Advertisement
Please note that it is not permitted to have referral links posted in your signature. Keep these links contained in the appropriate forum. Thank you.

https://www.boards.ie/discussion/2055940817/signature-rules
Help Keep Boards Alive. Support us by going ad free today. See here: https://subscriptions.boards.ie/.
https://www.boards.ie/group/1878-subscribers-forum

Private Group for paid up members of Boards.ie. Join the club.
Hi all, please see this major site announcement: https://www.boards.ie/discussion/2058427594/boards-ie-2026

Tesla Talk 2

1848587899094

Comments

  • Registered Users, Registered Users 2 Posts: 274 ✭✭evftw


    You work with developing the technology, the drivers in US and Australia drive it and seem quite happy



  • Registered Users, Registered Users 2 Posts: 60,762 ✭✭✭✭unkel


    Yes there are some vids out of the new Xpeng system being driven in a busy hectic city in China in bad weather. Far more difficult circumstances than in a typical US city. Crazy drivers, most people don't even indicate. Have a look for yourself. It seems capable

    And I guess you haven't seen any of the Tesla FSD beta public demos from maybe 50-60 EU cities from the last 3 months either? About a million vids of those about. And then there is the totally illegal vids like the cybertruck that was dumped into Prague city centre. FSD worked fine and no, it did NOT have maps!

    "Make no mistake. The days of the internal combustion engine are definitely numbered" - Quentin Willson, 1997



  • Registered Users, Registered Users 2, Paid Member Posts: 34,886 ✭✭✭✭listermint


    Yes I work within hardware and software that utilises imaging location GPS accelerometer firmware and AI technology to determine visual aspects of imagery to ascertain it's environment and what's going on within it.

    I've explained this multiple times but I get treated like a spoofer because I disagree with Tesla's lack of hardware for safe self driving . I've been consistent on this for years.

    Derided for evidenced based opinion.

    But I'm quite happy to listen to anyone on the thread that works with this on a daily basis and not just blurbing out twitter talking points . Quite happily.

    Oh btw US and Australia would be key locations that we utilise this in vehicles btw. 👍



  • Registered Users, Registered Users 2 Posts: 129 ✭✭antseanoifig


    Having stood idely by for a while now observing this conversation I'll finally bite and be the credible activist you seem to frame yourself as but also demand others need to have an opinion on this topic.

    You say you work with perception stacks, so you’ll know the key question isn’t how many sensors you bolt on, it’s whether the system can interpret the environment reliably at scale.

    Tesla’s approach is vision-first with neural nets trained on billions of miles of fleet data. That’s fundamentally different from the traditional sensor-fusion stack most AV programmes use.

    Right now in the US we’re seeing FSD run full end-to-end city driving with no route pre-mapping, handling roundabouts, unprotected turns, construction zones etc. The system is learning behaviourally rather than relying on high-definition maps and LiDAR point clouds.

    Whether someone likes the approach or not, the evidence is that it works in the real world:

    • Hundreds of thousands of vehicles running the stack daily
    • Billions of real-world miles of training data from the fleet
    • Continuous OTA improvement cycles measured in weeks, not model years

    Most other AV programmes are still limited to geofenced areas or require expensive sensor stacks that won’t scale to consumer vehicles.

    On the regulatory side, Europe isn’t rejecting the tech either. UNECE regulations for assisted driving are evolving and several updates already allow higher-level automated lane-keeping and driver assistance features. Tesla’s stack already exceeds what those frameworks were originally written for.

    So the real debate isn’t “Tesla lacks hardware”. It’s whether vision + AI at fleet scale can outperform traditional sensor fusion.

    Given what we’re seeing in the US right now, dismissing the approach outright feels more ideological than technical.



  • Registered Users, Registered Users 2, Paid Member Posts: 34,886 ✭✭✭✭listermint


    All true, however your summary is unclear

    I've said their hardware is not consistent with having an effective full safe driving.

    Because it doesnt give enough data points to fully determine it's environment. That's a sensor problem because camera by it's nature isn't capable of providing enough information.

    Il put it to you. Why has Tesla reported at a very minimum x4 times the rate of crashes per same mileage distance versus Waymo as a competitor. Those statistics don't even compare apples and oranges because the stats from last year are a fully autonomous waymo system and a tesla 'robotaxi' with a assistance driver for safety. So we don't even have clear data as to how many times the safety assistant intervened. Logically the incidents are far higher.

    The data doesn't show a good picture for vision only solutions. No matter how many people in here want to be chauffeured around Dublin putting other road users in a less safe position.

    I gather by the number of thanks on your post people simply don't give a shite about other road users. Fairly obvious.

    I just want to also add, can you expand on the 'what were seeing in the US now'. Is Tesla expanding robotaxi? How many are driving out there where ? What are we seeing I'm intrigued you seem to infer that it's now a fleet at scale and showing up what I'm saying to be false. Intrigued.... Details please.



  • Advertisement
  • Registered Users, Registered Users 2 Posts: 274 ✭✭evftw


    Theory vs. practice comes to my mind here. According to LM the FSD can't work without the sensors but does so in the real life, and better than the other implementations with the extra sensors.

