Self-driving cars UK Discussion

[Durzel]

All of these promises are perfectly fine to believe if you're a long TSLA shareholder. Not so much if you're an owner paying for FSD with any expectation of it being realised over the typical period of ownership.

I would caveat the shareholder benefit considerably in so much as Elon has been making big promises for several years, and we're at the point where the competition are actually delivering on systems that might be good enough for the majority of car owners. There will come a point, I feel, at which more and more people become less tolerant of Musk's incessant overpromising and missed deadlines.

Musk appears to be betting the farm on the general population wanting their cars to drive them to A to B completely autonomously. If Tesla can crack it then they could have a formidable USP in their cars. I am sceptical, though, not only of the possibility of Tesla achieving level 4 autonomy in a reasonable timeframe in any country. In Europe I don't think it will be a reality even in regulatory terms for probably 10 years.

I could be wrong but I feel like as every other manufacturer commits to a fully electrified portfolio that the shortcomings of Tesla vehicles will become more and more resonant. There seems to be more development towards basic stuff at all (headlights, wipers, other driver convenience features), all focus appears to be on FSD beta. It's a bit of a different kettle of fish in the States I think, because they have all of the autonomy stuff unencumbered, but in Europe our Teslas are not even as good as stuff coming out from Kia or Hyundai, it's just that the general public doesn't really know that yet, and still just believes that Teslas "drive themselves".

 

[Artiste]

I thought the days had long gone when people believed the myth that Tesla are the only outfit in town. That fallacy was busted a long time ago.

The penny dropped for most people some considerable time ago, but there are still a few people who have not yet smelt the coffee.

 

[Fursty Ferret]

It's a bit of a different kettle of fish in the States I think, because they have all of the autonomy stuff unencumbered

Trust me, you don't want to be living there. They might be able to play with FSD to their heart's delight but US regulations are remarkably backwards and is why we haven't got working matrix headlights etc.

 

[Wol747]

All of these promises are perfectly fine to believe if you're a long TSLA shareholder. Not so much if you're an owner paying for FSD with any expectation of it being realised over the typical period of ownership.

I would caveat the shareholder benefit considerably in so much as Elon has been making big promises for several years, and we're at the point where the competition are actually delivering on systems that might be good enough for the majority of car owners. There will come a point, I feel, at which more and more people become less tolerant of Musk's incessant overpromising and missed deadlines.

Musk appears to be betting the farm on the general population wanting their cars to drive them to A to B completely autonomously. If Tesla can crack it then they could have a formidable USP in their cars. I am sceptical, though, not only of the possibility of Tesla achieving level 4 autonomy in a reasonable timeframe in any country. In Europe I don't think it will be a reality even in regulatory terms for probably 10 years.

I could be wrong but I feel like as every other manufacturer commits to a fully electrified portfolio that the shortcomings of Tesla vehicles will become more and more resonant. There seems to be more development towards basic stuff at all (headlights, wipers, other driver convenience features), all focus appears to be on FSD beta. It's a bit of a different kettle of fish in the States I think, because they have all of the autonomy stuff unencumbered, but in Europe our Teslas are not even as good as stuff coming out from Kia or Hyundai, it's just that the general public doesn't really know that yet, and still just believes that Teslas "drive themselves".

I've been a big Musk fan for something over a decade and although he has to be credited with driving the industry towards EV domination I think he's rapidly becoming a liability for Tesla. Grown up manufacturers don't live by Twitter.
Most "consumers" (awful moniker IMO, but there we are) don't want high tech which works one day, not the next and has to be updated every week with new bugs and different fart noises: they want affordable reliable cars that just work. The days of selling on promises that never materialise are ending.
The legacy automakers are rapidly catching up on the things that are important and it's showing.
Tesla have the range, the efficiency, the BMS and the supercharger network - and remember, folks, that WE actually pay for that albeit indirectly. Opening it to allcomers may or may not make economic sense for the company but only time will tell if it does for owners.

