They don’t need a new CEO for that. Sounds like you’re describing a COO.
-
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
The worst companies are the ones that give you lots of uninteresting options. Just go look at your typical car company with all their options that exist only to hide the fact their main product sucks. I don’t need Tesla to provide lots of options, I need them to innovate and provide great products.
Lowchrg
Beta Blue-balled
Using my forum registration date as an argument against me sure is petty. I’ve been an owner since 2017 and never had an interest in the forums until FSD Beta was released.
If we’re gonna be that petty then check the like/dislike ratio on the comment. Still in my favor. It’s not just me.
I don’t care about 400000000% stock returns, or if Elon is going to space, or any of that noise. Their service and communication is currently awful, objectively. The hype driven by Elon continues to smokescreen the deeper issues with Tesla that are not being resolved. I’d love a CEO that stayed off Twitter and actively worked to make the company better instead of trying to sell the next thing (this week: the new Roadster preorders opening again, but no resolution for vibrating Plaids).
Anyways I’m sure I’m not making any friends right now but unless there are some major changes to the company I don’t foresee myself buying another Tesla. Again, I’m definitely not alone there.
Have a good one
Thanks for proving my point. If Musk’s only comparable CEO is Steve Jobs then he’s pretty special.
Easy for new brain to do better. Tell Elon: AP hardware 3 is not enough. AP hardware 4 is not enough. Stop fleecing its customers. Start experimenting with better hardware.
I agree their service and communication are beyond terrible. However because of Elon their products are second to none. I really need them to fix their issues but at the end of the day I spend 99.9% of my time with the product and 0.1% dealing with the company. I’ll take that trade off every time.
IMO they don’t need a new CEO to fix the issues, they just need a strong COO that Elon respects.
Knightshade
Well-Known Member
I'm glad you already know HW4, not even in cars yet, isn't enough. How much, specifically, is?
Remind me again, because I've asked before, where can we buy your self-driving software again?
Knightshade
Well-Known Member
Nope.
Read the rest of the posts.
The poster later admits he actually opted out of the beta and what he was seeing was the original radar-back-on 90 mph limit.
wknickless
Member
Exactly right - someone like Gwynne Shotwell over at SpaceX.
Unless they start changing cameras and placement, no amount of compute is going to going to solve the fact that the car just can’t see at certain intersections or in certain conditions (sun at specific angles/rain/etc…)
I think it’s safe to say if HW4 is just another compute upgrade, it won’t be enough.
Yes like Waymo or Cruise that have many lidars and radars but doesn't know that the car had no light
DanCar
Active Member
According to rumors HW4 also contains new parking cameras, which enables birds eye view. But yeah, FSD continues to be a much bigger problem than people think, including Elon. Elon continues to make the same mistake since 2015. Hopefully Tesla can squeeze level 3 out of hardware 4. Will be great when it can handle potholes that elon said several years ago, was coming.
Knightshade
Well-Known Member
FWIW I don't believe they can get to L3 or better on city streets without more (or at least better placed) cameras.
I DO think they can do L2 city good enough for wide release- and L3 or L4 -highway only- with current camera #/locations though... you don't need to see around tight obstructed corners, nor see below the hood-line, on highways.
As how how much compute they need for any of those-- nobody knows. Including the dude who insisted HW4 wasn't enough without even knowing how much HW4 has.
Knightshade
Well-Known Member
No, he did not.
At autonomy day someone asked what the primary design goal was- he thought about it, someone can be heard whispering to him "safety" and he then says "it'll be at least 3x better than the current system"
He has since reiterated that he was clearly talking safety, not raw compute, on that point.
For example when he more recently said "But it’s really just gonna be like, can we take it from, say, for argument’s sake, 300% safer than a person to 1,000% safer"
Which is, indeed, a little better than 3x safety improvement- not raw compute.
Though even if we pretended he DID say what you misheard- what evidence do you have 3x compute wouldn't be sufficient?
How much compute, specifically, IS enough- and how do you know?
Elon stated that hardware redundancy is crucial. They are using both nodes of hardware 3. If we assume that hardware four is 3x more capable compute wise then that means it is 50% more capable when compared against two nodes of hw3. You need at least 100% more capable to create true hardware redundancy.
On a different train of thought: if one follows advancements in machine learning, all of the big progress has been made by larger models. So large models are key to recent improvements. These models are orders of magnitude bigger than hardware 3 or 4. That will give you an idea of what is needed, just to meet state of the art.
On a third train of thought: No model is perfect. There is no foreseeable change to this. To make the system fault tolerant, need several different models running in parallel checking each other. Tesla has not indicated they are doing this.
On a fourth train of thought: Fundamental failures are evident in FSD. For example: no solutions proposed when sun shining into camera. So unless there are significant upgrades to the cameras, hw4 will be insufficient. There are low cost solutions for this. There are image sensors that can block out the sun, but haven't seen indications that Tesla will employ this tech.
On a different train of thought: if one follows advancements in machine learning, all of the big progress has been made by larger models. So large models are key to recent improvements. These models are orders of magnitude bigger than hardware 3 or 4. That will give you an idea of what is needed, just to meet state of the art.
On a third train of thought: No model is perfect. There is no foreseeable change to this. To make the system fault tolerant, need several different models running in parallel checking each other. Tesla has not indicated they are doing this.
On a fourth train of thought: Fundamental failures are evident in FSD. For example: no solutions proposed when sun shining into camera. So unless there are significant upgrades to the cameras, hw4 will be insufficient. There are low cost solutions for this. There are image sensors that can block out the sun, but haven't seen indications that Tesla will employ this tech.
Last edited:
Knightshade
Well-Known Member
But there's no basis for that assumption. His 3x statement was about SAFETY, not compute.
As I JUST pointed out in the last reply.
So we can easily ignore everything you posted after that.
Which was a really long post to once again refuse to answer the actual question you've been asked 3 times now.
How much compute, specifically, IS enough- and how do you know?
And where can we buy the self-driving SW that you wrote again?
You did ask “ Though even if we pretended he DID say what you misheard- what evidence do you have 3x compute wouldn't be sufficient?”
I think he was responding to this
Similar threads
