This just makes the correlation between safe driving and lower cost a little less abstract.

It will be interesting to see how this develops. Will they begin selling this data? I imagine it's valuable for rental car agencies, ride-share apps, etc etc.

They probably won't sell it but i can imagine them licensing access to the real time data and then other insurers could potentially insure the driver or some similar scheme

Selling access to the data in some fashion (think how Google, FB, etc allow advertisers to buy access to a certain user to shoaw ads) as a recurring business model is probably more profitable

Luckily for Tesla, pretty much everyone rates themselves as a good driver, so they'll all be expecting to get the good low-risk rates.

Nah, I'm pretty aggressive and I know I should stick with dumb insurance.

> But now other insurance companies know the only drivers that will not use Tesla insurance are high risk

Not necessarily. I'd pay more for car insurance to avoid being tracked this way, but that doesn't make me higher risk. Or maybe I just want to keep my home and auto insurance bundled so I can save money on both.

The fact that there exist other reason to use a 3P insurance company doesn't really matter.

At the end of the day, if the existence of a Tesla insurance removes a portion (not all) of low risk drivers from the insurance market, the remaining drivers would be higher risk. The insurance company doesn't care about the mechanism that causes it, at the end of the day the collision rate will stabilize into a new higher rate, together with the associated insurance rate.

Insurance companies are already over pricing Tesla coverage, which is why Tesla is bothering to get into this market at all. The end effect here might just be the rate they are offering now becoming justified as Tesla plucks all the lower risk fruit away from them.

> I'd pay more for car insurance to avoid being tracked this way, but that doesn't make me higher risk.

Can you even opt-out of being tracked by Tesla? Even if you're not buying their insurance, I'm sure their EULA will grant them access to your data.

There's options to opt-out of various kinds of data-sharing in the car menu system.

Didn't think about this. Smart comment.

The holy grail of insurance: each person’s individual risk behavior monitored by AI.

Progressive patented this in 1998

https://patents.google.com/patent/US5797134A/en

The worst nightmare of consumer privacy: A car that monitors your every move.

My wife activated this on her insurance. Got an email last week saying that she is in the top 5% of “worst drivers”. Which was both scary and funny at the same time. We investigated why they deduct so much points for her driving, and the majority of the points was conducted for her being on het phone while driving. Then we realized that we drive around with her car most often and she is on the the phone while I drive..

These models has it’s problems but they are getting better - it’s just so easy to make wrong assumptions from ML models which only include a subset of variables.

“Congrats you win worse driver in America award!”

How do you inspire the lowest performers?

Coming up next, a fingerprint sensor on the steering wheel to determine who is driving!

and a mental state, attentiveness, etc. sensor - "Your driving performance in the next hour is estimated to be 10% worst than your average, and will incur 20% insurance premium increase for that trip. Do you still wish to start the car? "

Me: No, I wish to sit in the car and bang my head on the steering :(

I think the new Genesis cars have this as a feature, on the start button. The upside is you don’t need to carry keys.

So it is measured with a phone app and not e.g. an OBD dongle with built-in cell modem? Why not install the app on a second phone which always gets left at home?

We're on the path where each of is monitored by (hopefully benevolent) AI from birth to death. Such a scenario is really the only way for humanity.com to continue scaling, as each of us has the potential to very negatively disrupt anything/everything.

If this were accurately reflected in the rates I'd be interested, but past experience suggests that it's going to be more like "if you drive perfect you might get $5/month savings, but if you brake hard that one time your rate will shoot up by $50/month".

You assume extra risk that the company won't screw you over for what is likely to be a very modest upside. There are lots of case studies of people who got massive unexpected bills because they were trying to save a few bucks. Especially since they will be able to assign the blame to the customer for their high rates. Never trust corporations more than you have to.

Hmm. Car insurance is one of the most competitive industries in the US. Lots of players, slim margins, commodity product, etc... I wonder how long a company could charge people $50 more for braking hard and not lose customers. I know if I saw a surprise like that, I'd be getting quotes immediately.

There's lots of places where tiny upsides come with huge downsides. And yet for some reason people still accept them! My partner's manager signed up for a summer energy program that could cut off her HVAC system in exchange for a super-paltry discount on her monthly electricity bill. She cursed her decisions a few hours in to the next heat wave.

I don't think they are all that competitive when it comes to insuring electric cars.

Teslas at least are very expensive to repair. Despite the high premiums, they may be competitively priced.

I suppose time will tell. If Tesla's insurance product becomes popular (in states/jurisdictions where it's offered), that would imply that big insurance companies are not pricing competitively.

Tesla gets to cheat a bit here, in that they don’t have to charge themselves markup on replacement parts for this own cars. So in theory they should always be able to outprice other providers for Teslas.

They also are only marketing to their own customers who generally have much safer lives than the Altima and Town and Country crowd.

When FSD matures it won’t matter how good of a driver you are it will matter what percentage of time you let it drive you.

Traditional insurance companies have been offering a version of this for a while now. Apparently the term of art is usage-based insurance. https://wallethub.com/edu/ci/usage-based-insurance/14118

The issue with this is that in the limit it just makes the insurance premium exactly equal the payout + some margin. Some of the point of insurance is cost sharing of extreme events.

It's perhaps a little better when we are talking about behaviors that are entirely under the driver's control, but gets potentially ugly if for example someone has slower reaction time due to genetics or health and thus gets priced out of insurance.

>The issue with this is that in the limit it just makes the insurance premium exactly equal the payout + some margin.

At which point there's no point in insurance, it's a glorified bond or escrow service depending on how you squint.

> Some of the point of insurance is cost sharing of extreme events.

