Tesla FSD Beta Update — How Much Has It Improved In 2 Weeks?

Sign up for daily news updates from CleanTechnica on email. Or follow us on Google News!

Like the ~1,000 others who got a Tesla Safety Score of 100 and got FSD Beta two weeks ago, I just got my first FSD Beta software update in the middle of the night last night. Here’s what the release notes say:

Naturally, I was excited to take my Model 3 out on the road and see how much the system has improved (or not) from those two weeks of work. Also, as you can see, there is a new option allowing you to choose how “Chill” or “Assertive” your car is when handling rolling stops, exiting passing lanes, making lane changes, following other cars, and responding to yellow lights (with a Goldilocks “Average” version that many may feel is “just right”). I wanted to compare those different profile options, so I took the car to a quiet, safe area and went through the same route for each profile. I’ll come back to that comparison in a separate article, though, and will update it throughout the week as I try out those different modes more and try to distinguish the differences between them.

First, let’s run through a few challenges FSD Beta had in the past and see whether it still has them now.

How Much Has Tesla FSD Beta Improved?

In the past two weeks, I have to say that I have been surprised at how poorly FSD Beta has executed (or not) basic operations on numerous occasions. Yes, I’m also excited and super impressed when it can drive somewhere on its own, making various turns, responding to lights and signs, navigating around parked cars or pedestrians, etc. But there have just been several super easy things that it hasn’t been able to do well, mostly involving intersections, but not only. I couldn’t retest all of these, or even most of them, but I picked two that stood out and were easy to test in order to quickly revisit them with the update. There’s also one general issue that happens in various places that confuses me, and I got to see how much it has improved. In total, I’ll go through the following three items in this article:

  1. Stopping at a stop sign on a quiet residential road.
  2. Handling a cul-de-sac with a little planter/median in the middle.
  3. Getting in the correct lane to follow the navigation’s guidance and make right and left turns.

Aside from all of the above, my biggest problem has been how quickly, slowly, and jerkily the car sometimes (often?) goes into intersections in which it is making a turn. Most dramatically, the system sometimes abandons its self-driving efforts in the middle of the turn, leaving me to take over. (Pro tip: If you get FSD Beta, do not use it going through intersections with cars closely following you or near you.)

Stopping at a stop sign on a quiet residential road.

FSD Beta has handled numerous stop signs in my two weeks with it. The main problem is that it’s often too slow while carefully stopping at the line, and, more so, while leaving the stop. If any other cars are around, it’s way too awkward and annoying for me to use it, so I just disengage it. Oddly, there’s one stop sign — on a street with some other stop signs before it and some after it — that the car didn’t see until too late. It stormed nearly halfway into the intersection before stopping suddenly. (To repeat: there was no real risk here because I was testing it with no other cars driving around me.) My friend and I decided to circle around and try it again. It was slightly better, but it still flew into the stop before braking hard.

With the update, I didn’t notice any such problems at this location. The car seemed to stop naturally and appropriately. I will retest it again soon and report back if the problem resurfaces.

In general, the car seems a little bit better at stopping appropriately at stop signs and then leaving them smoothly — not too quickly and not too slowly. But that’s not always the case and I would say that the car is very far from perfect on this. At this stage, I don’t see it as a good idea to use FSD Beta at stop signs with any other cars around. But maybe soon? Or in 6 months? Certainly within one year, right? I don’t know. This is what I’m most interested in right now — seeing how quickly and how well the system improves.

Handling a cul-de-sac with a little planter/median in the middle.

In that same residential area, my friend and I went down a street that has a little planter/median in the middle of a cul-de-sac. I was surprised the first time I was there that the car sped into the cul-de-sac without seeming to pay much attention to the planter/median and then had so much trouble turning around it that it had to disengage and quickly beep at me to take over. Of course, I was very ready to take over anyway — was about to do so — but I did find it interesting and surprising that this was one of a few occasions where the car disengaged FSD Beta/Autopilot by itself.

So, how did it do after the update?

Not well. Instead of speeding into the loop too fast and disengaging when it couldn’t make the turn well, it decided on a couple of occasions to go left around the planter/median (still far too fast). There was also a case similar to the first one where it was just flying into the right turn around the median too quickly and I decided to disengage a little ahead of time in order to have a less dramatic experience. There was no need to see how it would handle the rest of the turn, because it was certainly not going to be smooth and it would probably disengage itself again.

I will certainly check this spot again after future updates.

Getting in the correct lane to follow the navigation’s guidance and make right and left turns.

