I'll preface my frustration by saying after more than two years of ownership, my model 3 has been my favorite car out of probably 12 fun cars (mostly manuals) I've had as daily drivers. Further, despite his faults, Elon is on balance my favorite person in the public eye.
Tesla's top two objectives for its automated systems should be:
- Don't crash
- Don't impede traffic or otherwise impose more dangerous traffic conditions on everyone else.
Tesla is having some difficulty with #1, but I understand why it takes years of refinement to handle all situations. However, the failure to respect #2 makes me wonder if I'll ever let Tesla's systems do their thing in traffic without regular intervention.
As someone who (i) has had a 100% score with hundreds of miles since before 10pm PDT on October 8 and (ii) has a HW3 model 3 (yes, I know it's worse for plaid owners who never had a chance), it's incredibly frustrating to have no guidance on when Tesla will distribute the beta they should have provided last weekend. This indefinite period of continued waiting has put a lot of us in the awkward position of continuing to maintain a 100% score in the hopes Tesla will soon provide access to the beta or giving up after weeks of more dangerous (and definitely less fun) driving. Aside from the ill-considered motivation to roll through stop signs and run red lights (which I don't do even if it generates a penalty), worrying about the safety score has worsened my driving in almost every respect:
- Anytime traffic is moderately heavy, I'm currently stuck using autopilot to handle merges, and it frankly sucks at getting in the right position to merge safely in heavy traffic without cutting someone off.
- Anytime traffic is moderately heavy, I am often forced to let autopilot handle following (otherwise I'm far enough back that people constantly cut through my lane. The inconsistent following distance regardless of setting not only makes me look like an imprecise idiot, but also forces other cars to route around the hazards the car is creating. It's amazing we have to deal with this, given there were past iterations where following distance was fairly reliable. In a few cases, I just give up and take over, because ethically there's only so much you can do to fellow road users.
- A fair number of highway transitions in my area are 1 lane (meaning you inevitable impede traffic and generate angst if you slow down), and the radiuses are tight enough to exceed 0.4 g's if you go the speed limit. So the unfortunate choices if you want to avoid "aggressive turning" are (i) go on autopilot in the hope the car will maintain a reasonable distance to the car in front, and accept the fact autopilot is less than graceful in the turn; or (ii) do manual control and impede other traffic.
- Any time a stopping situation arises, I drag out the stopping as far as possible and end up carrying more speed than I like as I approach stopped traffic, when the safest thing to do would be to slow down more earlier.
- Exiting parking lots I have to wait until I've almost completed the turn before I can add speed. This means that even when I have chosen a large gap, the following car is closer than it should be by the time I'm able to accelerate at a reasonable pace. This is stupid because it reduces the margin for error, rather than preserving it.
- Sometimes the closest hole in traffic is quite tight and not the safest one to merge into. Maybe there's a large available space some distance in front that would require adding speed and then braking once you get there. Not hard braking, but more than 0.3 g's. It would be much safer because it provides much larger space buffers around the car, but you can't get there because the safety score will ding you if you do it manually, and autopilot will take a ridiculous amount of time to accelerate unless you're willing to goose the accelerator and get dinged.
- As all of us know, we are an absolute nuisance on the road in significant traffic when trying to drive for the safety score, and the amount of frustrated and aggressive driving we generate in our vicinity is vastly more than most of us generate with our normal driving behaviors.
- In lighter traffic I would normally stay as far away from other cars as possible while maintaining speed, but instead there's this perverse motivation to close the gap on other cars to cancel out those times you have to take over from autopilot too close to other cars.
- I avoid overriding autopilot when I should, because it will usually cause unsafe following, and can also penalize for turning or braking once autopilot has placed the car in a tricky situation. Typically when you need to override autopilot, there are conditions present that will penalize you for driving manually. In light of the stated purpose of the safety score, to identify drivers who can safely supervise FSD beta, this is the most counter-productive of the safety score attributes.
After reflecting on the myriad frustrations since Tesla released the safety score, I've realized that a fair number are related to the fact Tesla has all but forced autopilot use on those that want a perfect score, and at least one of the relevant factors (following distance) is much more inconsistent than it was when I bought the car a couple years ago. And most of the frustration boils down to Tesla's failure to diligently focus on objective #2 I identified at the top of this post.
The combination of Tesla's failure to seriously try to address objective #2 (don't make things worse for others) in its automated systems, coupled with the fact the safety score actively works against objective number #2 in many cases, has me wondering if I will ever feel comfortable letting a Tesla system handle everything, even if #1 is solved. So I've done all this painful safety score compliance to get early access to a system I'll probably feel ethically bound to override even when it's not trying to kill me.
The fact that Tesla dangled a carrot in front of us to drive in a way that generates
more anger against Tesla among the general public, and then failed to deliver the carrot to a bunch of us, is staggeringly stupid for a company that generally does cool things and does not want to invest in PR.