    Edit: And just to be clear, I don't talk about the Cybercab, but the L2-L4 consumer FSD for my personal car.



  • Registered Users, Registered Users 2, Paid Member Posts: 34,886 ✭✭✭✭listermint


    • Oh and just to follow up on your word smith. Which is right out of Teslas marketing speil. By trying to be a 'activist' as you put it.

    Tesla is using LLMs to determine the imagery, yes LLMs not "neural nets" .. this guff has to stop at some point.

    Their updated hardware is why they have to keep moving the goal posts as older boards don't have the GPU or CPU required for these image based LLMs to do their work at speed.

    Neural nets.... ..



  • Registered Users, Registered Users 2, Paid Member Posts: 34,886 ✭✭✭✭listermint


    Theory versus practise ? What ?

    It's got more than 4 times the crash rate of it's competitors that use lidar and video.

    What are you talking about ?



  • Registered Users, Registered Users 2 Posts: 129 ✭✭antseanoifig


    You’re conflating a few different things.

    The “cameras can’t provide enough data” argument doesn’t really stand up when Tesla is using an 8-camera surround system with temporal neural networks to build a full 3D occupancy model of the environment. Humans drive primarily using vision.

    The Waymo comparison also isn’t apples to apples. Waymo runs geofenced robotaxis in tightly mapped environments with extremely expensive sensor stacks. Tesla is attempting a generalised system that works anywhere a human can drive without HD maps.

    Of course the statistics look different when one system operates in a restricted domain and the other is deployed across the entire road network.

    You can disagree with the architecture, but saying vision-based autonomy “can’t work” while it’s already running on hundreds of thousands of vehicles feels like an outdated position.

    As I said I've stood back and watched your commentary on this thread for quite some time. You got what you wanted now which is someone who works on this daily to comment. However, it’s clear you’ve already made your mind up, so I won’t be continuing to engage if the discussion can’t move beyond that bias.



  • Registered Users, Registered Users 2, Paid Member Posts: 34,886 ✭✭✭✭listermint


    I just want to also add, can you expand on the 'what were seeing in the US now'. Is Tesla expanding robotaxi? How many are driving out there where ? What are we seeing I'm intrigued you seem to infer that it's now a fleet at scale and showing up what I'm saying to be false. Intrigued.... Details please.

    I'm still interested in what your saying here. You seem to be making claims about Tesla's success in the US here?

    And they're also Geofencing themselves too but you know that.

    I see you're a fire and forget having blurbed out nonsense. Being caught using neural net fluff, ignoring teslas own statistics and then legging it claiming the high road.



  • Advertisement
  • Registered Users, Registered Users 2 Posts: 129 ✭✭antseanoifig


    You’ve misunderstood the point.

    Nobody said Tesla has a robotaxi fleet operating at scale in the US. What’s happening is hundreds of thousands of vehicles running FSD across normal road networks, generating billions of miles of real-world driving data.

    That’s very different to a small number of geofenced robotaxis operating in mapped zones.

    I’m surprised someone claiming technical familiarity with perception stacks would conflate those two things.

    Goodnight.



  • Registered Users, Registered Users 2 Posts: 274 ✭✭evftw




  • Registered Users, Registered Users 2, Paid Member Posts: 34,886 ✭✭✭✭listermint


    You're proving my point, despite Teslas driving across more regions in the US and collecting data points from their vehicle cameras for what near a decade ?

    They still have a worse accident rate in Austin for example versus Waymo . Even and including the fact they are being monitored by in vehicle assistance drivers.

    You're proving, what I'm saying

    Good night indeed. And enough with the Jibes about my job. I haven't made a single assertion about yours. It's bloody rude. But I suspect you came on here to be just that. As your opening post demonstrated.



  • Registered Users, Registered Users 2 Posts: 316 ✭✭tppytoppy


    Tesla February sales in Turkey were only 720. This was a star market for Tesla last year. It is supplied from Germany. I fear the Norwegian market is the next to come under pressure mostly from Mercedes, BMW and later Id.tiguan and new peaq/enyaq. If the Germans provide 4WD models then Tesla will suffer as Norwegian market is affluent and 4wd is a selling point. The presales on Merc and BMW are over a year and one must assume a sizable minority of them are heading to Norway



  • Registered Users, Registered Users 2 Posts: 22,123 ✭✭✭✭Cyrus


    Someone should tell Tesla to sell 4wd versions of the 3 and Y!



  • Registered Users, Registered Users 2 Posts: 316 ✭✭tppytoppy


    They do and that is one of the few reasons why sales hold up in Norway along with short delivery times. Tesla are very vulnerable in what remains of their European market.



  • Registered Users, Registered Users 2 Posts: 5,236 ✭✭✭wassie


    From a mechanical standpoint, highly unlikely to have an EV that is a 4WD. It will be an AWD.



  • Registered Users, Registered Users 2 Posts: 5,236 ✭✭✭wassie


    By the classic 1990s/2000s definition, an ICE awd has a driveshaft connecting the front and back axles.