The most bewildering mistake the car routinely makes is not changing lanes based on the navigation route when a turn is coming up. The car routes itself once you put in a destination, and I recall Navigate on Autopilot being quite good at getting in the correct lane ahead of time, but FSD Beta/Autopilot on city streets keeps not changing lanes and then not being able to get in the turn lane in time. And, well, let’s just say that it can sometimes be super absurd with how it handles such situations after it has missed the lane change and gotten to the intersection.

It really seems simple — you have a left turn coming up, get in the left lane; or you have a right turn coming up, get in the right lane. I find it rather astounding that it hasn’t been able to do this most of the time. And mind you — I’m driving on basically empty roads in a Florida suburb with great lane markings and infrastructure! There is no traffic blocking me from getting in the correct lane ahead of time, or even at the last moment if they car tried to do that. Odd.

This seems like such a simple thing to improve that I’m optimistic that it will be resolved well at some point soon-ish. But we’ll see.

Wait, Three More Things!

After one of those lane-change misses, I was at a light that turned green and had re-engaged Autopilot after getting in the correct lane to turn left. I was sitting there in the lane with the green light in front of me (there’s a dedicated left-turn light) and the car didn’t budge. The car has made this turn several times. For some reason, though, it wouldn’t move. I’m not sure what was going on, but I simply had to disengage again and drive through it myself.

Before this update, one of the little problems I’d noticed with FSD Beta is that it would very briefly brake for shadows in some places. Or, at least, I think that’s what it was braking for (I’m 95–99% sure). Unfortunately, it’s one of the ~10 grey days of the year here in Florida, so there’s no testing the car’s perception of shadows today. Also, just as a note, this is very brief braking I’m talking about — enough to somewhat startle and annoy passengers, but not enough to create risk of accident from a car behind running into you (unless it’s a millisecond or so behind you).

One more issue I’ve noticed in the past couple of weeks is snaking/ping-ponging. This was quite a letdown because I loved the way the previous version of Autopilot locked in place for lane keeping. It felt like it was on the smoothest rails in the world or something. Apparently, though, changing to vision only (removing the use of radar) led to some changes in how Autopilot keeps itself in the lane. I haven’t noticed any snaking/ping-ponging today, but I also haven’t driven in the places where it seemed to be the most obvious. Also, the issue might have been a little worse the first few days I had FSD Beta and then got better over time. Or I imagined that.

In Summary, How Good Is FSD Beta Right Now?

If you want to lower your stress level while driving, FSD Beta is not for you. I’ll admit that I thought it would be much better at some basic things by the time I got it, so I’m still suffering a bit from my expectations being too high. This looked like a big update and I thought some of the system’s most glaring weaknesses would see stronger improvement than I’ve noticed so far. That said, it’s only been two weeks.

Despite apparently expecting too much, I am uplifted to notice improvements in just this short window of time. The system is far from a level necessary for robotaxis, imho. However, it’s also far better than any other driver-assist system on the road. It can do things that I would not have dreamt possible in my lifetime a decade ago. Can it learn to turn, go through intersections, not get confused in cul-de-sacs, and drive smoothly in the next few years? I hope so! But I have no real idea. We’ll see.

Given that my expectations were higher than they should have been for the state of the technology today, I’m a bit tempered in my forecasts and hopes at the moment. The main thing I want to see in the next few weeks and months is how well Tesla resolves twitchy behavior at intersections, improves the speed of the car going into intersections, improves the pace with which a car turns at different intersections without having to jump between quick acceleration and quick braking, and resolves problems following the navigation system when approaching a turn. I’m sure Tesla will continuously improve the system. But will it become robotaxi ready?


Have a tip for CleanTechnica? Want to advertise? Want to suggest a guest for our CleanTech Talk podcast? Contact us here.

Latest CleanTechnica TV Video


Advertisement
 
CleanTechnica uses affiliate links. See our policy here.

Zachary Shahan

Zach is tryin' to help society help itself one word at a time. He spends most of his time here on CleanTechnica as its director, chief editor, and CEO. Zach is recognized globally as an electric vehicle, solar energy, and energy storage expert. He has presented about cleantech at conferences in India, the UAE, Ukraine, Poland, Germany, the Netherlands, the USA, Canada, and Curaçao. Zach has long-term investments in Tesla [TSLA], NIO [NIO], Xpeng [XPEV], Ford [F], ChargePoint [CHPT], Amazon [AMZN], Piedmont Lithium [PLL], Lithium Americas [LAC], Albemarle Corporation [ALB], Nouveau Monde Graphite [NMGRF], Talon Metals [TLOFF], Arclight Clean Transition Corp [ACTC], and Starbucks [SBUX]. But he does not offer (explicitly or implicitly) investment advice of any sort.

Zachary Shahan has 7317 posts and counting. See all posts by Zachary Shahan