"Organization XYZ has godlike power thanks to Palantir, says Organization XYZ two years after investing half a billion in Palantir."

As far as I can tell, Sompo maintains a vested interest in hyping AI for business. Palantir has been on the rocks recently.

It's not every day that you see an insurance company violate several of the ten commandments in one fell swoop.

Article is about health and property insurance, but Tesla has a feedback loop that game-ifies safe driving: use PCA to identify which behaviors actually correlate to accident probability > give each driver a real-time safety score > use that safety score to determine their Tesla Insurance rate > driver is financially motivated to driver more safely.

Tesla Insurance saves drivers a ton of money, they rarely have to pay out because drivers become safer, and drivers make the roads safer by acting in their own financial interest.

I'm afraid to sign up for Tesla insurance because their collision warning system constantly goes off on me when I'm, say, driving slowly past an RV parked on the side of my street on a curve (which I do almost every day), or creeping up on a turning car in a parking lot at 5mph, fully aware and planning on stopping before I smash into it, but not coming to a complete stop because I know they're just about to make a turn. If they add those metrics to their system my insurance rates would probably sky rocket, even though I've never been in an accident or even got a moving violation ticket in 22 years/300,000 miles driving.

Yup. Mine recently slammed on the brakes and came to a complete halt for literally no reason. (Not using autopilot) Tesla’s answer? Couldn’t replicate. Updated firmware. Resolved. Next! Do not ever trust this company. Trying to get answers or actual resolutions out of them is like trying to catch a greased pig. I’m sure their insurance experience will be just as lovely as their repair service, their quality control, etc.

Looks like a PR puff piece.

The "godlike" adjective appears to refer to what is being predicted (longevity) as opposed to the quality of the model.

There aren't any numbers shared in the article which would indicate a "godlike" prediction performance improvement.

How do you argue or reason with someone using a black box model? If you can't, what is it really.

I don't think you argue. You're always free to choose some other insurer if you disagree with the premium they're quoting you.

Where I am, insurers form a "radical monopoly" where they are an esconced tax on people as a legal requirement for a whole set of services. They all use the same models and data, so there isn't choice. There is what you are legally mandated to have at the lowest possible price, and they have no value add beyond that. It does not resemble a market.