-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Tesla, TSLA & the Investment World: the Perpetual Investors' Roundtable
- Thread starter AudubonB
- Start date
In one of the many parallel universes (hopefully not this one), Elon stops in front of Zuckerberg's house... Zuckerberg was watching the stream and comes out angry. Words are exchanged, Elon exists the vehicle and the (non)cage fight begins, live on camera ...
Interesting live stream. Elon is giving lots of info about v12. Zero c++ code, all neural net, trained by just watching video. No labeling. No explicit designation of what a traffic light is, what a lane is, what a roundabout is, etc. it just learns what to do.
The drive is no better than whole mar's video but what is coming out of Elon's mouth is mind blowing. How V12 is being trained is just 100% videos. In fact the only way for V12 to stop at stop signs fully is by finding <1% of the video data from the fleet of people who actually stopped at stop signs. They did not hard code anything into the software such as traffic lights or roundabouts. Just show videos to the NN and it'll figure it out.
Seems like V12 is a full re-write by AI and not by humans.
Seems like V12 is a full re-write by AI and not by humans.
Anyone outside California would think it was staged, given how many other Teslas there are.
I’m not watching the stream but is Elon saying that the big difference between V11 and V12 is that in V11, all of these things were coded to tell FSD what to recognize and what action/function to perform while in V12 all operations and functionality are run by the neural nets now, which require no actual coding, just data training?
Yes, exactly that.
Basically, no road rules have been programmed. It literally just drives based on what it has learned from watching good drivers drive. They have lots of traditional code in the data center to curate the training videos to find good driving and exclude bad driving. I can’t wait to try it on my funky roads I have around my house.
They did have one intervention in about 30 minutes of driving around Palo Alto. Major intersection, stopped at red light, driving straight, left turn arrows came on and the car tried to go straight but the straight ahead lights were still red.
They did have one intervention in about 30 minutes of driving around Palo Alto. Major intersection, stopped at red light, driving straight, left turn arrows came on and the car tried to go straight but the straight ahead lights were still red.
Elon’s risk tolerance was on full display here. He joked that assassins could find him if they wanted to, but he thought the number of assassins in Palo Alto was pretty small.
Makes me wonder how he knows this.
That was an interesting stream by Elon with FSD v12.
I have to wonder if the learning rate of FSD will increase with V12 being pure AI training? This could be quite the step change for FSD once v12 releases, and it could be quite the boon for TSLA if the training rate does increase drastically. It could mean TSLA feels the impact of a solved FSD sooner than most people believe.
I have to wonder if the learning rate of FSD will increase with V12 being pure AI training? This could be quite the step change for FSD once v12 releases, and it could be quite the boon for TSLA if the training rate does increase drastically. It could mean TSLA feels the impact of a solved FSD sooner than most people believe.
It took what, 2 years+ to do roundabouts from "non zero% chance" comment from Elon to at least V10.69? Remember how terrible roundabouts were when FSDb was first streamed by youtubers? V12 went from not knowing the concept of driving to doing roundabouts. I feel like V12 has surpassed 5 years of FSDb performance in like 4 months(and remember they just got V12 online because occupancy network was developed recently which is required for V12 training).
With V12 being fully neural nets, now computing power becomes essential for a S curve of improvement in FSD.
Luckily it sounds like Dojo has gone live and Tesla has gotten a big order from Nvidia to help out.
While not exactly fun to think that they have to re-train for Hardware 4,0, the reality is that if Tesla's computer power combined with how neutral nets can exponentially learn faster than hard coding, we may see a re-write by the neural nets do what took Tesla's coders/engineers 2 years to do in the matter of 3-6 months. If that actually happens, then the computing knowledge of the neural nets is about to reach escape velocity.
Edit: What @Singuy just said.
Elon confirmed on the livestream @ 9:30 that FSD V12 with pure NN runs faster on HW3 than the C++ code + NNs and estimated to go up to 50fps (cameras are only capable of 36fps).
So where are the "experts" at on this now?
He also confirmed that the "ADAS test operators" (Tesla job listing) all over the world are testing V12.
Don't worry.. Elon will be giving us cheques* for free lol
(* checks for mericans)
So where are the "experts" at on this now?
He also confirmed that the "ADAS test operators" (Tesla job listing) all over the world are testing V12.
Don't worry.. Elon will be giving us cheques* for free lol
(* checks for mericans)
Last edited:
"Watson, come here. I need you!"
This glitchy livestream is history in the making.
It's amazing watching this revolutionary transition to NN visual training as the most important industrialist of the 21st Century casually and humorously describes what's happening while giving viewers a look behind the curtain. And the fact that all of this will function when not tethered to the Internet is an astounding testament to the capabilities of Tesla AI and its coming applications.
Summary? Freaking BULLISH!!!
This glitchy livestream is history in the making.
It's amazing watching this revolutionary transition to NN visual training as the most important industrialist of the 21st Century casually and humorously describes what's happening while giving viewers a look behind the curtain. And the fact that all of this will function when not tethered to the Internet is an astounding testament to the capabilities of Tesla AI and its coming applications.
Summary? Freaking BULLISH!!!
Last edited:
dhanson865
Well-Known Member
Not "anything" but "some things", you can get to some content on Twitter URLs that don't work on an X url still and you can get to some types of content without logging in (like a single message) but not all content (like a full thread or this livestream).
IMO Tesla will at least look at what Lucid has done and see if any of it can be applied to CT.
I'm talking about motors and the drivetrain voltage here, not frontal area and vehicle shape.
Similar threads
- Locked
- Locked
- Locked
