Discover and read the best of Twitter Threads about #FSDBeta

Most recents (17)

Die Frage nach der #FSDBeta für Europa könnte kaum grösser sein. Da sich hier auch die Gesetzgebung auf Neuland befinden, ist es extrem schwierig, eine klare Antwort auf diese Frage zu erhalten. Doch gibt es immer wieder implizite Hinweise – so auch heute.
Auf YT machte sich Steven Peeters die Mühe und nahm den Entwurf der DCAS-Regelung (Driver Control Assistance System) unter die Lupe. Dieser wird durch die @UNECE vorangetrieben & ist insbesondere für #FSDBeta in Europa relevant, da es sich dabei um ein Level 2(+) System handelt.
Good News: Es wird an div. Regelungen gearbeitet und man beweist darin genügend Weitblick, dass auch zukünftige Autonomiestufen (bspw. 3 und 4) direkt integriert sein können. D.h. das ab dem Zeitpunkt des Inkrafttretens auf einen Schlag, viele Hürden aufgehoben werden können.
Read 7 tweets
I'm really disappointed. Can anyone really read this and feel like we are where we thought we were? I am all in on #FSDBeta and supportive of the path we are on, but this wording sounds so "legal" that I am not sure what to expect going forward. tesla.com/support/recall…twitter.com/i/web/status/1…
Quote from the FAQ..
Will owners continue to receive FSD Beta?
Until the software version containing the fix is available, we have paused the rollout of FSD Beta to all who have opted-in but have not yet received a software version containing FSD Beta.
From the FAQ.. FSD Beta is an SAE Level 2 driver support feature that can provide steering and braking/acceleration support to the driver under certain operating limitations. With FSD Beta, as with all SAE Level 2 driver support features, the driver is responsible for operation… twitter.com/i/web/status/1…
Read 3 tweets
🧵Measuring #FSDBeta progress as a community after 1 year:
Many people like to use the metric “miles to disengagement” to measure or compare progress for autonomous vehicles. While this is a possible metric, is it the right one and the only one? I don’t think so…
1/11
Waymo doesn’t think miles to disengagement provides relevant insights into capabilities or comparison of performance to others:
Their thread goes on to state that the data may fluctuate depending on geography and other factors.
2/11
Cruise provided 4 different reasons for disengagements: medium.com/cruise/the-dis…
1 reason under ‘True AV limitations or errors’ stated about disengagements “are sometimes reportable to the DMV”. This raises a big question: Are companies only reporting SOME disengagements?
3/11
Read 11 tweets
#FSDBeta Community Tracker: Thank you
This has come a long way in a year thanks to the community participation & feedback. We have collected nearly 70k miles to track & measure progress. I was initially collecting data on my own in Oct 2021 & others wanted to join me.
1/4
🙏@energyvest @PaperBagInvest & @freshjiva for pushing me to expand the data collection to the community 🙏@xskidave for all the feedback & support since the beginning! 🙏@EVGuyCanada for building a great app to track data while we drive & for supporting the data integration.
2/4
🙏@CYBRLFT for all the feedback to improve the details collected including critical disengagements. There are honestly too many people to thank, but here are some mentions…sorry if I missed anyone as I don’t know all off the top of my head @GreyTesla @SantoroSystems
3/4
Read 4 tweets
On December 15th I left to drive from Los Angeles to near Philadelphia. Got back yesterday, after a total of 6,392 miles in the Model S Plaid with #FSDBeta.

