Tesla’s FSD Beta 10.3 Coming This Friday 10-22-2021

[Smh2210]

My wishlist for 10.3 is:

• Ability to recognize flashing yellow arrows and proceed through when safe
• Moving into the turn lane quicker across multiple lanes
• Toning down the speeds into turns

Has anyone received 10.3?

No

 

[PianoAl]

I'm thinking Tesla will catch on to the opt out/opt in hack pretty soon.

 

[jabloomf1230]

Bruh I drove 1400 miles last weekend to try to get to 100 and only got to 99

 

[BCNY]

Good morning! ANOTHER DAY, ANOTHER LIE!

Can I get my money back for not even having a beta version of the product I bought, even a couple of years later????

 

[vogtaction]

Fsd 10.3 beta download status ? Did anyone get it yet ?

nope, no update from 10.2 to 10.3 for me.

 

[M3Denver]

simply put..
what the FSD? where's my 10.3!?

 

[rubyrube]

I have a 99 in New York City and still no beta 10.3 for me in my MYP.

 

[DenisC]

Better than 98. I did the same thing only to read about opt out/in, although I haven’t figured out how to do that.

 

[Mojito]

Nothing in Orlando as of 9:41 am

 

[LowlyOilBurner]

I'm thinking Tesla will catch on to the opt out/opt in hack pretty soon.

I’m thinking the same thing. I was thinking about doing this cute trick, but then I was thinking: I’ve driven 2,6XXmiles since 9.25.21 to get to a 99, why risk possibly throwing all that away not to get “it” when 99’s are theoretically next in line? I’ll wait it out. What a few more days/weeks/months/years?

 

[Usain]

Well, we know it's not capable TODAY because the current code needs to borrow significant compute from node B to run at all.

I know a few people think Tesla can somehow magically add a bunch more, better, capabilities (which from the current beta it clearly still needs) yet somehow drastically reduce the compute needed to do it... but when you ask how they just waive their hands and say optimization.... which seems like pretty magical thinking to me.

Tesla thought HW2 had enough compute for L4. Then they thought 2.5 did. Then they thought 3 did. Ultimately until they actually solve it, they don't know how much compute they need. It's possible even HW4 isn't enough, but we have no real info on that at all. We do have a fair bit on 3 and the vast preponderance of evidence is that it's insufficient.

The ideal case for Tesla is if they can get "perfect" L2 out of HW3 using both nodes... because we know HW4 is about 3x the compute, so that would mean they could run the "perfect" code in a single node with room to spare on HW4 and that could then support L4 since it could run redundant code on both nodes. But again from the current state of the beta, they're not there yet.[/B]

I don't think it's magical thinking to believe that the car's NN can be optimized. They probably haven't tried very hard as they've been more focused on just getting it to work. And truthfully, I don't see Tesla spending a lot of time on optimization for HW3 as they would rather spend that energy on migrating to HW4. So it's probably a moot point.

I agree with most of what you are saying, especially the part about how we really don't know what it will take to get to L4. All we know is that Telsa appears to be on a path of continuous improvement until they either hit L4 or another local maximum. And we don't know which will happen.

 

[EV3_]

Elon:
Regression in some left turns at traffic lights found by internal QA in 10.3. Fix in work, probably releasing tomorrow.

 

[tdiggity]

Elon:
Regression in some left turns at traffic lights found by internal QA in 10.3. Fix in work, probably releasing tomorrow.

So, Monday. Maybe.

 

[n.one.one]

 

[DenisC]

So, Monday. Maybe.

Thanks for posting.

 

[B1081358]

No 10.3 in Michigan Go Green!

 

[n.one.one]

 

[dan3413]

I’m thinking the same thing. I was thinking about doing this cute trick, but then I was thinking: I’ve driven 2,6XXmiles since 9.25.21 to get to a 99, why risk possibly throwing all that away not to get “it” when 99’s are theoretically next in line? I’ll wait it out. What a few more days/weeks/months/years?

That’s my thought too. Let’s be honest folks they put the safety score system in place for a reason… Even if it does have some bugs in it (eg someone cuts you off etc). So if you have a low score like someone said they were at 74 - then drive more miles and pay closer attention.

I used this app here to input my daily driving miles and let it calculate how much more driving I had to do at 100 or sometimes lower to get to the desired score of 99

Safety Score Calculator

safety.sfnet.app

Also remember the app is a rolling 30 days … so in a few days (Oct 25) the older dates will roll off. I’m not cheating the system they setup for a very good reason.

 

[JADCa]

Can anyone confirm the 30 day rolling average? My first day was an 84, but nothing under 97 since so I am at 98. Easy 99 30 day average with real driving so I would rather not cheat the system.

Found it. The Safety Score Beta is the first release of the Safety Score which is an assessment of your driving behavior based on five metrics called Safety Factors. These are combined to estimate the likelihood that your driving could result in a future collision. We combine your daily Safety Scores (up to 30 days) to calculate the aggregated Safety Score, displayed on the main ‘Safety Score’ screen of the Tesla app. You can find details around your daily Safety Score by selecting ‘Daily Details’ at the bottom of the screen.

 

[Jarvis20]

There you go