I guess this borders on a philosophical point, but I don't agree. car insurance is not (or shouldn't be, imo) a payment plan for likely crashes. if, for whatever reason, a person's reaction times are so slow that they would be priced out of insurance entirely, they shouldn't be behind the wheel at all.

Seems like the most important thing you can do in regards to this is put a little camera on the dash and monitor for sleepy behavior.

I would bet this trumps all of that junk wrt crashes by a significant margin.

Or "texting while driving" behavior.

There is a camera pointed at the driver, at least in the model 3, near the rear view mirror. I believe it was indeed installed to monitor the driver.

Both the 3 and the Y, and it actively monitors your attention while Autopilot is engaged starting with Tesla software rev 2021.36.

Looking down at a phone, for example, causes Autopilot to alert, and three alerts in succession reduces your safety score and kicks you out of Autopilot [1] until you go through a Drive->Park->Drive shift cycle.

[1] https://www.tesla.com/support/safety-score (Control-F "Forced autopilot Disengagement")

Oh goodie. I have a Model Y and it literally always goes insane with forward collision warnings in the parking lot for my local grocery store. I guess I'll have to pay more in insurance eventually because tesla thinks I am going to run into a parked car off to the side of me while doing 7 mph.

While Tesla Insurance has potential, I'd don't recommend it at this time.

“Have become inattentive” is doing a lot of work in that description. It means “you fail to respond to the prompt to touch the wheel”.

I’ve gotten double digit red-text autopilot nags (the new ones driven based on camera data) on a single drive, but have never had a Forced AP Disengagement or been put in AP jail.

Wait, Autopilot, or FSD? Or both?

It's interesting that early user feedback indicated that if you wanted to max your Safety Score you actually had to drive unsafely in specific ways. For example, stopping correctly at red lights and stop signs would show up as braking, so gently running yellows/hollywood stops would produce a higher safety score.

Considering that the car also sees the stop signs and traffic lights, that would have been an early days bug more than any other thing.

A yellow light is not a stop sign, you are allowed to drive through it if you can make it. And it is safer to do so than to slam on the brakes, which might cause the car behind you to hit you.

in the USA, it's just the reverse.

You must stop for a yellow (amber) light if it is safe to do so.

Nope.

States vary somewhat in their laws, but in my home state of California, the laws state only:

California Vehicle Code (CVC) Section 21452(a): "A driver facing a steady circular yellow or yellow arrow signal is, by that signal, warned that the related green movement is ending or that a red indication will be shown immediately thereafter."

CVC Section 21454: "When lane use control signals are placed over individual lanes, those signals shall indicate and apply to drivers of vehicles as follows: (a) Green indication: A driver may travel in any lane over which a green signal is shown. (b) Steady yellow indication: A driver is thereby warned that a lane control change is being made. (c) Steady red indication: A driver shall not enter or travel in any lane over which a red signal is shown. (d) Flashing yellow indication: A driver may use the lane only for the purpose of making a left turn to or from the highway."

A yellow light on a traffic control signal in the state of California only indicates an upcoming transition to a red light, which a driver must stop for. There is no requirement that a driver must stop for a yellow.

Thanks for the citation. I'm hereby wrong.

States vary significantly on this one. Oregon expects you to stop unless it is unsafe to do so. This difference is part of why we hate California drivers.

https://oregon.public.law/statutes/ors_811.260

The fact that you say "in the USA" on regarding an issue that did not exist in the 18th century and is not one of the handful of things we have chosen to federally regulate since is a massive red flag that you're wrong.

And FWIW the driver's manual for my state offers no guidance and simply says it indicates the light will soon be red.

Since the other person who replied said their state on the other side of the country does this I suspect it's one of those things where there's a federally issued suggestion (probably from the NHTSA) that some states implement and some don't.

Part of the safety score is based on g forces. So if you do a smooth stop by braking ahead of the stop, you’ll have less g forces. If you slam on your brakes, well, you’ll have a lower score due to higher g forces.

I'm very curious to see if tracking the actual driving data turns up any surprises in terms of safe behavior. What if say tailgaters turn out to be safer than average drivers? I don't expect that to happen, which is exactly why I'm curious to see if it does.

I'd put a good wager that "speeders" may fall to that category.

I find that I generally travel above the speed limit and feel far safer than I do when I set my cruise to the proper speed limit. Why? Because when I'm passing others I get to choose how long another vehicle is in my blind spot, or how often someone is trying to park themselves in my back seat. When I go the speed limit I always end up in a pack of cars all jockeying lanes with someone far too close to my rear bumper and a guy in front of me swaying up and down in his speed... I feel generally unsafe leaving control of my speed/position up to the other cars on the highway.

* Very obviously, these are subjective takes and likely don't reflect reality. No vehicle accidents in 17 years makes me think I'm doing something right though

My brother's motto is: if you're the fastest person on the road you don't have to worry about anyone behind you.

Apparently he learned that driving boats in the Navy.

Sorry, watch Canada's Worst Driver. Every time someone comes on and says they drive better by going fast they fail almost all the driving tests in braking, cornering, dodging sudden run-outs into the road and even just the simple drive better two cardboard boxes with two feet extra distance apart compared to the width of the car.

Don't believe me? Try: https://www.youtube.com/results?search_query=Canada+worse+dr...

Sure but if they are on that show they’re going to be bad no matter what their opinions are.

True. I thought a bit about that after submitting my comment but I figured going back to write in a whole bunch of caveats would just derail everything.

I do a lot of highway driving, and Saskatchewan doesn't see the same level of congested traffic that you'd get in bigger metropolitan areas. So my comment really does have a lot of particulars that aren't universal. I take a different approach to driving while in the city which does not involve frequent speeding.

Reply on news.ycombinator.com