-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Tesla Software updates - Australia
- Thread starter Grenadine
- Start date
Wol747
Active Member
>>Surely the size of anything in the display is pointless? All that matters is what the driver sees, which is either you looking out the window or the cars camera looking through its lens. Not sure why everyone wants tesla engineers wasting time on dash visualisations. Surely perfecting the current ‘autopilot’ and making it more reliable is a better starting point?<<
I find that an extraordinary point of view.
You are in charge of a lethal machine, which is driving you using algorithms which none of us know yet for which we are responsible. Knowing what the machine is about to do - or not do - is surely almost the definition of "responsible"?
For my part I find the instrument cluster impossible to follow: today, in reasonably strong sun, the only part of it constantly legible was the large speed readout: the left and right displays were just about visible and the centre with cones (AKA just about anything to one side or the other), vehicles and especially traffic lights and the like unreadable.
From what I've seen on YouTube videos, the Model 3 is a little better.
I find that an extraordinary point of view.
You are in charge of a lethal machine, which is driving you using algorithms which none of us know yet for which we are responsible. Knowing what the machine is about to do - or not do - is surely almost the definition of "responsible"?
For my part I find the instrument cluster impossible to follow: today, in reasonably strong sun, the only part of it constantly legible was the large speed readout: the left and right displays were just about visible and the centre with cones (AKA just about anything to one side or the other), vehicles and especially traffic lights and the like unreadable.
From what I've seen on YouTube videos, the Model 3 is a little better.
Would be interesting to know. I suspect they do vet new version to some degree & don't roll them out when the local offices are closed. Once the version is known to be stable they would push it any time, even on Anzac day. Your version 2020.12.20 looks like unicorn as it does not exist according to Teslafi. Either way not much update activity in AUS at the moment.
I think you misunderstood my point. I was trying to convey that spending time getting autopilot working more reliably should be a priority over visualisations. Visualisations are going to be useful once we are permitted to have a hands off wheel experience, and until then, you as driver are very much in control. You also shouldnt assume the visualisation is about to tell you what the car is about to do. It never warns for example that its about to slam the brakes on for no particular reason. It never says that it can see the construction zone coming up.
I have a similar attitude to these in car games, supposedly there to keep us entertained whilst the car is driving itself. Well maybe more focus on getting the car to drive itself is more beneficial than games.
Meanwhile my tesla cannot react to pedestrian crossings, work zones, roundabouts, traffic lights, stop signs, give way signs. It wont slow down from a 60 zone to a 50 zone, and it wont slow down for the blue flashing lights on the side of the road. When choosing a priority between these, visualisations, and games, to me the driving attributes should come first. Fluff later. I’m sure plenty wont agree with my position, and I openly acknowledge that my tesla is the most advanced and incredible car on the planet.
I see this argument a lot but not all programmers are created equal. Just like you probably don't want your neurosurgeon operating on your colon, you probably don't want your Cuphead games developer jumping into and coding your Tesla's self-driving features that other AI developers have been specialising in for years.
But Tesla probably has enough money for both, so let them both be developed seperately.
It's not one or the other. There wouldn't be that many cross-skilled developers able to work as both to the point of providing positive benefit.
To be fair though, I'd still prefer the neurosurgeon operating on my colon than the Cuphead games developer.
ARMARM
Member
2020.12.11.1 has gone bananas overnight. Over 850 worldwide installs on Teslafi alone in the past 12 hours! Looks like > 100 cars in Australia have got it overnight. Nice.
I thought the games were there to amuse us while the car was charging. Would never dream of playing them while the car is moving.
@paulp the visualisations are our only window into the AI that's driving the car, as such, I feel it is very important. It provides us with a narrative of what the car thinks it knows. Without the visualisations, it's a complete blackbox. The vis at least gives us a peeping tom hole... arguably a very small one with only a small fraction of the car's knowledge displayed, but nonetheless it's a start.
Top priority items that need fixed based on my observations still are:
- reliable detection whether side traffic is two lanes over or one (it gets that wrong often)
- stop setting TACC speed limits from map database information (the car rarely knows where it is with sufficient precision to do so reliably)
- but, if it must: when TACC speed limits are reset when driving into a new speed zone, it sometimes sets it to the limit, sometimes to the limit +3 km/h (the offset I have it set to). As a software developer myself, I dread thinking why that is happening... looks like there are two separate subsystems setting that speed. One does not respect the speed limit increase I have set. Shitty coding has no business being in a safety critical system.
- address the phantom braking issues
- give us individual autosteer disconnect ability (e.g. push stalk once: AS disconnect, push twice: AS+TACC disconnect, same as stepping on the brake).
Top priority items that need fixed based on my observations still are:
- reliable detection whether side traffic is two lanes over or one (it gets that wrong often)
- stop setting TACC speed limits from map database information (the car rarely knows where it is with sufficient precision to do so reliably)
- but, if it must: when TACC speed limits are reset when driving into a new speed zone, it sometimes sets it to the limit, sometimes to the limit +3 km/h (the offset I have it set to). As a software developer myself, I dread thinking why that is happening... looks like there are two separate subsystems setting that speed. One does not respect the speed limit increase I have set. Shitty coding has no business being in a safety critical system.
- address the phantom braking issues
- give us individual autosteer disconnect ability (e.g. push stalk once: AS disconnect, push twice: AS+TACC disconnect, same as stepping on the brake).
Similar threads