Drove more than 99% autonomously, and I've some not great @Tesla Autopilot + FSD thoughts and experiences to share... 🧵
Let me start off by saying I recorded most of the trip in 4k@60fps from a GoPro inside of the car, and so I _should_ have footage of many of the things I’m complaining about here. Might be hard to find it though, amongst all the footage.
The second is that even with the issues I’m raising, it did reduce the personal cost of driving so many miles. It was a lifesaver. That said, it caused stress in other ways that wasn’t present years ago, some of which I think are legitimate safety concerns for me and others.
Read 21 tweets
I got #FSDBeta on Christmas Eve 2021. (Thanks @elonmusk & @Tesla) I got it originally on our 2018 Model 3 and then again when we traded Tess in for our new 2022 Model Y (Y Not). I am not on the new 10.69 version yet, (1/26)
so all these observations will be on overall progress from late 2021 to now and the current version I am on 2022.12.3.20 I am a super nerd, (2/26)
so I have purchased FSD twice to be able to watch the progress in real time and to hopefully provide useful feedback via the appropriate use of send snapshots to Tesla and my interventions.
The progress between releases is obvious. (3/26)
Read 26 tweets
After enjoying $TSLA #TESLA #FSDBeta, at last. My 2 impressions:👇
#1 - Awesome! Quality of the experience is exceptional. I was stunned and I was expecting much less. My car drove me to Mass today, and back, Alleluia! 😂. The quality of perception, handling complex and uncertain situations... Waow. Best A.I.-based product I know in the world.
#2 - One may think the gap to driverless (L4-L5) is tiny. It is kind of true. It is tiny, but of abysmal depth. I continue to firmly believe the technology will never scale-out effortlessly over a fleet, for driverless services. Let me explain three reasons why:👇
Read 8 tweets
Update in question is “Upgraded modeling of lane geometry from dense rasters (‘bag of points’) to an autoregressive decoder that directly predicts and connects ‘vector space’ lanes point by point using a transformer neural network.”

So, $tsla #FSDBeta creates a 3D model of…
…reality.

As I interpret the notes, previous versions, the 3D geometry is based directly on the camera inputs, which generates excessive qty’s of 3D points.

This is now fed into a new neural net that can distill them into a more sparse-yet-adequate 3D model.

Importantly…
The new network is regressive, meaning that it can base predictions on past and current “knowledge.” Essentially, it has short-term memory and can make determinations based on info from a timeframe instead of solely from an instant moment.
Read 4 tweets
For about the next 10 days, this is my playground. 😄🎆🚗🛣️

Will post interesting bits of the trip as a thread. Image
Starting the trip off by making my first visit to the Garner, NC @Tesla supercharger. Follow @superchargefeed for crowdsourced updates on the status of superchargers coming online. See a map at supercharge.info. Image
Not the biggest fan of the Glenn Allen supercharger but the upside is that I can have this nice healthy lunch at @panera and update Tableau since it had hiccupped overnight. Thanks for the WiFi and food Panera! Image
Read 46 tweets
@elonmusk @Tesla Tired of sending my feedback to the Tesla FSD Beta e-mail address and sending autopilot videos only for no reply and to see no change in how car handles things. Will just post them here and perhaps someone might see it. #fsdbeta #fsd
#FSDbeta is not beta, or even alpha, and is barely an engineering prototype if we go by standards of hi-tech companies I've worked with/for. First of all, it suffers from a severe regression problem when compared to production software.
But first, a word about where I drive my car. I live in the midwest, a rural area with many state highways that have 55 mph speed limits, T-intersections, and many small villages where speed limits will rapidly change between them. It's clear @Tesla doesn't invest much time here.
Read 21 tweets
Tesla Earning Call Day!
$TSLA $TSLAQ Image
How late does $TSLA Q4 2021 Earnings Call get started?
Don't expect fireworks from a DOJ NEUTERED Musk. There's a reason Musk didn't post a video of him driving the latest Cybertruck prototype--that reason is NIKOLA/Trevor Milton federal indictment.
I mean, Musk'll still be indicted, he's just not adding to the list.
$TSLA $NKLA Image
Read 18 tweets
The #Tesla #FSDBeta thread, for documenting the abilities of #FSD and advancements/regressions with each update on the way to full #autonomy!

Contains detailed timestamps for interesting events in each of many instructive videos.

