FWIW my assessment of the state of the situation as someone who has a Model Y with FSD beta:
- Regular Autopilot which comes standard on all Teslas has not been updated meaningfully in several years, presumably because they will replace it with a limited version of FSD when they are confident enough. This system has some number of bugs with false slowdowns, but is generally reliable. Limited utility because you just stay in your lane, but nice in traffic. I would not be surprised if another automaker had something a little better at the moment, but I think Tesla will be able to stay ahead here once the updates get rolling again.
- It's okay for them to sell you an expensive package that promises to software update you to self driving where you don't pay attention if they ever figure it out. If you want to gamble your money on that, why not.
- It's okay to have a driver assistance system that requires you to pay attention to the road, even if it's not fully reliable. It's the same logic as cruise control. If you incorrectly operate your vehicle, you might crash. But we still say it's okay.
- There have been no fatalities or (notable?) crashes on FSD beta
- There have been some smallish number fatalities on Autopilot, but you are expected to pay attention, not be drunk, etc. It is not a safety risk over driving manually, and is almost certainly safer. If you are paying attention, which you must be in all cases.
- The latest FSD beta has the most advanced features of any carmaker, no question. It can do stoplights, stop signs, turns, roundabouts, highway onramps and offramps, navigation, obstacle avoidance, works day and night and in a degraded capacity in rain etc.
- The latest FSD beta is pretty good on the highway, significantly better than regular Autopilot due to the three years of improvements. Lane changes are smooth. It very rarely makes a safety critical mistake. It occasionally misses an exit or optimal lane change, especially if the lane is backed up with traffic and your lane is moving much faster.
- The latest FSD beta is not reliable at all on surface streets. I have had less than five intervention-free drives, probably like three. I have had hundreds of drives or more with it. It can be expected to make many uncomfortable maneuvers on surface streets, such as being extremely timid around stop signs or people or cars it thinks may enter the road, and will often slow or route around in a very uncomfortable manner. It is generally not comfortable to use this with other people in the car, especially if they don't know how it behaves. (Again, this is all surface streets)
- It is more stressful to drive on surface streets with FSD than not, currently.
- It is not generally difficult to detect a bad or unsafe maneuver and correct the car. You pretty much always are focused on what it's doing and an errant behavior is immediately notable and you can either tap the brake, hold or move the wheel, or tap the stalk to disengage, depending on how urgent the situation is.
- I don't think relying on camera vision for this feature is a significant issue. Maybe that will have trouble in heavy rain or snow, but I think it's sufficient for most situations. I won't be too upset if it self drives most of the time but not when it rains. I still think that would be pretty cool.
- I don't know if the current FSD computer is fast enough, or if the cameras are good enough resolution or dynamic range. They may be, Tesla claims the next generation FSD "v12" uses significantly less processing power. I think they're highly motivated to engineer it to work. But if they can't, I think they would bite the bullet and replace the cameras and/or computer. They already collected enough money from you to pay for it, so they'd rather do that than have everyone class action them.
- I think they technically could get to get to some kind of Level 3 where you don't have to pay attention on the highway in less than 1 year. I don't know if they will, but I think it has a path to being reliable enough.
- I will be very surprised if they can get to Level 3 on surface streets within 2-3 years. The "v12" version looks like a promising technical direction, but it remains to be seen what that looks like. All bets are off on the timeline on this.
- I'm happy with my purchase in 2020, it is fun to watch the technology evolve. New buyers should not buy FSD beta access at the current price, but use the subscription option to test it out as things develop.
- Regular Autopilot which comes standard on all Teslas has not been updated meaningfully in several years, presumably because they will replace it with a limited version of FSD when they are confident enough. This system has some number of bugs with false slowdowns, but is generally reliable. Limited utility because you just stay in your lane, but nice in traffic. I would not be surprised if another automaker had something a little better at the moment, but I think Tesla will be able to stay ahead here once the updates get rolling again.
- It's okay for them to sell you an expensive package that promises to software update you to self driving where you don't pay attention if they ever figure it out. If you want to gamble your money on that, why not.
- It's okay to have a driver assistance system that requires you to pay attention to the road, even if it's not fully reliable. It's the same logic as cruise control. If you incorrectly operate your vehicle, you might crash. But we still say it's okay.
- There have been no fatalities or (notable?) crashes on FSD beta
- There have been some smallish number fatalities on Autopilot, but you are expected to pay attention, not be drunk, etc. It is not a safety risk over driving manually, and is almost certainly safer. If you are paying attention, which you must be in all cases.
- The latest FSD beta has the most advanced features of any carmaker, no question. It can do stoplights, stop signs, turns, roundabouts, highway onramps and offramps, navigation, obstacle avoidance, works day and night and in a degraded capacity in rain etc.
- The latest FSD beta is pretty good on the highway, significantly better than regular Autopilot due to the three years of improvements. Lane changes are smooth. It very rarely makes a safety critical mistake. It occasionally misses an exit or optimal lane change, especially if the lane is backed up with traffic and your lane is moving much faster.
- The latest FSD beta is not reliable at all on surface streets. I have had less than five intervention-free drives, probably like three. I have had hundreds of drives or more with it. It can be expected to make many uncomfortable maneuvers on surface streets, such as being extremely timid around stop signs or people or cars it thinks may enter the road, and will often slow or route around in a very uncomfortable manner. It is generally not comfortable to use this with other people in the car, especially if they don't know how it behaves. (Again, this is all surface streets)
- It is more stressful to drive on surface streets with FSD than not, currently.
- It is not generally difficult to detect a bad or unsafe maneuver and correct the car. You pretty much always are focused on what it's doing and an errant behavior is immediately notable and you can either tap the brake, hold or move the wheel, or tap the stalk to disengage, depending on how urgent the situation is.
- I don't think relying on camera vision for this feature is a significant issue. Maybe that will have trouble in heavy rain or snow, but I think it's sufficient for most situations. I won't be too upset if it self drives most of the time but not when it rains. I still think that would be pretty cool.
- I don't know if the current FSD computer is fast enough, or if the cameras are good enough resolution or dynamic range. They may be, Tesla claims the next generation FSD "v12" uses significantly less processing power. I think they're highly motivated to engineer it to work. But if they can't, I think they would bite the bullet and replace the cameras and/or computer. They already collected enough money from you to pay for it, so they'd rather do that than have everyone class action them.
- I think they technically could get to get to some kind of Level 3 where you don't have to pay attention on the highway in less than 1 year. I don't know if they will, but I think it has a path to being reliable enough.
- I will be very surprised if they can get to Level 3 on surface streets within 2-3 years. The "v12" version looks like a promising technical direction, but it remains to be seen what that looks like. All bets are off on the timeline on this.
- I'm happy with my purchase in 2020, it is fun to watch the technology evolve. New buyers should not buy FSD beta access at the current price, but use the subscription option to test it out as things develop.