FSD video completely fake?

[3Victoria]

Most replies here are rather over-reaction. Take a look at the disengagement reports, they are mot hard to find, Google is your friend, but here is the Tesla 2016 report: https://www.dmv.ca.gov/portal/wcm/c...a/Tesla_disengage_report_2016.pdf?MOD=AJPERES

The Demo was in November 2016.

Here are the miles driven by each car:
Car: SYJXxx7708 SYJXxx7809 SYJXxx6777 SYJXxxx6199
Oct 129 97 211 93
Nov 0 20 0 0

The disengagements in October were almost entirely 'Planner output invalid' and 'Follower output invalid' which suggests that they ere working on the route-planning and execution part of the system. There were *no* disengagements during November.

Its pretty obvious that once they had the route-planning and execution working, the demo functioned well. This does not mean that FSD was any more than a very promising demonstration. Obviously it needed much work to generalize it and ensure corner cases were included. I do not think the current EAP progress tells us anything about FSD other than they have competence.

I agree that the 2017 Disengagements Report in January may give us more information on progress -- with the caveat that that report only applies to testing on *public* roads.

 

[lunitiks]

with the caveat that that report only applies to testing on *public* roads

... in the golden state

 

[thecloud]

You do realize that "NIPS papers" refers to scientific papers at a well-known machine learning conference?

What!? We were led to believe this was a standard sleazy Reddit PM_ME_YOUR_$BITS account. Now you're saying the wording of this account name is misleading, and not what it seemed??

 

[AnxietyRanger]

... in the golden state

Well, better that than the other golden thing.

 

[Swift]

Most replies here are rather over-reaction.

Not if you bought the car and paid for FSD based on the 2016 video along with Elon’s associated comments and tweets...

 

[KyleDay]

This guy actually works for Tesla on their AP team, or so he claims. Read his other reddit posts. Their pretty insightful; such as the effectiveness of front-facing sonic sensors having trouble with wind noise, etc. He's got several really informative posts.

Correcting my own dumb self - "they're" not "their". I know better and I blame autocorrect.

 

[lunitiks]

*sugar*

 

[McRat]

Most replies here are rather over-reaction. Take a look at the disengagement reports, they are mot hard to find, Google is your friend, but here is the Tesla 2016 report: https://www.dmv.ca.gov/portal/wcm/c...a/Tesla_disengage_report_2016.pdf?MOD=AJPERES

The Demo was in November 2016.

Here are the miles driven by each car:
Car: SYJXxx7708 SYJXxx7809 SYJXxx6777 SYJXxxx6199
Oct 129 97 211 93
Nov 0 20 0 0

The disengagements in October were almost entirely 'Planner output invalid' and 'Follower output invalid' which suggests that they ere working on the route-planning and execution part of the system. There were *no* disengagements during November.

Its pretty obvious that once they had the route-planning and execution working, the demo functioned well. This does not mean that FSD was any more than a very promising demonstration. Obviously it needed much work to generalize it and ensure corner cases were included. I do not think the current EAP progress tells us anything about FSD other than they have competence.

I agree that the 2017 Disengagements Report in January may give us more information on progress -- with the caveat that that report only applies to testing on *public* roads.

While this is not a new discussion, there is something unusual about the data.

They ran 4 cars about 500mi over 4 days. Lots of disconnects. Stopped for 4 days. Pulled only #xx708 out Oct 22, and again, disconnects.

Then some time in November, #xx809 was run for only 20mi, this time no disconnects. No other cars were tested.

Think of that what you will.

My take is they were trying to make a demo video Oct 14-17. All 4 efforts were not deemed adequate. They tweaked the code for a few days and loaded it in #708 only. Tried again, no dice.

Then they went to Plan B some weeks later using #809. They ran it for about 1 hr (20mi, no disconnects) and got their video, then stopped all public testing for the year. It's up to you to come up with why and how.

 

[KyleDay]

Recall they created two videos

 

[3Victoria]

Recall they created two videos

Yes, the first was criticized for having multiple cuts -- people wanted to see a continuous video -- hence the second video.

 

[croman]

They can just take my S and give me one of those magic Xs that FSD...

 

[harry]

You do realize that "NIPS papers" refers to scientific papers at a well-known machine learning conference?

But it's funnier the other way.

 

[TaoJones]

Obviously it is a preplanned route with a lot of hard coding. Afterall it is a future capability demo.

The visualizations are a representation of what the system infers. Very similar to the color images we see in astronomy of nebulae.

As long there is no human driving it in REAL TIME, sitting inside the car or even remotely it is not fake.

This is blatantly false.

At no time during either video’s release was it made obvious that NONE of the demonstrated capabilities were going to be available. Parity was promised by the end of 2016, and significant EAP differentiation within 3-6 months. Based upon the video, that included managing stop signs - functionality that has been referenced since late 2014 at the AP1 event.

It’s now a year later and three years after the October 2014 event, and it is painfully obvious what happened. And what didn’t happen.

Revisionist history serves nobody except blatant fanbois/gurls who should, as I’ve noted elsewhere, have “Rah RAH sis boom BAH” tattooed across their foreheads and sigs.

 

[Super_Popular]

I would be really surprised if that person actually works at Tesla, maybe a subcontractor but actually working at Tesla....doubtful. Why?, Tesla has a very strong, strict policy that it's employees do not post stuff online.

 

[buttershrimp]

Finally, the real missing FSD video:

 

[Electroman]

Parity was promised by the end of 2016, and significant EAP differentiation within 3-6 months. Based upon the video, that included managing stop signs - functionality that has been referenced since late 2014 at the AP1 event.

It’s now a year later and three years after the October 2014 event, and it is painfully obvious what happened. And what didn’t happen.

What is the lack of sticking to the timeline Musk promised, anything to do with the issue at a hand that the video was fake or not?

Sure it was orchestrated on a carefully planned and rehearsed route with a few controlled and uncontrolled variables. But isn't that what ML and AI is all about? in a much larger scale with lot more variables

 

[oktane]

Fraud

 

[Electroman]

Fraud

Thank you.

 

[3Victoria]

Certainly no fraud. EM was clear that this was in development. Anyone that has done any software development knows that the best estimate is likely 2-3 times better than reality. Also, the progress of FSD is likely a trade secret and not shared. So I am not holding my breath on when we get to see the results.

 

[alloverx]

Certainly no fraud. EM was clear that this was in development. Anyone that has done any software development knows that the best estimate is likely 2-3 times better than reality. .

Er no, it comes down to planning with resources and risks, along with estimating skill assuming the estimator has a handle on the technical platform. So yes it can go 3 times if some of the risks go true.