This is likely why the EU is taking its time making sure FSD will actually work and not kill somebody. It's also notable that Tesla and Waymo only operate within fixed geofences so they probably need to load a heap of data into their systems
I think the only way for FSD to become a reality is if every manufacturer works on the same software at the same time
Of course I haven't made any such bet 10 years ago. At least I hope not 😂 It would have been purely based on faith. Like people believing in God. I'm an atheist.
Can you show me where I made that bet? Not until some time last year did FSD beta become really convincing to me.
I think the point is that you’d have made that same bet any year of the past decade.
Having a vehicle capable of driverless travel, and having government policy that allows its >safe< usage, in >all< conditions, are two very different things.
Tesla can talk until they are blue in the face able how they’ll get around weather and low sun. The Austin trial forbids them from having these vehicles on public streets in anything other than optimal weather conditions for this very reason.
I think ultimately it’s the safety requirements that will be their downfall.
Haha with me? Can't find anything I am looking for on boards with the boards search function or even with google, so I gave up using it years ago
Pretty sure I already have a bet on here from a couple of years ago reaching to the end of 2030
You don't think it would happen in the USA within the next 10 years? Wanna bet €1000 with me that it will?
Europes roads are ancient relative to the geo US locations with FSD trialling, some of our L roads are shocking twisty/hilly with absolute zero visibility of where to go. Let them perfect it on better roads then see how it goes….
Considering the first cars sold as 'coming with full self driving hardware' are now reaching end of life (2014 model S were described as having fully self driving capable hardware), and we're almost 10 years into Tesla autopilot (as opposed to the aforementioned Mobileye product now known as AP1), I love the optimism of 'the next few years', I really do. No one would love a legal and free taxi home from the pub! But I don't think it will happen in the next 10 years. There may be betas, geofenced tests, etc etc, but until the blinkers are removed and it can be done anywhere without a driver, it doesn't count.
The musk of 2014-19 would have made it work, but I think he's … lost his charm…
Yep if you lived in the USA likely within the next few years. In Ireland later. We are under the control of the UNECE and they are not only extremely slow to adapt to developments, they seem to keep throwing up barriers. No idea how much later it will happen here after it happens in the USA. Even China allows FSD beta now. Europe is falling behind in all sorts of areas, we will be the laughing stock soon enough, if we aren't already.
the Canada Q1 and Q2 deliveries really highlight the effects on sales of tariff wars, EV rebate removal, and public sentiment.
In the 8 quarters of 2023 and ‘24, Canada delivered avg 12,000 each quarter.
They delivered just 860 3/Y vehicles in Q2 2025, and by all accounts online, many of the 35 nationwide Canadian locations sold no cars at all in June. Most were Q1 hangover deliveries in March.
https://x.com/TroyTeslike/status/1941156737886879848
I was outlining where we are at now. I said exactly:
Self driving means no one in the driver's seat, and no company currently sells that
No one currently sells it to private companies. Tesla are not even the closest. Do you think Tesla will have a car that I can own that can legally drive me home from the pub under their insurance, and not only that, be first to market to do so? Note I live rurally, approx 15km from the nearest town.
You're throwing all 6 levels of self driving into one big bucket there 😂
What I will predict again is that Tesla will be the first to reach level 5 autonomous driving for cars owned by private customers: no driver needed under any circumstances. The car (manufacturer) is responsible.
Tesla do not sell any cars capable of 'self driving' right now. They sell cars with driver assists branded as FSD, EAP, AP, etc.
GM and Ford sell cars in the US that are legally capable of self driving. Bluecruise and supercruise (eg check out the latest cadillac) offer hands off self driving on geofenced roads. Ford bluecruise is even available as hands off in the UK. Now, *IMHO*, that doesn't qualify as self driving either, because the insurance is not covered by the OEM and you can't be drunk in charge etc. Mercedes has a limited pilot of self driving where the OEM covers the insurance, however you still need a driver as it is level 3-4 and not level 5 .
Self driving means no one in the driver's seat, and no company currently sells that. Mercedes, GM and Ford are closer than other car companies in terms of current purchaseable options, and Waymo are ahead of other companies in terms of products in the market.
… 😀
I feel like this debate goes in circles constantly. Some people do not understand how much Musk has kneecapped Tesla's FSD ambitions by going the camera-only route. The limited onboard hardware resources for real-time inference, and the lack supporting hardware(radar, lidar) to reduce computationally expensive inference of objects shape, distance and velocity. And that's before we even talk about limitations of current model limits due to their inability to generalise.
There's a concept named Moravec's paradox which observes that it's relatively easy to make computers perform at an adult level, and even excel to a super intelligence level on rule based tests or playing complex games like chess or go(see AlphaGo), and impossible for humans to match that speed and accuracy. So essentially computers excel at rule based logical operations.
