Comment by SecretDreams
1 day ago
They make cars that are very safe from a testing protocol perspective, but cars that might also fall off their performance cliff quicker in some real world scenarios where the energy input is just a tad higher.
They engineer to robustly pass tests, seemingly without a lot of consideration of "beyond" the test. It's not uncommon for most OEMs to target the test metrics, but you can still make design choices along the way where you may recognize with EJ when there is a "design cliff" and you may avoid it.. if desired.
Test metrics are just plain silly. For example, if your car emits 100 ppm of some gas, it is illegal to drive it. If it emits 99 ppm, it's good to go!
They're bang-bang functions, which aren't very smart.
OK so we allow a car with 100ppm to go. Ok now my car is 101ppm that should be fine too right? How about 105ppm? Where is the actual limit?
In reality, when you're getting close to 99ppm you start getting "borderline" notices on your MoT that you should look into this before your car ends up in the failure zone.
The solution is to use a continuous function, not a step function.
2 replies →
Well, while I agree, they have been incredibly useful in improving car safety. Cars used to be death traps. Now, most people will be able to walk away from most car accidents. Modern car safety, especially on the restraints and bags side, is incredible.
My first comment was just meant to highlight why* a certain car might pass all tests but have more injuries in real life. Another reason could be the nature of Tesla accidents, with people either driving silly or people sleeping in the car while it drives itself into a wall, e.g.