News
Tesla 3D labeling is the next big leap for Autopilot
Tesla’s 3D labeling efforts are integral to the development of its Full Self-Driving suite. Using over 2.2 billion miles of real-world driving data from its electric vehicle fleet, the electric car maker has a treasure trove of information about how human drivers behave.
Elon Musk recently confirmed that Tesla is finishing work on Autopilot core foundation code and 3D labeling, and once these are done, users can expect the electric carmaker to roll out more functionalities in a potentially more efficient manner. More advanced features such as Reverse Summon will also be rolled out.
Tesla 3D Labeling: The Next Big Thing
The Tesla CEO has tagged 3D labeling as the next big thing for the company’s efforts to achieve full self-driving. “In terms of labeling, labeling with video in all eight cameras simultaneously. This is a really, I mean in terms of labeling efficiency, arguably like a three order of magnitude improvement in labeling efficiency where Tesla vehicles use all of its eight cameras simultaneously, and that the company has improved significantly in terms of labeling efficiency,” Musk said during the Q4 2019 earnings call.
During Autonomy Day last year, Tesla’s AI head Andrej Karpathy gave the electric vehicle community an idea of how labeling is done. He said annotating data is a very expensive process that initially involved people processing data, but Tesla has also been using information from its fleet to automate the process of labeling using different mechanisms.
For example, in predicting cut-ins, Tesla taps into its fleet for data on such incidents. This information is then automatically annotated and used to train the neural network, which in turn learns from recognizable patterns. This information is then spun until the neural network is trained enough. Improvements in the neural network can then be rolled out as an update for Autopilot.
The same is true according to Karpathy when it comes to object detection. Tesla sources data from its fleet to learn more about different objects and anomalies on the road. With automated 3D labeling, the neural network can more efficiently process the information and learn even about the rarest things one can encounter on the road.
Karpathy and Musk explained how annotations from its fleet help with path prediction. Using trajectories collected from the real-world, the neural network can improve its driving behavior, say while approaching a corner that it doesn’t actively see. This smarter neural network is perfectly demonstrated by an older Model X with early-gen Autopilot negotiating a muddy rural backroad recently, after a storm in the United Kingdom.
All of these things form part of the equation to achieve Full Self-Driving capabilities. Likely through 3D labeling improvements in the past year or so, Tesla has immensely improved driving visualizations in vehicles equipped with Hardware 3, which now identify traffic lights, garbage cans, and detailed road markings, among others. Thus, Elon Musk’s explanation about rewriting the Autopilot foundational code and 3D labeling could be a way of emphasizing that Tesla owners’ investment in the company’s Full Self-Driving suite would be proven worth it and more soon.
Tesla’s FSD computer and autonomy software will transform how humans travel. The company’s vehicles will be smart enough to drive like humans and eventually make the roads a few times safer for everyone. This may also pave the way for Robotaxis and help achieve Musk’s vision of Teslas earning for their owners while they are busy with work or even while relaxing at home. Tesla Robotaxis would be an attractive form of transportation as they will be more cost-efficient compared to driving personal cars, as predicted by ARK Invest.
Autonomy As Key To Profitability
Autonomy will spell profits for Tesla, as Elon Musk explained during the company’s Q4 2019 earnings call. In order to achieve sustained profitability, Tesla needs to produce high volume units with high margins. Musk appears to consider autonomy as key to Tesla’s high margins as well.
“As we’re close to Full Self-Driving, that is just going to become more and more compelling. So that’s for our financial standpoint, that’s the real mind-blowing situation is high-volume, high-margin because of autonomy,” Musk said.
With FSD capabilities, Tesla adds more value proposition that can help sway even more customers to purchase its electric vehicles from the Model 3, Model Y, Model S, Model X, or the Cybertruck. Depending on regulations in specific regions, Tesla can tap into most of its earnings potential, which bodes well since the company has current plans to expand its presence worldwide with Gigafactories in multiple regions.
Tesla’s path to autonomy is only one of the aspects that make it the leader in the electric vehicle industry. Add to that its advancements on car connectivity and battery technology and one will complete the equation why legacy carmakers with the deepest of pockets can only watch in amazement as a relatively young electric car maker dominates the emerging EV industry.
A plant manager at a Tesla factory just tipped off the fact that the company will begin production of an affordable model in the coming weeks, all but confirming that a new car will be unveiled tomorrow.