 

[pgkevet]

Tesla may have the 'hoover' name in electric vehicles but still only a tiny percentage of global car sales. Range and efficiency inevitably will be matched by rivals in due course. More importantly even the newspapers are making the point about electricity prices making EV's less attractive financially - and that's what matters to the end users. Who in Europe is going to buy new EV next year? Most folk will stick with ICE and wait to see if power prices drop or petrol prices go up again. Tesla sales will be fine in the US where power always seems to be cheaper.
The whole climate change, green energy thing remains serious but sadly any effort we put into it is overshadowed by population growth, aviation growth, china's 3 child policy etc - let along Russians tanks doing 2 miles per gallon on diesel. We even have the nonsense of Saudi Arabia buying Russian oil while selling us theirs for more.
Self-driving remains a curiosity that the majority will avoid on cost grounds until its mature enough that no-one needs to own a car and just calls up a really reliable robotic taxi as a vehicle that can negotiate double-parked side streets, negotiate the dust-cart or ambulance, spaghetti junction etc. Anyone;s guess when that will really happen - 15yrs minimum?

 

[Dilly]

When you look back two or three years to the time when electric was 15p/kWh, EV’s were a no-brainier to many. That situation has changed dramatically. There was a seemingly clear path to halting climate change and EV’s were part of it.
When you now look at the various crises across the world, the ‘green’ agenda though obviously essential and urgent, has been outweighed by panic to produce power.
Suddenly the likelyhood of public transport making a comeback is ever more real. A whole lot of people won’t be able to afford a car, let alone a self driving one!
In 1939, one assh*le turned the world upside down. Another is doing a pretty fair job right now!

 

[AnthonyLR]

I thought the days had long gone when people believed the myth that Tesla are the only outfit in town. That fallacy was busted a long time ago.

Show me another company with one hundred thousand members of the general public driving their cars with autonomous driving software in beta…

 

[GeorgeSymonds]

Show me another company with one hundred thousand members of the general public driving their cars with autonomous driving software in beta…

Well.. there's mobileye with 1.5 million cars feeding back data while running their latest software

Cloud-Enhanced Driver-Assist™ Takes ADAS to the Next Level | Mobileye Blog

Mobileye’s latest solution combines our expertise in advanced driver-assistance systems with the power of our cloud-based, crowdsourced mapping technology.

www.mobileye.com

 

[gangzoom]

Well.. there's mobileye with 1.5 million cars feeding back data while running their latest software

Cloud-Enhanced Driver-Assist™ Takes ADAS to the Next Level | Mobileye Blog

Mobileye’s latest solution combines our expertise in advanced driver-assistance systems with the power of our cloud-based, crowdsourced mapping technology.

www.mobileye.com

If you think that can be compared to FSD Beta development there is little else to say....

 

[GeorgeSymonds]

If you think that can be compared to FSD Beta development there is little else to say....

Yep, you’ve no idea where Mobileye are in their development.

It’s well known the AP1 was built using Mobileye tech, but not the previous iteration of Mobileye technology, not even the version before that, but the version before that. About 30 million cars a year get eyeq5 hardware in different states of function. They’re all capable of feeding back and supporting eyeq6. But the big difference is when they release a version,, they don’t make false promises, they don’t stick a ‘beta’ moniker on it, they just shop stuff that does what it says it does, which is why loads of people driving competitive cars say their systems are every bit as good as their Tesla, if not better.

 

[gangzoom]

Yep, you’ve no idea where Mobileye are in their development.

But I have just signed off on a final report for a real life AI project, that is tiny in comparison to anything discussed here, and even with my very little limited understanding of AI/NN development what you linked shows no signs of data gathering at a level needed for NN training, not even close.

Mobileye may be gathering data for AI/NN training, but the thing you linked has makes no suggestion of that at all. It's at most augmenting mapping data, a totally different thing for what FSD beta/Waymo etc are trying to do.

 

[GeorgeSymonds]

But I have just signed off on a final report for a real life AI project, that is tiny in comparison to anything discussed here, and even with my very little limited understanding of AI/NN development what you linked shows no signs of data gathering at a level needed for NN training, not even close.

Mobileye may be gathering data for AI/NN training, but the thing you linked has makes no suggestion of that at all. It's at most augmenting mapping data, a totally different thing for what FSD beta/Waymo etc are trying to do.

Watch this then

Doing in multiple countries what Tesla are only doing in the US
Have a level 3 approval
Have a approval for level 4 in dev in multiple counties
Have robotaxi deals
Have a roadmap that’s credible.

 

[gangzoom]

Watch this then

Doing in multiple countries what Tesla are only doing in the US
Have a level 3 approval
Have a approval for level 4 in dev in multiple counties
Have robotaxi deals
Have a roadmap that’s credible.

Interesting and actually very similar to Tesal, but can you point me to the bit where they are gathering mass amounts of video data from the public to than train their NN like Tesla are doing?

