Charity bets: $100/$1K/$10K -> EAP/FSD feature release by 12/31, 1/31, 3/31

[gearchruncher]

Can we please un-ban Calisnow so we can make him pay up tomorrow?

Not a single new feature on AP since June 2016, much less an EAP feature. Just a week short of 300 days so far.

Yes, significant progress on actually keeping a car in the lane, but didn't we all assume that was a solved problem back in October 2016 when they showed us a FSD video, not something they would be working on well into 2018?

 

[croman]

Can we please un-ban Calisnow so we can make him pay up tomorrow?

Not a single new feature on AP since June 2016, much less an EAP feature. Just a week short of 300 days so far.

Yes, significant progress on actually keeping a car in the lane, but didn't we all assume that was a solved problem back in October 2016 when they showed us a FSD video, not something they would be working on well into 2018?

What scum. He pushed a false narrative worse than Russian trolls.

 

[BinaryField]

Yes, significant progress on actually keeping a car in the lane, but didn't we all assume that was a solved problem back in October 2016 when they showed us a FSD video, not something they would be working on well into 2018?

I feel that the progress on the lanekeeping is somewhat over-hyped, possibly due to the recent relative big jump in an otherwise slow development environment. The control scheme still leaves a lot to be desired. I still feel rapid micro-oscillations in straight and well-marked lanes; the acceleration and deceleration profile is still way too aggressive; and the recent update has dramatically increased the false braking rate and severity. The list of deficiencies can go on.

It's nice to see progress on AP2, but it still has a long way to go for even a basic functionality.

 

[gearchruncher]

I feel that the progress on the lanekeeping is somewhat over-hyped, possibly due to the recent relative big jump in an otherwise slow development environment. The control scheme still leaves a lot to be desired.

I totally agree. I threw that in because the apologists for Tesla will tell you how there have been "updates" to AP over the last 9 months and how much better lane keeping is. I'll agree it's way better than it was 9 months ago, but it's still not good enough, and better lane keeping is not a feature release, it's an improvement.

This is that company that used a video that opens with "the driver is only here for legal purposes" to introduce AP2 over 500 days ago, and immediately started charging for EAP, saying AP1parity and EAP features were only 70 days away. With that in mind, It's not impressive when they manage too keep the car in the lane a bit better (while causing other regressions), nor a good sign for overall AP feature progress.

So, who's taking the next round of bets for when we'll hit AP1 feature parity, and when the first EAP feature will appear?

 

[croman]

I'm so sick of these Tesla apologists. They are dinging me in the HWY 101 thread for pointing out EAP shouldn't have put the driver in danger in the first place. Product liability is a clear avenue for a system that does not operate as designed (and clearly it wasn't or the driver wouldn't have been driven in a non-drivable area into a barrier).

Tesla is disgusting in that its shifting blame to the driver to make up for the fact their system still does not function adequately. Its certainly better but grossly inadequate for what Tesla claims it does.

Its amazing how far the delusional Tesla fans will go to protect a poorly implemented EAP ("E" still MIA 15 months later and AP1 parity only appearing a beta release that still doesn't have local autolane change).

 

[u00mem9]

Seriously. Where are the defenders these days? Tell me one more time that Elon didn’t lie...err, misrepresent the status. Or solarcity. Or model 3 status.

Seems like the tide has turned this quarter. Where denial was the norm, there is suddenly realization that the stack of false claims over the last 2 years is staggering. I have friends that still work at Tesla, but aren’t high enough to bail at the moment with a vested options parachute.

Back to False self driving, I don’t see the point in continuing this path. Wayno / Google have the nut cracked. Tesla’s last 2 years of production will NEVER operate autonomously. Move on, make some refunds, settle some suits, and pay the company that spent time solving the technical challenge instead of publicity touring, flame thrower fundraising, and bullsh*tting. Show’s over folks.

 

[croman]

@u00mem9 -- they are all dinging me for pointing out EAP is not functioning as designed in the HW101 thread in the autonomous vehicle subforum. They are all hit and run disagrees without any substantive debate because there is nothing to debate anymore. Even with Karpathy's rewrite, Tesla acknowledges the past 15 months was just polishing a turd of EAP.

 

[BinaryField]

Even with Karpathy's rewrite, Tesla acknowledges the past 15 months was just polishing a turd of EAP.

This reminds me: does anybody remember the shareholder letter a few quarters ago talking about how the foundation of the neural network was finally right? That completely flies in the face of the replacement in 10.4. Apparently, the old one was so good that it earned a retirement!

 

[u00mem9]

@u00mem9 -- they are all dinging me for pointing out EAP is not functioning as designed in the HW101 thread in the autonomous vehicle subforum. They are all hit and run disagrees without any substantive debate because there is nothing to debate anymore. Even with Karpathy's rewrite, Tesla acknowledges the past 15 months was just polishing a turd of EAP.

Unfortunately, the opinion of those posters is irrelevant. Tesla is going to get a hard lesson from grownups (regulators AND the tort system ) on that accident. I’m guessing we will find out that previous SW updates did handle the spot, and the driver was caught out by an update that suddenly didn’t.

 

[gearchruncher]

@u00mem9Even with Karpathy's rewrite, Tesla acknowledges the past 15 months was just polishing a turd of EAP.

Not EAP. AP. Remember, for people with AP1 orders converted to AP2, Tesla told them they could keep AP1 features for what they had already paid, or upgrade to EAP features for additional money. At this point, nobody is missing a feature because they didn't pay for EAP.

