I certainly don't think that SpaceX's test failure or a car catching on fire four months ago had a whole lot to do with TSLA being down today.

I'm watching the presentation because I came across it in my YT feed.

Musk is so awkward.

The neural network trainer is now on stage, it's a literal mind****ing circus.

Elon feels very desperate. If you really figured out a different route to open up a new field of technology potentially worth billions and billions do you really throw down the gauntlet like that? LIDAR not only suboptimal but a FOOL's ERRAND and anyone pursuing it is DOOMED?

Will google et al laugh it off or perceive this as an actual threat to the valuation of their companies?

Fully self driving explained via iguanas! This is amazing

They are talking about basicly image recognition software atm what the hell

musk is talking. vol is expanding and the stock is rising....

From this talk it seems clear they're taking the overthinking dickhead route toward FSD (which isn't suprising given that Musk is a "big thinker - this is just like Musk's dickhead 10 year old notion of full robot factories). The guy talking reminds of me pure CS clowns who tried to solve poker using pure algorithms. Meanwhile the non-tards solved it by a knowledge matrix and got working profitable bots a decade earlier.

The Lidar guys are mapping a 3D space at ultra high resolution with extreme accuracy, which forms a highly reliable base for decisions. Tesla are trying to build a replica of a human mind driving with a probabilistic model with no hard 3D environment. They're 5+ years away from having reliable level 4 using this strategy while Google/GM are taking over the self driving world. Complete morons.

The big data stuff stuff to train the "neural net" - which they only just built the chip for by the way - is a whole lot of nonsense. It's big thinking nonsense from someone who can't see details and can't build complex models and get the right answer.

So after watching the rest, basically Elon is placing his trust in this kid's view of how to build FSD, which seems to be non-practical purist. Neural nets are nowhere near fast enough for this, with nowhere near enough data - you need a human-level mind - and cameras are nowhere near reliable enough - insufficient resolution for fine detail edge cases, insufficient range for dealing with flares, rapid lighting changes, etc. You'll basically never get enough data to train these the way they're training them.

The idea of learning how to drive by using data from actual drivers is absolutely comical. It indicates to me that their engineer is trying to throw neural nets at everything, which is a smart-people dumb. Driving strategy (not object and world recognition) needs to be programmed and not neural net learned from drivers. That's just crazy.

If anything they made the case for lidar, its necessary for depth perception even in this nonsensical methodology.

Musk is saying the neural net approach is much longer road to travel, but the results will be 3x better, so its worth it. Come on shareholders, do you believe?

"ONCE YOU SOLVE VISION [LIDAR] is worthless."

Quite a caveat there, bro.

Yeah, let's just make Skynet then we can throw away Lidar. No **** bro.

This is intellectual purist bull**** from people without the smarts to see problems 10 steps down the road (e.g. Elon).

Q: "I live in Minnesota, and you don't have lane lines in snow. How do you handle that?"
A from engineer: "...We didn't try training the neural network in snow yet!!!...we think it'll work ok"
A from engineer: "As long as you as a human has driven through this environment, the neural network will learn it".

They're in pure fantasy world here. What a clown show. My favorite of all:

Musk weeks ago: "We'll have FSD by the end of the year"
Also Musk just now: "We'll have much better snow data after next winter, it should be very good by then"

Haha awesome question:

Q: "What about parking lots and similar situations where it's more an art than a science and you're dealing with other drivers"?
A: "There's a long tail in negotiations...we have a lot of confidence that there must be something in machine learning that will eventually learn that"

What a frigging clown show.

Musk on the time line:
- (paraphrased) We expect to be feature complete this year. Regulatory approval though.

Just another day of tsla fsd rigs on the road



They poached the software guy from SNAP, at least they got something right

Basically their FSD strategy is a moonshot by purists. A bit like Musk's "machine that builds the machine" "alien dreadnaught" with "air friction limited robots" that failed so spectacularly it ended up with people building cars in a tent with 1980s technology.

This whole thing is a joke.

edit: more gold...

10,000 Model 3 + S + X per week by the end of the year. So a massive walkback from "absolutely no concern about 10,000/week by the end of 2018". He's not even claiming he'll hit it in 2019 now.

Fully Self Driving Robotaxis next year: "exponential", "exponential", "exponential", "very hard for the human mind to understand exponential", "I feel very confident predicting autonomous robotaxis next year".

Solar roof now V3, rolling out later this year, lol.

"I have it on my house." Which one, Elon?

Current Teslas sans the battery pack are designed for a million miles of operation...new battery pack is designed for a million miles of operation with "minimal maintenance".

Cost of running a robotaxi: 18c per mile right now - and dropping, vs 62c for ownership and $2-$3 for ridesharing.



He's the greatest bull****ter of all time.

Turns into a megastutterer when asked about how they worked out the bull**** pricing.

Why is the cabin designed like that if the cars don't have pedals or a steering wheel? Even his mock ups don't go through qa.

Elon admits they just threw in a number for cost of ROBOTAXI.

It seems incredibly dishonest to:

*Sell the Tesla as a solid investment based on future promises of robotaxi.
*Saying they'll produce 2,5 million cars / year in 5 years.

"Extremely cash flow positive once the robotaxis are enabled!"

This is some crazy stuff.

Remain cash flow neutral during ramp!

Question about Musk's claim that Tesla fully self driving will be feature complete this year:

Q: "I assume you're talking about level 5, no geofence"
Musk: "Yes"

According to Musk Level 5 will be here in less than 8 months, folks. Still gotta work out that snow though over the winter.

Definition of Level 5: Under all roadway and environmental conditions that can be managed by a human driver

As a reminder: Tesla has zero real world autonomous miles in California in 2018 where their team are. Zero. And he's claiming they will be level 5 in 8 months.

People are going to look back and wonder how the entire world got conned by this guy, not unlike Elizabeth Holmes and Bernie Madoff.

How much is Tesla spending on Autonomy on an annual basis?

"Its basically our entire expense structure."

Huh?

They don't have a custom FSD asic? What are they using?

o/u on months the chip guy is still around after they 'finished' their chip now?