-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Elon: FSD Beta tweets
- Thread starter Terminator857
- Start date
MarcG
Active Member
“We’re too dumb to realize how simple the answer was”
diplomat33
Average guy who loves autonomous vehicles
Could be. Wayve also has a "foundational world model". They showed some clips of it performing driving tasks in a simulation.
It certainly looks like an exciting and promising new tool that should help with training E2E.
Yeah, that's my point. It takes time and effort to convert a possible or promising solution into a viable product. These foundational world models look like a very promising tool for training L5 autonomy. I can see why people are excited. But turning it into a viable product will take a lot of work and effort. In the recent keynote, Shai Shalev-Shwartz mentions that these new large model AIs like chatGPT, while they demonstrate some impressive capabilities, are also still very unreliable (can do weird, stupid stuff). I think achieving the very high reliability needed for safe AVs will be the huge challenge of these new "foundational world models". So while they are an exciting new tool, I don't think they are the quick and simple solution that Elon seems to think. I think achieving affordable, safe, reliable L5 (which is the ultimate goal) will require redundancy and a complex, robust, solution. There is no quick, easy solution to L5.
Last edited:
Not even, there has not been a single NN/ML/AI Breakthrough from Tesla.
All they do is use what Google has invented, they even make it clear in their presentations.
This is Elon doing what he does best. Being the biggest liar the world has ever seen.
The gullible will easily fall for it.
Last edited:
Foundation model is now an unmbrella term that has become the latest hype buzzwords since chatGPT. Now everyone is claiming to have a foundation model. it’s not some AI breakthrough. Not surprised Elon/Tesla is getting in on the action to drum up more HYPE. The biggest liar in the history of mankind.
Last edited:
diplomat33
Average guy who loves autonomous vehicles
True. It is not new. Tesla copies others. So now they have apparently built their own generative AI world model or some E2E model and Elon is touting it as the latest Big Thing that will "solve" FSD. We've seen Elon tout the Next Big Thing before, with "4D" and "stitching vision together", then "occupancy networks", then "transformers", and not it's "end to end" and "foundational world models".
By the way, the hallmark of any scam is whenever someone makes a bold claim without showing anything for it but just wants their followers to accept it at face value (and of course pay for it in advance). It is amazing to me to see all the Elon followers on Twitter just totally believing that Tesla has solved L5, done deal, when Elon won't even say what the "simple solution" even is and has no viable product to show for it. Apparently, you can solve L5 by just claiming that you did and it was simple too, without showing any proof. LOL.
I maintain that there is no quick, easy solution to autonomous driving. Achieving safe, reliable, mass market autonomy will require redundancy, robustness, and lots of thorough validation. There is no magic pill.
Last edited:
willow_hiller
Well-Known Member
The issue is that Tesla critics haven't changed their tune in decades. At first they said the Model S was a scam, and then they said the Model 3 was a scam, and now they're saying the Cybertruck is a scam.
In the face of constant criticisms saying Tesla will never accomplish something, and then watching them smash expectations, it's hard not to do the same with FSD.
There is not a single Tesla critic in this subforum. There are tesla fans and ADAS/AV fans period.
This is why you would never find a post (maybe 1) of me talking about tesla in general (outside of ADAS/AV context) ever in this forum.
I do remember saying something once and i pointed out this is the only time I have ever commented on Tesla outside of AV/ADAS context. The same is the case on reddit, twitter, etc.
The problem is that superfans like you consider anyone pointing out reality and Truth to be a critic and Tesla hater.
Heck today you still believe that Tesla fans were and are right about FSD. After 8 years without Level 5 FSD materializing after 30-50 statements of Elon saying it will show up imminently and Tesla fans swearing it to be so.
Do you realize that Tesla can release Level 5 that's objectively and statistically better than humans tomorrow and I would still be right?
This is something Tesla fans can't wrap their heads around. Reality. If Tesla released it in 2030 (15 years after announcement and 13 years after first date). Tesla fans would still claim that I'm wrong and they were right. After they have claimed every year would be the year. We call that delusion. This is why they will argue vehemently against radar one day and the next day after Tesla adds it claim Tesla surpassed everyone because of it's radar. There's no logic or reasoning involved.
willow_hiller
Well-Known Member
Have you considered a career in stand-up?
Seriously though, "Everyone who agrees with me is balanced, nuanced and fair, while everyone who disagrees with me is a cult member" is not as great of an argument as you think it is.