So the turd they were polishing still wasn't even EAP, it was just trying to do what Mobileye had been doing for 5+ years in a small ASIC that fits in a camera housing instead of needing a 200W water cooled GPU.

 

[gearchruncher]

I’m guessing we will find out that previous SW updates did handle the spot, and the driver was caught out by an update that suddenly didn’t.

That's a fascinating observation. When exiting the express lanes here in Seattle, 10.4 dives at a barrier almost every time because it can't figure out the lane split even though there is a nice white line on your right. Older AP code had done thus too, but had stopped about 4 months ago. It was quite a surprise when it suddenly did it again. Based on that catastrophic behavior, I bet 10.4 is actually less safe to humans than the previous version, which didn't hold lanes perfectly, but also seemed more reasonable in picking your lane location when there wasn't two perfect lines. 10.4 just wants to sit in the middle of whatever it thinks is a lane at that moment while previous versions appeared to hug the right line when available.

 

[daktari]

When this thread started, I really did not think we would pass this date without EAP features. Sad. Wonder how OP feels.

 

[outie]

@u00mem9 -- they are all dinging me for pointing out EAP is not functioning as designed in the HW101 thread in the autonomous vehicle subforum. They are all hit and run disagrees without any substantive debate because there is nothing to debate anymore. Even with Karpathy's rewrite, Tesla acknowledges the past 15 months was just polishing a turd of EAP.

I still remember the responses when I was calling out the self driving demo in early 2017. More than one year later today we have yet to see anything new.

 

[Canuck]

I still remember the responses when I was calling out the self driving demo in early 2017. More than one year later today we have yet to see anything new.

Well in the post you quote, he did say "it still has a long way to go before it's ready for release". I said things in support without that caveat since Tesla sure fooled me with that FSD video, and I thought it was just around the corner. But I didn't go so far as to post pictures with money going to charity. That requires actual proof of payment if you lose, or reveals serious character flaws if you lose and then jump ship. I sure hope the charity got paid.

 

[croman]

No charities got anything.

 

[BigD0g]

No charities got anything.

I tried!! I guess a fool and his money aren't easily parted

 

[u00mem9]

No charities got anything.

I guess you aren’t counting Tesla / Elon ?

 

[TaoJones]

As predicted, he who has welched repeatedly now has at least $11,120 reasons to never darken these forum doors again.

Meanwhile, it's worth noting that we do have Elon's most recent estimate for when FSD will be enabled across the board. I mention it because presumably we'll receive at least one EAP feature *before* FSD (makes sense, right?). That FSD estimate was by year-end 2019.

I almost believe Elon's latest timeline estimate for this FSD stuff (and EAP prior thereto).

Why? Two reasons - first, recall that Tesla has committed to their own SoC/board design (by 2019-2020-ish) and that can only help. Second, you've got to love the work that Karpathy has done so far, and you can sense that he's just getting started.

I don't care about auto lane changes to overtake slow traffic or half of the other silliness that's in the marketing material. What I want is what we were told was coming "soon" back in October 2014 - and that would be stop sign *reaction* - not just recognition. Youse may recall this functionality in the contrived December 2016 video *upon which many based their Q1 2017 purchases in conjunction with Tesla's claim that included supercharging would go away after April 15, 2017*. Little did we know that they would bring it back a month later. Some might say there's a term for that, and that it rhymes with "sait and bwitch".

Specific rollout milestones and feature sequence aside, 18-24 months from now should be a Great Time in TeslaWorld.

If we get some of these features sooner, then great. I doubt the competition will have much to offer that's better by late 2019 anyway.

 

[gearchruncher]

Second, you've got to love the work that Karpathy has done so far, and you can sense that he's just getting started.

Kaparthy took almost 300 days to roll out a new NN just to keep a car between two lines better, and it has significant and dangerous regressions that have likely already killed one person. What about that timeline makes you think FSD is on any kind of track?

 

[TaoJones]

Kaparthy took almost 300 days to roll out a new NN just to keep a car between two lines better, and it has significant and dangerous regressions that have likely already killed one person. What about that timeline makes you think FSD is on any kind of track?

What, you didn't get the latest Kool Aid? It's quite tasty, really.

But (more) seriously:

While I find AP2 to remain significantly inferior to AP1, the reasons for which I've enumerated in another thread, here are my reasons for continued optimism as long as Karpathy remains at the helm:

1. Significant recent improvement in AP2. It's still a nervous child missing features relative to the confidence inherent with AP1, but still.
2. Regular releases every 2 weeks like clockwork, relatively speaking. As if there were sprints and program increments and everystuff.
3. Elon stopped saying "by the end of the month" and "3 to 6 months" and instead gave a 2-year projection that actually fits with the SoC estimate.

Putting #3 another way, I don't see the bulk of FSD happening until we get that SoC. Granted, not to mention better car connectivity which means 5G, and that's a good 3+ years out - on the front end anyway.

So that's how I come up with late 2019, which is to say perhaps the first half of 2020. I have no idea how Elon makes those of us whole who purchased vehicles based upon the Dec 2016 video and who prepaid for FSD, but if it takes a board swap and perhaps a component upgrade or two, that's not the end of the world.

One video of 10 Teslas lined up at a stop sign (*with the first one stopping therefor*), all accelerating in unison when safe to do so, will send the stock soaring. Just in time for the shorts and the unions to beat it back down at the end of the quarter, but still.

It will be a beautiful thing. Just that one feature.