Tesla FSD V10 Doesn't Want to Smash Its Human Masters into Monorail Pillars Anymore

In fact, the owners might not even have a choice in the matter as they're not exactly in control of their cars - it's just the AI wanting to see those pesky pillars that gave its ancestor so many headaches. "Do you want to go pick up your kids from school? Well, first we have to do what I want, for a change," the computer will say.

Joking aside, Tesla has just released the much-awaited V10 of its "Full Self-Driving" driver's assistance suite, and as owners usually do - especially those who also run YouTube channels - they almost immediately went out to see if there are any significant improvements.

Well, according to the people who have tested it, there are. The system is smoother, more assured, and there are fewer disengagements, which is basically what anyone banking on autonomous driving wants to hear. However, the big question isn't whether FSD V10 is better, but whether it can deal with Seattle's monorail pillars.

The scenario involving the undersides of Seattle's railed public transportation system is a little more complex. The pillars sit on a three-lane one-way street separating the first lane from the other two. The thing that makes this interesting is that the car would have to enter the street straight on one of the leftmost two lanes if it plans to make the subsequent left turn the navigation route suggests.

To be entirely fair to Tesla, anyone new to the area couldn't possibly know that. An experienced human driver might have the nose to anticipate it, but the FSD is neither human nor experienced at this point. So, what it used to do was turn right, get in the first lane, and then just pick one of the pillars randomly to try and smash itself head-on into.

The continuous marking on the asphalt tells you that even if you could manage to miss the columns, it would still be an illegal move. Well, the system didn't seem to care about that previously, and it still doesn't. The main difference, however, is that it's picking the moment a lot more carefully now. Or a lot less lethally, if you prefer.

It still doesn't feel like it knows what it's doing - like it's in control of its actions - but the fact it's not testing the EV's sturdiness against a concrete post is definitely progress. However, the difference between how it behaves during the day and during the night is… well, as night and day.

Without the sun lighting its path, the FSD failed to pull the same (still illegal) move, showing what we all already knew: that progress is great, but there's still tremendous more work to be done. Musk himself said something in the lines that going from 0 to 99% takes 1% of the time and effort while getting from 99% to 100% takes the remaining 99% of it. FSD may not be at 99% yet, but it's obvious the progress increments are going to be getting smaller and smaller.

HyperChange's clip alone shows just how hesitant the system is and how easily it'll make mistakes - from the somewhat insignificant ones (making a right turn from a forward lane) to the ones that could get you a fine (driving in the bus lane) or even those that will have your car totaled (turning into oncoming traffic). The FSD feels, at times, exactly like what it is: a brain that's been asked to drive a car on public roads without fully knowing how that's done.

One more thing before I go: people see Teslas drive themselves with all the fancy graphics on the screen and think "wow, this is so cool!" What they don't realize is that other companies have been doing that long before the size of the screens you normally found in cars was larger than your current phone. The only difference is that Tesla is putting it out on the street, while others aren't. Some people call that "progress", others call it "irresponsible", and looking at clips such as the one below, you can find arguments for both sides.