Knightshade
Well-Known Member
God damn it, this is like the third irony meter of mine you've broken this month!
diplomat33
Average guy who loves autonomous vehicles
I've always found this argument to be illogic and more based on emotion. Building EVs and solving FSD are two completely different engineering challenges that have no correlation at all. Doing one difficult thing does not automatically mean that you can do a completely different difficult thing. It does not prove that Tesla can't solve FSD but it does not prove that they can either. It is also a convenient argument because no matter how late Tesla is on FSD, eventually Tesla will solve FSD, so no matter what, you will argue that the critics claimed it would never happen and therefore they were wrong.
And if the criticism is that Tesla will never "solve FSD", that is ridiculous. Of course, Tesla (and others) will eventually solve FSD some day. It is a question of when, not if. If the criticism that Tesla will not solve FSD with the current hardware/software or that they will not solve FSD on Elon's timeline, those are fair criticisms. Maybe they will be wrong but they are fair.
I also think we need to define "solve FSD". What exactly are critics saying Tesla cannot do with FSD? Is the criticism that Tesla will never achieve a self-driving car? Will never achieve L4 at all? Or never achieve L4 on the current hardware? Will never achieve L5? Will never remove driver supervision? Will never offer driverless robotaxis? Will not even achieve reliable "hands off" L2 point to point? Those are very different goals. It makes a big difference what we are talking about. I've seen different definitions of "FSD" on this forum which makes it hard to agree on what Tesla has "solved".
Last edited:
TBH this all feels like an argument for the sake of arguing.
It’s like we've got to the point of essentially trying to define what constitutes the color white and what shade is acceptable before it comes grey.
Its more position defense now.
It’s like we've got to the point of essentially trying to define what constitutes the color white and what shade is acceptable before it comes grey.
Its more position defense now.
Not one? What about @Bladerskb?
There may even be a few others, some of which I have put on ignore due to their content-free posts. However that @Bladerskb person has many informative posts (and a few not so much). He really adds value to the conversation, IMO.
GSP
There's really no shortage of people and companies who doubt full transference of driving task ownership will ever occur. Talking about "solving FSD" doesn't really add value to anything as I think most people see this, which is robotaxis driving themselves around + generating $30k/year revenue + massive increases in vehicle utilization and all those other benefits. The only thing of consequence for consumers is the SAE Levels and ownership of the driving task, neither of which I've ever seen Tesla mention outside of private communications with the California DMV.
Last time Elon was asked about this on an earnings call he just rambled on about the March of 9s
Last time Elon was asked about this on an earnings call he just rambled on about the March of 9s
DanCar
Active Member
Theory: What Elon meant by breakthrough: He is talking about the perception that people don't believe the car is driving itself. The breakthrough is to remove the nags.
2nd theory: Elon is good at hype. He throws out a bone, and we eat it all up with wild theories. We have fallen into the Elon hype machine trap. Elon thanks BladerSKb for fueling the free marketing fire. haha
2nd theory: Elon is good at hype. He throws out a bone, and we eat it all up with wild theories. We have fallen into the Elon hype machine trap. Elon thanks BladerSKb for fueling the free marketing fire. haha
Last edited:
Knightshade
Well-Known Member
But it fundamentally is NOT driving the car "by itself"
The OEDR is insufficient for L3, per Teslas own statements to CA DMV, thus the human remains driving and responsible for the overall DDT.
The OEDR is insufficient for L3, per Teslas own statements to CA DMV, thus the human remains driving and responsible for the overall DDT.
DanCar
Active Member
How does the human remain driving if they aren't controlling the car? Who is driving, the one controlling the car the one that isn't controlling the car? Yes the human is responsible and can and should take control when something is sus.
diplomat33
Average guy who loves autonomous vehicles
Driving is not just the act of steering, pressing the accelerator or pressing the brakes. Driving is the entire spectrum of watching the road for possible risks, planning actions and then applying steering/accelerator/braking appropriately. With L2, it can do some steering/accelerator/braking tasks but it cannot do the entire spectrum of driving from watching for all risks, planning all actions. For example, L2 might do lane keeping and cruise control but maybe it cannot respond to debris on the road or a fire truck intruding into the lane. So L2 still needs a human driver to watch for risks and handle any driving tasks that the L2 cannot do. Since the human is watching for risks, planning, and taking over as needed when the L2 cannot handle something, the human is still driving even if they go a long time without touching the controls. So L2 can do the steering/braking but is not driving since it is not watching for all risks, planning for all events and objects. That's the difference with L4/L5. L4/L5 can do all driving tasks. So L4/L5 is driving. It can do the entire spectrum of all driving tasks, from watching for all objects and events, watching for risks, planning driving decisions and doing steering/braking as needed. So it does not need a human driver.
Last edited:
Similar threads