It's also interesting the use of 'augmented' data to simulate real life situations, Tesla are doing this too, but as the saying goes 'life is stranger than fiction'

 

[Durzel]

But I have just signed off on a final report for a real life AI project, that is tiny in comparison to anything discussed here, and even with my very little limited understanding of AI/NN development what you linked shows no signs of data gathering at a level needed for NN training, not even close.

Mobileye may be gathering data for AI/NN training, but the thing you linked has makes no suggestion of that at all. It's at most augmenting mapping data, a totally different thing for what FSD beta/Waymo etc are trying to do.

Let’s see what happens in the fullness of time shall we? Rather than completely writing off a competitor based on a passing knowledge of AI/NN and a transparent zealotry towards Tesla, eh?

Bias never helps anyone.

 

[GeorgeSymonds]

They use real world captured data which they run, it’s 4 mins into the video, 50M runs across 500K hours of driving each month. This is what Tesla are talking about doing. Tesla don’t really do much in the car, at best they capture the scenario when they think it’s failed and recreate it back at base. Different ways to do things.

 

[Wol747]

Again: there's a VAST difference between testing beta and an autonomous car. The first effectively is testing a driver aid: no more, no less. Testing an autonomous car can only realistically happen when they are released autonomously (if they ever are) and we can compare statistics meaningfully.

 

[gangzoom]

They use real world captured data which they run, it’s 4 mins into the video, 50M runs across 500K hours of driving each month. This is what Tesla are talking about doing.

Just watched it these numbers mean almost nothing. What if I told an AI project for 27k individual events, had 7 million data points, with training runs of 60k steps, does that sound good or not? These numbers have zero meaning without context or output.

Again I cannot see any where Mobileye has any where near the capability of doing what Tesla is doing. Taking raw data, sensor feeds, from live beta tester (which is exactly what beta testing is), labelling them, and feeding it into their NN.

Everyone can talk the talk, but Mobileye still hasn't let any public user near their products interms of automonoy - tech demo videos really dont count, Tesla did that in 2016.

Waymo and Tesla have given public access to their products.

 

[gangzoom]

Let’s see what happens in the fullness of time shall we? Rather than completely writing off a competitor based on a passing knowledge of AI/NN and a transparent zealotry towards Tesla, eh?

Bias never helps anyone.

Absolutely, however as an owner of a Tesla car, all I really care about is the progress Tesla is making. Is Mobileye or Waymo going to deliver automonoy capability into my current car without me having to spend ££££££ on a new car?

What it clear plenty of very very very clever people believe automonoy is possible, and very soon. Competiton is good, and there is plenty of it in this area. Someone will get there sooner rather than later, that is good news for all of us.

 

[GeorgeSymonds]

Just watched it these numbers mean almost nothing. What if I told an AI project for 27k individual events, had 7 million data points, with training runs of 60k steps, does that sound good or not? These numbers have zero meaning without context or output.

Again I cannot see any where Mobileye has any where near the capability of doing what Tesla is doing. Taking raw data, sensor feeds, from live beta tester (which is exactly what beta testing is), labelling them, and feeding it into their NN.

Everyone can talk the talk, but Mobileye still hasn't let any public user near their products interms of automonoy - tech demo videos really dont count, Tesla did that in 2016.

Waymo and Tesla have given public access to their products.

So Mobileye numbers mean nothing to you, but Tesla numbers are magnificent? Touch of bias there?

And there are multiple videos driving through various world cities doing what Tesla can only do in the US but you can’t see any evidence?

And the regulators have approved level 3 in certain countries yet Tesla haven’t declared 1 mile of L3 driving anywhere.

To me you seem blown away by pretty graphics and enthusiastic tesla fanboys but aren’t comprehending the task, rigour, roadmap required, etc etc and what it’s going to take for Tesla to close down flaws to the point when there are none for L3 and above. I’ve watched the latest videos, you don’t need to look far to find driver intervention in any of them.

 

[Type2]

I believe he believes them. 10.69 is a big step forward. I believe Tesla will be the first (and probably only) company to solve FSD. I think they do this by 2024. Let’s see.

I presume you mean you think FSD will be solved in the US by 2024? Personally I think there’s zero chance of that.

But I’m more interested in what’s happening in the UK. Or rather the complete lack of anything happening. Would you care to predict when I’ll get the “auto steer on city streets” that I was promised by the end of 2019?