Відео

Yeah you can’t really compare New York driving to anywhere else . That’s why it’s such a good testing ground for FSD. It’s got its own rules lol.

Yeah it’s a little uncomfortable but if it’s perfect every time would you care.

@ as long as Tesla says its drivers have the responsibility then, yes I’d care. Fender benders can be expensive. Tesla would be the first to say people should still worry about things like this because it’s not perfect yet. However, I get your point and it is valid, in future when Tesla is going to cover mistakes, no problem.

I was emptying that you would not be responsible but I agree if for now it should take all the space it needs.

@@FSDRoboRides Same thing for me with following distance (12.5, HW 3), the car can react so much more quickly than I can it does OK, but still doesn't feel "right". Takes time to get used to.

Yeah the mirrors did not fold lol I wish they would have.

Thanks for watching. These are my favorite but it’s so hard to find people who want to join me lol

@FSDRoboRides if I could fly up I would :)

Yeah the way the system reacts to situations is so human like.

Please do! Thank you so much for your support.

I live of RT22 in Bridgewater Area so i take it every day. But not down by Newark. Are you from that area?

@@FSDRoboRides I'm in Passaic county but I pass through there whenever I go down to the shore

@@bener289 If you want we can meet up and go for a FSD ride shoot me an email at fsdroborides@gmail.com.

It seamed so.

That and much much more.

My passenger was buying big after this ride.

@@FSDRoboRides Can t blame him! I ve been buying pretty much every 20 dollar jump myself :)

I'm glad my videos are helping you build your conviction in FSD. If you keep watching them I will keep making them.

I would only agree to that in a Robovan lol

We believe it!

Only Time will tell.

@originalambival Your perspective as an engineer is very valuable, but there are several arguments in favor of using DOJO as a backup for Tesla’s Full Self-Driving (FSD) system: 1. Scalability and Efficiency: DOJO is designed specifically for handling the vast amounts of video data required for training neural networks, offering significant computational power. This could enable rapid iteration and improvement of FSD algorithms directly from real-world data, which is essential for scaling up to handle the diverse scenarios encountered in autonomous driving. This scalability could surpass the capabilities of human-operated solutions in terms of speed and coverage. 2. Continuous Learning and Adaptation: While you mention the risk of hallucinations, Tesla’s approach with DOJO isn’t just about solving individual cases but about learning from them to improve overall system performance. DOJO’s architecture allows for continuous learning, where the system can refine its models based on new data without solely relying on predetermined outcomes. This could potentially lead to fewer hallucinations by constantly updating and correcting the model based on new driving scenarios. 3. Cost and Operational Efficiency: While human remote operators can address specific scenarios, they come with significant operational costs and logistical challenges. DOJO, once set up, could offer a more cost-effective solution for backup by automating the learning process, reducing the dependency on human intervention for every new or edge case scenario. 4. Data-Driven Decision Making: Tesla’s fleet generates an enormous amount of data daily, which DOJO can process to enhance the FSD system. By leveraging this data, DOJO can identify patterns or edge cases that even experienced human operators might miss or not encounter frequently enough to learn from effectively. This data-driven approach could lead to a more robust and safer backup system for FSD. 5. Integration with Existing Systems: Tesla already uses a classification and auto-tagging pipeline for its FSD system. By incorporating DOJO, this pipeline could be enhanced, allowing for immediate feedback into the training loop, thus improving the system’s capability to handle new situations without solely relying on human intervention for each new scenario. 6. Overcoming Human Limitations: Human operators, while insightful, are limited by fatigue, attention span, and reaction times, especially in emergencies or complex scenarios. DOJO, being a machine learning system, doesn’t suffer from these limitations, offering a consistent backup that can react instantly to any scenario within its training scope. While your point about human oversight is well-taken for current safety and regulatory reasons, integrating DOJO as a backup could significantly enhance the system’s self-improvement cycle, potentially leading to a future where human intervention is less frequently required. This doesn’t negate the value of human oversight but complements it by providing a scalable, data-driven approach to continuous improvement.

@@FSDRoboRides Your points are very valid, and I agree that its scalability and resiliency are fantastic assets. I also agree that it can (and probably will) serve as a fantastic classification and training tool. My point was limited to addressing the idea about letting it generate solution data in cases when human intervention would otherwise be required. At such times, having human intervention (as drivers, supervised CC employees are doing today, and remote will do in the future) to create solid training data for these (hopefully) edge cases will be more valuable. After all, the "successful human driver" golden path is the target. Once it gets past that point, I expect the auto-tagging/classification, modelling and synthetic regression tests will all take place on the large scale, as you suggest.

