Search results

Sure but she was granted Veoneer shares worth $900,000 as a "Director". If someone with $900,000 of Tesla stock was put on the NHTSA people would be crying foul just as well (just different people).

Surprised she nuked her entire twitter account. I guess it's harder to be called objectively biased against Tesla when you bury the evidence. https://twitter.com/missy_cummings/with_replies

Slight differences in how the roads appear visually and the background environment can affect neural network perception very easily. Look how easily a neural network can get the wrong answer even though the image looks identical to a human: How to trick a neural network into thinking a panda is...

Wanted to take FSD Beta from CA to Texas to see the Starship under construction, but recently chose to abort those plans

From what I understand watching AI day, the inference stack does not use a 3D vector space point cloud like you describe. There was a visualization with a 3D point cloud but that's an output of the neural network rather than how it functions. The Transformer combines the 2D feature maps from...

As a child, being dragged around on road trips and joyrides and trips to the store was just the worst. One time my dad even almost drove us off a cliff. Better to keep children locked up indoors and on the internet where they belong, regardless of whether FSD is enabled! edit: and I did not...

The paper explains optical flow and much more about all of your questions. The trick to reading these white papers is just skip over the equations and read the overviews, look at pictures and impressive results: https://arxiv.org/pdf/2105.07014.pdf

Practically there are also constraints on AI training time. If it takes 2 weeks to train a vision neural network and 2 weeks to train a radar neural network, you can drop radar and instead train 2 vision neural networks in 2 weeks with slightly different parameters and use the better one. Or...

Hopefully AI day will include more details related to what's mentioned in any of these tweets There are a lot of interesting ongoing AI advancements so hearing them talk about specifically what they're doing would be great!

I appreciate that Tesla has avoided added visualization interpolation since it would clean up the jiggle and wiggle but add latency and inaccuracy. Actually it's interesting that the most recent neural networks look at the last 15 frames since they must have developed a way to prevent it adding...

Same for me: fixed in .15, service appt cancelled!

Thanks for the updates, I've scheduled mobile service with this info. Will see how it goes!

I've tested many options and confirmed the only way to prevent this issue is to keep Sentry Mode enabled (uncheck Exclude Home). Anything that prevents the car from going to sleep will actually help reduce network traffic, battery usage, and the battery disconnect/reconnect sound. Nothing to do...

I also have this issue but without teslamate, just from watching network traffic and battery and hearing it wake up constantly. Other thread is here No longer sleeping after 2021.4.11

I've had the same issue since 2021.4.3. Daily network traffic went up ~100x as well. Have tried soft reset, hard reset, turning off standby mode, turning off scheduled charging, logging out of spotify.

The ultimate solution would combine the visualizations by drawing map data onto the car visualization and drawing nearby objects onto the map. I could see the car visualization looking similar to a video game with a glowing path projected along nearby roads for navigation. Done well enough you...

Not really normal. Right lane and left lane should be similar. OP do you see this issue with left lane ever?

I think teleoperation and remote assistance give Waymo a crutch that we then perceive as advancement. But if you disabled teleoperation and remote assistance for a Waymo van, or if you added teleoperation and remote assistance to FSD Beta, how would they truly compare? To me teleoperation and...

The craziest thing about fsd beta is that Elon can release it whenever he chooses. AP can already be engaged in many unsafe scenarios on city streets and there's currently no strict regulation in the US to stop him. Whether he should release it is each person's own opinion but he gets the final...

"merging code" should not be taken as "merging systems". When using source control and changes are brought in from a separate line of development they are "merged" in. Instead it should be read as "Lot of new code was added, deleted, or changed." edit: this is my assumption because Elon tends...

I'm not sure about the new version causing new problems. DirtyTesla's roundabout is completely missing from OpenStreetMap and appears to be brand new (not even visible yet in Google or Bing satellite images). This appears to be another case of bad map data overriding vision. Interestingly, the...

On OpenStreetMap the roundabouts that Brandon has trouble with look like: Possibly it's having trouble because it thinks it needs to follow the circular path when really no human driver would follow this tiny circle that closely. I edited the next closest roundabout to have paths that are...

In Tesla's patent for training FSD they talk about automatically training the neural net based on sensor data from vehicles, then deploying that updated model, and testing if it increases the accuracy. GENERATING GROUND TRUTH FOR MACHINE LEARNING FROM TIME SERIES ELEMENTS - Tesla, Inc. "In...

Found a side camera video in a very similar situation from greentheonly. He got it from a salvage wreck so the video is from a crash caused by humans driving, to be clear nothing to do with autopilot. The oncoming car is clearly visible beyond 110m. Note this video has been compressed once by...

I just confirmed the same thing with the repeater cameras. Unfortunately we can't check the side cameras until they are recorded to TeslaCam. Will be very trivial to test then... Looking at a recent random repeater cam video, a car at 150m (measured on google maps between intersections)...

I've been testing Autopark and Summon every few trash days (maybe once a month) and with 2020.44.10.1 Autopark finally works for me! Since my garage has a steep incline right before the door it would always get stuck at "Grade too steep". Now it's able to power through the incline and after...