    4WD has a transfer case for low and high range.



  • Registered Users, Registered Users 2 Posts: 316 ✭✭tppytoppy


    Are all four wheels driven?

    Try translating four wheels drive in to German and you will get Allradantrieb i.e. AWD. You can nitpick on spurious grounds all you like but Tesla Europe is in a precarious position at present and AWD or 4WD won't steady their course.



  • Registered Users, Registered Users 2 Posts: 5,236 ✭✭✭wassie


    "You can nitpick on spurious grounds all you like".

    Lol, no idea what that is meant to mean.

    Was simply offering a nostalgic viewpoint of what was accepted as a 4WD/AWD back when we all used to enjoy the stink of filling up our cars. As a former 4WD enthusiast, many a debate was had on this subject.



  • Advertisement
  • Registered Users, Registered Users 2 Posts: 274 ✭✭evftw


    Alternatively it's also a possibility that Tesla are just good quality cars that also provide good value for money and the most mature EV market in the world knows this.



  • Registered Users, Registered Users 2 Posts: 7,502 ✭✭✭MojoMaker


    Thanks @antseanoifig - finally an actual expert on this topic is contributing.



  • Registered Users, Registered Users 2, Paid Member Posts: 34,886 ✭✭✭✭listermint


    He doesn't work in video analysis.

    He also owned two Tesla's. He has a bias which isn't backed up by an operational involvement in this technology other than driving the car.

    But I do thank you for your back handed Jibes. It's like a small child who can't articulate a defence to what I'm saying so you've decided to tarnish my credentials. All I've done is shown facts and figures and explained why it's not workable.

    Anyway, as I'm sure all of the folks who are keen for their car to run people over will be happy I've had enough of this topic. I won't be commenting on it any longer as I'm being personally attacked and tbh I don't need it

    Run someone over, run into a parked vehicle whatever floats your boat . 👌



  • Registered Users, Registered Users 2 Posts: 7,502 ✭✭✭MojoMaker


    Please forward LinkedIn profiles guys, and we'll settle it with some basic checking here.

    1 of you is an expert, and the other isn't. Or maybe both of you are, or maybe neither of you are.

    How are we to know? You both sound knowledgeable but I'm going with @antseanoifig because he's shouting a lot less about how he's an expert and the rest of us are blind or deranged.



  • Registered Users, Registered Users 2, Paid Member Posts: 34,886 ✭✭✭✭listermint


    I'm not forwarding my private information for you.

    Ive years on this website explaining I'm in technical engineering . Why would I bother making this up in 2026 just to get one over on a Tesla owner.

    Bizarre stuff from you today



  • Registered Users, Registered Users 2 Posts: 274 ✭✭evftw


    So according to you the opinion of people who drive Tesla is somewhat less valuable than yours? Quoting you here: How bizarre.



  • Moderators, Sports Moderators Posts: 18,873 Mod ✭✭✭✭slave1


    Not really, there is plenty of evidence online that Tesla AWD is virtually all RWD with the front motors using very little energy unless called up by flooring it etc. from memory a typical drive around town and country roads had 1% of energy spent via the front motor. There are plenty of aftermarket apps that can demonstrate this data.



  • Registered Users, Registered Users 2 Posts: 5,236 ✭✭✭wassie


    My understanding is that Tesla, Hyundai/Kia and VW typically use the rear motor as the primary drive unit. How they do it also differs. Tesla does it electronically while Hyundai/Kia do it mechaniically by disengaging the front wheels via a clutch.

    Meanwhile the likes of Audi & Volvo/Polestar prefer FWD. This is not uniform across all models though.



  • Registered Users, Registered Users 2, Paid Member Posts: 34,886 ✭✭✭✭listermint


    I was asked about expertise in image and video analysis, industry experience.

    So yes I don't think the opinion of people owning a car and driving it has any value in how imagery is analysed by LLMs both on vehicle or post processing .

    That is not bizarre.

    In fact it's a symptom of today's world where people seem to think opinion trumps facts , science , data and engineering. It's rampant.

    Perhaps if I throw my words into chatgpt to ask it to restructure it in a nicely worded format you'd take me as seriously as you did anto. I could use neural networks to make my words fancy. (I guess)



  • Advertisement
  • Registered Users, Registered Users 2 Posts: 7,502 ✭✭✭MojoMaker


    The point you're missing is that we have absolutely no evidence one way or the other that you have any of the experience or knowledge you claim. That is at the same time the beauty and the curse of an online bulletin board I guess.

    Anto sounds credible, in fact some would say more credible than yourself, but it is nuanced I'd imagine. Someone could come on here and spout anything they liked, using fodder served up from any number of AI tools going back years to make one sound convincing and somehow authoritative on any random given topic.

    The reality is we're all bumbling around here offering opinion in an unsubstantiated capacity, no more no less. Unless you want to pop up a LinkedIn profile with references we'll all just tip along wondering idlily is yer man serious etc. But it is entertaining to see how openly annoyed you get when you're not taken as seriously as you believe you are 😄



Advertisement
Advertisement