Tesla has been teasing some kind of product unveiling for October 7 on its social media accounts. It has now dropped two separate indications that a new product is coming on its X account.
Fans have been anticipating two things: either the company’s planned affordable model, which has been codenamed “E41,” or the Roadster, a long-awaited vehicle that Tesla has kept under wraps for much longer than it would likely care to admit.
Tesla all but confirms that affordable Model Y is coming Tuesday
André Thierig, Tesla’s plant manager at the German production plant Gigafactory Berlin, tipped off what is likely coming tomorrow at the product unveiling as he revealed during an internal event today that a light version of the Model Y will begin series production and deliveries “in a few weeks.”
Thierig’s revealing of plans was reported by Handelsblatt, a German media outlet.
The description of a “light version of the Model Y” aligns with what CEO Elon Musk said earlier this year, as well as what we have seen on public roads, both covered and uncovered.
Last week, we finally saw an uncovered version of what the affordable model likely is, as it was cruising around near Gigafactory Texas, just outside of Austin.
Tesla coding shows affordable model details, including potential price
Musk said earlier this year, candidly during an Earnings Call, that the affordable model Tesla planned to release was a Model Y.
“It’s just a Model Y. Let the cat out of the bag there,” Musk said.
The images of what we assumed to be the affordable model lined up with Musk’s candid statement:
🚨 It looks like the new affordable Tesla Model Y was spotted near Giga Texas
Model Y body with the Model 3 fascia, no glass roof, and looks as if there is a front bumper camera!
Should be coming soon! https://t.co/UAXQMHjM23 pic.twitter.com/9lC5te9GnW
— TESLARATI (@Teslarati) October 3, 2025
Tesla is expected to unveil its affordable model tomorrow during the planned event, which has been teased twice. Pricing and other details are still pending, but the company is expected to reveal this information tomorrow.
News
Three things Tesla needs to improve with Full Self-Driving v14 release
These are the three things I’d like to see Tesla Full Self-Driving v14 improve.
As Tesla plans to release Full Self-Driving version 14 this week after CEO Elon Musk detailed a short delay in its rollout, there are several things that continue to plague what are extremely well-done drives by the suite.
Tesla Full Self-Driving has truly revolutionized the way I travel, and I use it for the majority of my driving. However, it does a few things really poorly, and these issues are consistent across many drives, not just one.
Tesla Full Self-Driving impressions after three weeks of ownership
Musk has called FSD v14 “sentient” and hinted that it would demonstrate drastic improvements from v13. The current version is very good, and it commonly performs some of the more difficult driving tasks well. I have found that it does simple, yet crucial things, somewhat poorly.
These are the three things I’d like to see Tesla Full Self-Driving v14 improve.
Navigation, Routing, and Logical Departure
My biggest complaint is how poorly the navigation system chooses its route of departure. I’ve noticed this specifically from where I Supercharge. The car routinely takes the most illogical route to leave the Supercharger, a path that would require an illegal U-turn to get on the correct route.
I managed to capture this yesterday when leaving the Supercharger to go on a lengthy ride using Full Self-Driving:
You’ll see I overrode the attempt to turn right out of the lot by pushing the turn signal to turn left instead. If you go right, you’ll go around the entire convenience store and end up approaching a traffic light with a “No U-Turn” sign. The car has tried to initiate a U-turn at this light before.
If you’re attempting to get on the highway, you simply have to leave the convenience store on a different route (the one I made the vehicle go in).
It then attempted to enter the right lane when the car needed to remain in the left lane to turn left and access the highway. I manually took over and then reactivated Full Self-Driving when it was in the correct lane.
To achieve Unsupervised Full Self-Driving, such as navigating out of a parking lot and taking the logical route, while also avoiding illegal maneuvers, is incredibly crucial.
Too Much Time in the Left Lane on the Highway
It is illegal to cruise in the left lane on highways in all 50 U.S. states, although certain states enforce it more than others. Colorado, for example, has a law that makes it illegal to drive in the left lane on highways with a speed limit of 65 MPH or greater unless you are passing.
In Florida, it is generally prohibited to use the left lane unless you are passing a slower vehicle.
In Pennsylvania, where I live, cruising in the left lane is illegal on limited-access highways with two or more lanes. Left lanes are designed for passing, while right lanes are intended for cruising.