On the other hand, it's extremely difficult to give computers the skills of a 1 or 2 year old when it comes to perception and mobility, such as picking up an object because humans had millions of years of evolution to continuously improve sensorimotor skills. These abilities are so intrinsic for humans that they are largely unconscious. We don't need to think about the complex physics of balance and coordination to walk across a room, we just do it. When we pick up an object, our brain instantly processes a huge amount of sensory data. It takes in visual information such as an object's shape, size, location, tactile feedback from your fingers such as texture, pressure, temperature, and proprioceptive data (which is knowing where your hand is in space). A computer needs to be programmed with complex algorithms and equipped with numerous sensors to even begin to replicate what we percieve as a simple act. To do all of this at a real, fully automous level, a model would need to generalise(AGI) and have multisensor inputs, and do all this with limited onboard resources.
ok what car can joe public buy that has better current self driving than tesla (i am not trying to be smart my assumption is its the best available currently but i am not that well up on it to be honest)
I think that's very very optimistic. It will be an extreme pareto. Vision can do 80-85-90% of the FSD functions perfectly and much cheaper than lidar or radar hardware. Especially with all the machine learning that tesla are using. However that last 20-15-10% or less will be very very difficult to get with vision only.
I used to be an optimist but when it comes to tesla, not only is my glass not half full or half empty, it's exploded.
Joe public can't buy a robotaxi either.
Waymo is light years ahead of the current tesla self driving. Other companies are further down the no driver/hands off autonomous driving now, think of GM and Ford in the US with bluecruise and supercruise. Mercedes in Germany have some limited hands off too.
Currently however, other than the waymo cabs and a limited test by Mercedes, no other company has a self driving car where the user of the car is not classed as a driver and the insurance is covered by the OEM.
When fully self driving teslas are sold to the public, level 5, with no legal driver required and insurance covered by the OEM, then it's a fair comparison.
is there a better autonomous driving car that joe public can buy? ive no skin in the game, i think this is the future, dont know if itll be tesla or someone else but they have certainly pushed the agenda further than anyone else so far.
Call me an optimist, but I think it can be done with vision alone. It's very, very difficult, but not impossible. Sun glare? I put on the sun glasses. Rain, or snow, on go the wipers. Fog….while human eyes can't penetrate that too well, plenty of cameras can.
10 years ago, it might have seemed impossible to land a rocket upright and reuse it. Tesla are going for the cost solution and while redundancy might seems absolutely now, it may not seem that way soon.
Yes, and that is that it doesn't work 100% of the time. Of course it works some of the time. But some might as well be zero, you can't get in an FSD driverless car that works some of the time.
Yes it works some of the time. I've seen the videos, I love the idea, it's just not good enough and Tesla are not even the best autonomous driving product out there anymore.
Yet if you goto reddit there are lots of us folk that gush about fsd
Might have come from this sentence
Humans do not rely on sight alone, it is impossible to safely create a drive anywhere robotaxi without a second source, preferably lidar, to complement vision.
Humans do not rely on sight alone, it is
impossible
to safely create a drive anywhere robotaxi without a second source, preferably lidar, to complement vision.
I don't think Elm has ever said vision doesn't work but rather it doesn't work 100% of the time which is what's required.
No, I agree about the ridiculous overpromise and the wait. But up to now you have been giving out about the tech. It will never work, it is not possible, vision doesn't work, need lidar, etc. But now you say ah sure I know it is only a trial but I have a problem with the wait
How is it disingenuous, people have paid for this software for 6-8 years without being able to use it. You can't just arbitrarily decide that today is day 0
That's disingenuous. You criticise the technical aspects, the vision based approach, FSD in general will never work, needs lidar, etc.
Of course Musk has been absurdly optimistic in his promises. Nobody is denying that. I have no idea why there hasn't been a class action in the USA for early customers of FSD to get their money back and then some. If I had paid a cent for FSD myself, I would have got it back in full in the small claims court already. Tesla's lawyers in Ireland probably all know my name now 😂
My criticisms are in the context of FSD being promised for 10 years. Not in 10 days of a trial.
"Remote monitoring and a guy sitting in the car" while true is without context. It is the first live trial and of course it is going to be with some limitations, used to provide feedback loop, learn, and respect safety, so a degree of caution is to be expected …to do otherwise would be silly.
This first trial/test does not represent the ambitions of perfecting the technology and ironing out issues before being scaled. I am not making a case for it either way, other than to inject some balance as criticisms for the first test of a new technology for being done at a small scale with built in contingencies to address safety issues are to be expected, praised rather than criticised.
The current robotaxi fsd is geofenced.
Humans obviously don't use lidar. But try and drive with only your eyes, remove sound, sense, gravity, etc. How do you hear an emergency vehicle upcoming? Or adjust for the wind when passing a truck? Or hear a young child about to run on to the road? Or even simpler, hear the beeping of a horn?
We drive in 3 dimensions, not two. We have multiple inputs, not just vision. Some are conscious like sound and others are somewhat subconscious, like the effect of gravity. I said it the day they changed to tesla vision, and I stick with it, you cannot have level 5 drive anywhere FSD/robotaxi/whatever other name, where it can pick me up at the pub and drive me home, without multiple inputs, including at least one of radar or lidar. Level 5, no steering wheel, no remote monitoring, no insurance required for the driver (insurance covered by oem or operator like waymo).
Humans use lidar?
FSD is by definition not geofenced. That's the whole point. It can work anywhere. Waymo is geofenced. As in the entire infrastructure is programmed in. That's why it would struggle if there is a major change in the roads until it has been updated. Of course the Austin trial is limited to a certain area of Austin that is the safest to drive in, so temporarily geofenced of sorts