Begins at #FSDBeta10 👇
Read 3 tweets
$TSLA $TSLAQ likes to boast quarterly about how incredibly safe #Autopilot is in their Vehicle Safety Report. The numbers they claim are impressive, but let's look past the grand marketing claims typical of Elon Musk and dig into the numbers. 🧵
First of all, let me just state upfront that I have no doubt that driver assistance features like the dangerously misnamed "Autopilot" do improve safety. Most of those safety features are now widespread, e.g. on your average Toyota Corolla. (BTW, $TM Toyota marketing sucks)
Now where do $TSLA's numbers come from? Their own numbers are self reported and cannot be independently verified. The NHTSA number they quote, "an automobile crash every 484,000 miles", of comes from Overview of Motor Vehicle Crashes (page 3): crashstats.nhtsa.dot.gov/Api/Public/Pub…
Read 20 tweets
I will be live tweeting @Tesla AI Day Here. Personally looking at the data flow, latency, low power and performance of #Dojo and how it compares to @Google #TPU version 4 and other HPC systems and how it directly relate to #AutonomousVehicles #FSDBeta
Well as usual, Tesla is fashionably late and we are almost 18 mins past the start time. And the beat goes on...
#AIDay starts with a demo of FSD with the driver griping the steering wheel with his left hand in what looks to be the streets of California. #DojovsTPU #Dojo @Tesla Image
Read 22 tweets
A follow up 🧵 on Tesla's FSD "rewrite" to explore the Bird's Eye View Networks.

"There are still a few of neural nets that need to be upgraded to video training & inference. And really, as we transition to each net to video, the performance become exceptional." - Elon Musk
We can gather from this statement, not all networks are running the BEV mode "rewrite" even for #FSDBeta users. But why is a BEV so much more "performant" compared to a single camera? Let's see what @karpathy has to say about this video:
Per camera: "We developed this occupancy tracker that stitches up these image level predictions. The stitching is highly error prone across cameras and time."

BEV: "We don't have the occupancy tracker in C++ land, its in a network."
Read 15 tweets
A #FSDBeta thread: Executive Summary - 1) I have been an @Tesla FSD skeptic for many years, largely unimpressed by Autopilot/FSD's claims versus its abilities. This beta appears to be a marked improvement and far more inline with those claims than previous iterations. 2) Safety..
is paramount and should not be overlooked. Within the context of Tesla's approach to FSD objectives it appears they HAVE taken steps to ensure owner safety of those in the beta program. 3) Most of what I've seen from $tslaq so far around this matter appears to be
focusing on the 5% POTENTIAL risk instead of the 95% upside. Let me explain. Before FSD Beta, critics said it's misleading, dangerous, and deadly. This perspective puts the blame 100% on Tesla. It completely ignores any owner responsibility for any accidents AND the driver's...
Read 13 tweets
最近给 Model 3 升级 2019.36 版本后,小试了一下 #Autopilot,没想到第一次测试就来了个完美的路口左转。

当然,目前 Tesla 尚未上线自主左转的能力,所以前导车辆仍是必要条件。虽然以前偶尔也能在小路上左转成功,但在有四车道的主干道路口肯定没戏。

继续见证自动驾驶的演进~

@vincent13031925
看了最近大量涌现的 Tesla #FSDbeta 视频,一个直观感受是 FSD 在城市道路的驾驶风格朝着人类老司机迈进一大步。

Autopilot 基本只认道路线,遇到挡道的东西,不是减速刹停就是直接撂挑子。要是道路线不清晰或紊乱,简直就六神无主了。FSD 已经具备相当的机动意识,为避开单车,压双实线也是常事儿。
目前看起来 #FSDbeta 处理右转弯是没有技术障碍的,只是有点像新手司机,转弯前过于踌躇,一遇到直行的来车就龟缩不前。

左转弯还是目前出错比较集中的场景,遇到稍微复杂一点的路口,一不小心就在优雅的左转后入错车道,华丽的逆行。

至于狭窄的小巷、道路施工啥的,对 FSD 来说简直就不是个事儿。
Read 4 tweets

Related hashtags

Did Thread Reader help you today?

Support us! We are indie developers!


This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!