Yeah... I don't think they are going to go from non-HD maps to HD maps by adding a single fixed size bitfield. Any additional HD data of significance should require an order of magnitude more data (5GB to 100 GB would be a better indicator for example).

Don't think an increase from 4.5GB to 5GB means anything besides half a year's worth of new map data. Increase of 500mb is nothing...

There are many intersections with blind turns where you can't see fast oncoming traffic until you are in the road. In this situation it doesn't matter what resolution or framerate the camera is at because even a human can't see oncoming traffic until the vehicle moves forward into a lane. Humans...

Another part from that link describes how it runs in shadow mode: "FIG. 4 is a flow diagram illustrating an embodiment of a process for training and applying a machine learning model for autonomous driving. In some embodiments, the process of FIG. 4 is utilized to collect and retain sensor and...

Tesla describes the answer to OP's question here (bolding for emphasis is mine, except for the referenced numbers): GENERATING GROUND TRUTH FOR MACHINE LEARNING FROM TIME SERIES ELEMENTS - Tesla, Inc. "In various embodiments, the result of vehicle controls such as a change in speed, application...

We can read how it works in Tesla's own words (bolding mine except for the numbers) GENERATING GROUND TRUTH FOR MACHINE LEARNING FROM TIME SERIES ELEMENTS - Tesla, Inc. "At 405, deep learning analysis of the sensor data is initiated. In some embodiments, the deep learning analysis is performed...

Some more interesting quotes from GENERATING GROUND TRUTH FOR MACHINE LEARNING FROM TIME SERIES ELEMENTS - Tesla, Inc. which I think describes both pre-rewrite (209) and post-rewrite (409) autopilot system. "For example, the model may be trained to identify road lane lines, obstacles...

How about we look at Tesla's related patent? https://www.freepatentsonline.com/y2020/0250473.html https://www.freepatentsonline.com/20200250473.pdf "In some embodiments, the various outputs of deep learning are used to construct a three-dimensional representation of the vehicle's environment...

I appreciate the video and information but there's no reason for insults. Just trying to have a conversation and learn more. You say I'm wrong about Mobileye but posted a video from Cruise. They only briefly mention labelling multiple frames at 16:45 and it doesn't seem that relevant. If you can...

The video is great and I love the level of detail. However I think if anything it actually proves that Tesla is taking a different approach. Direct link to their pdf: https://newsroom.intel.com/wp-content/uploads/sites/11/2019/11/Mobileye-Investor-sensing-status-presentation.pdf Mobileye is...

Do you have any diagrams for how Mobileye compares to this? All I've seen is lists of features supported by EyeQ4, not how it's built. Just because two systems result in similar features doesn't mean they're functionally identical. Will watch the sensing video you linked in a bit.

Yet it also says this, hmm...

I don't think that video proves it without a doubt. For the cases I see in that video they could detect crosswalks as features of the intersection based on the brightly lit red hand / white walking person crosswalk indicators. Just like a human ;)

Karpathy explained in a talk that Autopilot is actually the combined result of over 100 separate tasks running in real-time and each can be tuned separately or left alone by Tesla in a software update. Even the FSD Beta people could still have a Smart summon that is using the old perception...

I think you mentioned in the Waymo thread how self driving cars will eliminate certain classes of accidents like rollovers. I bet driving straight into a roundabout is one of those classes that humans have a higher percentage than most SDC or FSD already. Personally there is a two lane...

Haha just a bit of bias in that article. "Seemingly mundane maneuvers for human drivers, like circling a roundabout, have drawn intense praise from the Tesla owners and friends they've allowed to ride with them." I'm not sure they've seen the gifs of humans driving cars straight into and over...

Thanks for posting this, very interesting and relevant! On Tesla Dojo George says it's too expensive to train neural networks, your only options are Google or NVIDIA and Google doesn't allow training autonomous driving. He doesn't like the monopoly and wishes Tesla would sell their own chips to...

Tesla has been intentionally encrypting and obfuscating their software in increasing degrees so I think it would require an insider to really compare their neural network approach to Mobileye or any other competitor. I don't think that they're still labeling all their secrets and methods in...

To put it a different way, a successful global release of FSD could force a change in strategy from other competitors. Honestly reading the sudden rush of all the other self driving car articles posted recently alongside the FSD beta announcement, I feel like it already has.

What seems key to me is that Tesla is the only major car company committed to equipping every single vehicle it manufactures with FSD capability. GM, Honda, Mercedes, etc. may have the capability to manufacture 400k+ FSD capable cars per year but they are not currently doing so and will take...

Had this same thought as well especially when they describe the rear ends during a left hand turn. "The remaining rear end struck collision involved a deceleration to a near stop by the Waymo Driver while making a left turn in an intersection with a following vehicle that was traveling at a...

Very glad that verygreen snooped the beginnings of a driver monitoring system in Tesla code since it means they're likely working on a better solution to driver monitoring and fixing these ratings.

The Waymo article and report mention that "nearly all" collisions were caused by the other driver. I can't find where they talk about which collisions that was not the case for. https://storage.googleapis.com/sdc-prod/v1/safety-report/Waymo-Public-Road-Safety-Performance-Data.pdf "Nearly all...