Tesla FSD 12.3.4 - Roundabouts and Yellow Lights

Our first update to FSD Beta 12.3.4 is here! I drive the brand new FSD Beta 12.34one short 16m drives in my 2021 Tesla Model S Long-Range Version. A much more natural driving experience. We tackle roundabouts and FSD even goes through a yellow light!

[@DerekDavis213]

FSD did very well. But some issues: 2:30 FSD is weaving left and right. *Confusion* . 12:07 Not perfect! If gate did not auto-raise, FSD would have hit it. No slowing down. Other FSD reviews show that seeing gates is an ongoing problem.

[@xiaoandmatt]

New sub here! Great video. Keep sharing.

[@pucaslife]

Thanks! I appreciate the sub! Let me know what you think of the channel and if there are any recommendations or questions are certain features on the car.

[@dennissteul5445]

Thanks for the video! I heard of problems with Teslas clipping curbs while making right hand turns. Have you experienced any of the curb problem?

[@pucaslife]

Not yet, and I do look for that when it's turning. I would imagine on the East cost where roads are tighter and smaller, there may be more issues.

[@dennissteul5445]

@@pucaslife Last week I took control during a FSD right turn from a two lane to two lane turn. It just seemed like it was cutting the corner too sharp. But maybe it wasn’t.

[@BigBen621]

*@DerekDavis213* _FSD did very well. But some issues:_ Your obsessive cataloging of every FSD shortcoming on virtually every YT video on FSD has no relevance, without a critical missing component-a control group; without a control group, these are only anecdotes. Since one of the major goals of FSD is achieving a level of safety greater than that of human drivers, the obvious control group is human drivers. And luckily, we have some good statistics for human drivers, from NHTSA. One safety statistic for human drivers is 13.4 fatalities per billion miles, equivalent to about 43,000 fatalities a year in the U.S. alone. The FSD Beta/Supervised test group has recently driven a total of one billion miles with FSD engaged, and so the number of fatalities in *this group* driving one billion miles would give a useful comparison with the human driver control group. And it turns out that there have been *no* confirmed fatalities among this group driving one billion miles; and only one claimed but highly unlikely fatality in Colorado in 2022. Of course, to be relevant, the human driver statistic would have to be adjusted for age of vehicles, driving conditions, driver demographics, etc.; but no adjustments conceivable to me would adjust the 13.4 fatalities per billion miles among average human drivers down to the zero confirmed fatalities of the FSD Beta/Supervised group driving one billion miles. This does not demonstrate that FSD by itself, without supervision, would have a similar record of safety; it almost certainly wouldn't. And it won't graduate to ADS status until its safety can be proven to meet standards of each state; but even if it *never* gets beyond ADAS status, it'll still provide convenience, relaxing travel, and astonishing safety to those who use it.