@originalambival While I acknowledge the value of human intervention for handling edge cases, especially in the current state of technology, I’d argue that becoming as good as a human driver is indeed only the first step. Here’s why: 1. Human Limitations: Humans have inherent limitations when it comes to handling stress, fatigue, or complex, unexpected situations. Even the best human drivers can make errors or react sub-optimally under pressure. An AI system like DOJO, as part of a larger AI ecosystem (let’s call it “Big Mama AI”), could analyze and learn from these stressful scenarios in ways that humans can’t, providing a safer, more reliable backup for FSD. 2. Edge Case Management: While human intervention can provide valuable data for current edge cases, the sheer volume and variety of situations a vehicle might encounter would be overwhelming for human operators to cover comprehensively. A system like DOJO, could not only react in real-time but also predict and learn from potential future scenarios based on a broader dataset than any human could handle. 3. Real-Time Updates and Fleet Learning: With an AI system monitoring and updating the fleet’s capabilities continuously, any new edge case encountered by one vehicle could be instantly analyzed and the learnings disseminated across the entire fleet. This is far more efficient than waiting for human intervention, analysis, and then manual update distribution. 4. Beyond Human Performance: The ultimate goal isn’t just to match human performance but to surpass it. An AI system that can learn from every vehicle in the fleet simultaneously, update its understanding of the world, and refine its operational strategies offers a path to a safety and efficiency level that humans can’t achieve due to our biological and psychological constraints. 5. Synthetic and Scenario-Based Learning: Once a certain baseline of performance is established, DOJO could use synthetic data, simulations, and predictive modeling to prepare for edge cases that haven’t yet been encountered in the real world. This could significantly reduce reliance on human intervention by preemptively addressing potential issues. Your point about the “successful human driver” as a target is valid for initial stages of FSD development. However, the vision should extend beyond this. With AI’s capacity for continuous learning, adaptation, and data processing, we can aspire for a system where the AI doesn’t just react to human solutions but proactively improves upon them, making autonomous vehicles safer and more capable than any human-driven counterpart could be. This is where having a Big Mama AI overseeing and updating baby AIs in each car becomes not just beneficial but necessary for true autonomous driving at scale.

I love my model Y, because of Max

Hey mom

Yeah now we just have to convince my dad lol but I think he is holding out for the cybertruck lol

Hey, love you❤​@@FSDRoboRides

My too❤

Thank you I appreciate your support.

Thanks for always watching my videos.

@vinod8941 I hope you share this sentiment on X with them channel link????

Thank buddy I’m assuming the more emojis the better the video. Let me know if that the case.

@FSDRoboRides yes I love your videos

The progress is incredible. The scenarios that it now can handle are wild.

Thank you. Obviously it’s speculation on my part but I imagine that Tesla would want humans out of the loop as much as possible.

I agree with you 💯. Tesla is right in the middle of it all.

Hahaha yeah this a was a fun one. When are you coming on a ride?

​@FSDRo😂😂😂❤boRides

Glad you enjoyed it

I cut the dead space of us rolling up behind the taxi ahead of us in the distance. Unfortunately the emergency vehicle turned off and never caught up to us to interact with us. It would’ve been great to see what FSD would do. I pride my self on being supper transparent in my videos. I never cutout anything other than silence or mundane driving like in this case of us catching up to the taxi in front of us. Thank you for bringing this to my attention. I will be more mindful with my cuts moving forward.

@@FSDRoboRides A lot of other channels fast-forward through those parts instead, especially when there's some movement but nothing interesting. I think it helps build trust and is also smoother if you don't cut, personally. Looking forward to part 2 :)

Awesome. Part 3 will have a situation where the passenger I was with is preparing for the fist disengagement of the ride lol so that should build some trust between us.

Thanks for sharing good luck with growing your channel.

@ just sharing the love, but you too!!

O yeah hopefully retire in few years on those Tesla shares.

Thank you so much I look forward to your feedback about the content or presentation.

Looking forward to it. Let me know when your state side. But don’t wait too long lol. FSD maybe solved this year.

I appreciate your support.

It's so cool to meet tesla fans from all over the world. getting a perspective of investor in europe is very informative.

Glad you enjoyed it. This is as a fun one. They are always better with a partner in crime.

What do you mean?

@@FSDRoboRides Just watching you driving on the bumpy roads made me wonder if the car's internal computer system wouldn't be affected. If not, then I guess it's a real sturdy system.

@@tekmepikcha6830 That's an interesting point. If i had to guess I would be confident Tesla engineers have considered it during design.

@@tekmepikcha6830 There are 7 million Tesla's with system boards and I've never heard of one internal computer system that had an issue with bumpy roads.

💯

Wait till you see the upcoming once next week.

@@FSDRoboRides Looking forward to that! Do you agree with poor map data and sun glare are the two main issues now?

@ yes that’s my experience. Just posted let me know what you think.

Thank you for sharing this information

Honestly there is rarely time for blinkers lol. Your head is constantly spinning on a swivel lol.

The car was driving 100% of the time. If I hit the brake the FSD would have turned off. That maneuver was all FSD just like every scenario in this drive. Wait till you see the next few videos things get wiled. So make sure you check back in.

I like most of the upgrades inside. Not crazy about the front end but I will reserve judgement till I have a chance to see it in person. It seams like it’s geared yours Robotaxi work with the rear screen and auto folding backseats.

According to Grok - Based on recent data, the daily average of reported accidents in New York City is around 625. And it drives me around for three hours without any intervention.

@@FSDRoboRides 625 a day!?

O yeah NYC is definitely top boss lol.

The software has gotten that good that I trust it to do the right thing in most scenarios.

Thank you have good weekend.

Glad you enjoyed it! More to come with.

Thank you I’m trying to live up to all of your expectations lol

Yeah what are thoughts? I'm trying it out for next few videos.

I def think it keeps it more focused on what I care about at least which is the visualization and the front view. I think one problem with the side mirrors is that because of the camera angle we weren’t even able to see what you were seeing, so this gives us a better view without really sacrificing much.

@@sshrdl Cool

@ hey any thoughts on the new model Y refresh

Keep the trend going.

Thanks a lot.

I’m having the same experience. It’s really exciting. I’m looking forward to see the progress in the future updates.

Fingers crossed!