Full Self-Driving, especially on the “Hurry” drive mode, which drives most realistically, cruises in the left lane, making it in violation of these cruising laws. There are many instances when it has a drastic amount of space between cars in the right lane, and it simply chooses to stay in the left lane:
The clip above is nearly 12 minutes in length without being sped up. In real-time, it had plenty of opportunities to get over and cruise in the left lane. It did not do this until the end of the video.
Tesla should implement a “Preferred Highway Cruising Lane” option for two and three-lane highways, allowing drivers to choose the lane that FSD cruises in.
It also tends to pass vehicles in the slow lane at a speed that is only a mile an hour or two higher than that other car.
This holds up traffic in the left lane; if it is going to overtake a vehicle in the right lane, it needs to do it faster and with more assertiveness. It should not take more than 5-10 seconds to pass a car. Anything longer is disrupting the flow of highway traffic.
Parking
Full Self-Driving does a great job of getting you to your destination, but parking automatically once you’re there has been a pain point.
As I was arriving at my destination, it pulled in directly on top of the line separating two parking spots. It does this frequently when I arrive at my house as well.
Here’s what it looked like yesterday:
Parking is one of the easier tasks Full Self-Driving performs, and Autopark does extremely well when the driver manually chooses the spot. I use Autopark on an almost daily basis.
However, if I do not assist the vehicle in choosing a spot, its performance pulling into spaces is pretty lackluster.
With a lot of hype surrounding v14, Tesla has built up considerable anticipation among owners who want to see FSD perform the easy tasks well. As of now, I believe it does the harder things better than the easy things.
Elon Musk
Elon Musk teases previously unknown Tesla Optimus capability
Elon Musk revealed over the weekend that the humanoid robot should be able to utilize Tesla’s dataset for Full Self-Driving (FSD) to operate cars not manufactured by Tesla.
Elon Musk revealed a new capability that Tesla Optimus should have, and it is one that will surely surprise many people, as it falls outside the CEO’s scope of his several companies.
Tesla Optimus is likely going to be the biggest product the company ever develops, and Musk has even predicted that it could make up about 80 percent of the company’s value in the coming years.
Teasing the potential to eliminate any trivial and monotonous tasks from human life, Optimus surely has its appeal.
However, Musk revealed over the weekend that the humanoid robot should be able to utilize Tesla’s dataset for Full Self-Driving (FSD) to operate cars not manufactured by Tesla:
Probably
— Elon Musk (@elonmusk) October 5, 2025
FSD would essentially translate from operation in Tesla vehicles from a driverless perspective to Optimus, allowing FSD to basically be present in any vehicle ever made. Optimus could be similar to a personal chauffeur, as well as an assistant.
Optimus has significant hype behind it, as Tesla has been meticulously refining its capabilities. Along with Musk’s and other executives’ comments about its potential, it’s clear that there is genuine excitement internally.
This past weekend, the company continued to stoke hype behind Optimus by showing a new video of the humanoid robot learning Kung Fu and training with a teacher:
🚨 Some have wondered if this is ‘staged’ or if Optimus is teleoperated here
Elon Musk said this is completely AI https://t.co/N69uDD6OVM
— TESLARATI (@Teslarati) October 4, 2025
Tesla plans to launch its Gen 3 version of Optimus in the coming months, and although we saw a new-look robot just last month, thanks to a video from Salesforce CEO and Musk’s friend Marc Benioff, we have been told that this was not a look at the company’s new iteration.
Instead, Gen 3’s true design remains a mystery for the general public, but with the improvements between the first two iterations already displayed, we are sure the newest version will be something special.
Tesla plant manager tips off affordable model production
Three things Tesla needs to improve with Full Self-Driving v14 release
Elon Musk teases previously unknown Tesla Optimus capability
-
Elon Musk2 weeks agoTesla FSD V14 set for early wide release next week: Elon Musk
-
News1 week agoElon Musk gives update on Tesla Optimus progress
-
News2 weeks agoTesla has a new first with its Supercharger network
-
News2 weeks agoTesla job postings seem to show next surprise market entry
-
News2 weeks agoTesla makes a big change to reflect new IRS EV tax credit rules
-
Investor's Corner1 week agoTesla gets new Street-high price target with high hopes for autonomy domination
-
Lifestyle1 week ago500-mile test proves why Tesla Model Y still humiliates rivals in Europe
-
News1 week agoTesla Giga Berlin’s water consumption has achieved the unthinkable