Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Tesla autopilot HW3

This site may earn commission on affiliate links.
given that it runs same NNs today - that's kind of expected.

The big question now is how long until Tesla deploys new NNs that take advantage of HW3. On one hand I'm setting expectations low because of the HW2 debacle where cars shipped in 2016, only recently got features advertised as Enhanced Autopilot (NOA). On the other hand, one could discount some of that because HW2 was rushed because of the breakup with MobileEye. They have had Autopilot 3 hardware on the test bench and in cars for a while, so hopefully they have the development efforts further along and it won't be another year or 2 until we actually see software shipped to leverage HW3.

Presumably they will be demoing a HW3 car with new NN running at the upcoming Autonomy Investor's day. If we don't see it in some form there, then my expectations for them shipping the new NNs is atleast a year out, perhaps 2.
 
  • Informative
  • Like
Reactions: rabar10 and OPRCE
Could we see higher resolution images and video for the dashcam and sentry mode with HW3 due to better performance? Thought I've read that the cameras are higher res, but the AP2.5 computer can't process at a higher resolution because of cpu limitations.
 
  • Like
Reactions: OPRCE
Presumably they will be demoing a HW3 car with new NN running at the upcoming Autonomy Investor's day. If we don't see it in some form there, then my expectations for them shipping the new NNs is atleast a year out, perhaps 2.

I think there is a 99% chance that the April 22 event will demo HW3 and the new NN. After all, that's what the entire event is about: to show off what FSD on HW3 can do. Not showing HW3 in action would be an epic fail.
 
I think there is a 99% chance that the April 22 event will demo HW3 and the new NN. After all, that's what the entire event is about: to show off what FSD on HW3 can do. Not showing HW3 in action would be an epic fail.

question is, is it gonna be hands free and done in the day to allow recording and show the experience from a ride hailing riders POV? OR is it gonna be something done in the dark with alot of hand waving to hide failures?

 
Last edited:
Could we see higher resolution images and video for the dashcam and sentry mode with HW3 due to better performance
no.

the cameras are 1270x960 - so what you get from the dashcam is as good as it gets, no higher resolution. It's just hww.5 downsamples it to 1/4 of the resolution (1/2 linear) before feeding into NNs.
 
question is, is it gonna be hands free and done in the day to allow recording and show the experience from a ride hailing riders POV? OR is it gonna be something done in the dark with alot of hand waving to hide failures?

We know its going to be done in the day as it was scheduled for the morning. I say was as they rescheduled it for the 22nd, but they didn't give a time. I assume it would be in the morning though.

We know it's also going to involve test drives.

As to hand waving? When does a demonstration of new features ever not have a little of that? So I do expect a fair amount of "ignore that. We'll get that figured out".
 
If it is a demo what we see in April 22, it will belong to the "demo" universe and cannot be compared to real released products. Especially, if it is a pre-recorded demo on a geofenced area.

Demos from Tesla, Mobileye, NVidia, etc are interesting to see what are they developing and what can we expect in the future.

Now finally we will be able to compare recent demos from all major manufacturers (Tesla was a bit behind as his latest demo was from 2016).
 
I would give Tesla a lot of slack had they not pre-sold me Level 5 capable hardware in 2016, Tesla Network details in 2017, coast to coast summon and whatnot — and dissing on those using Lidar etc in the process. But because they have done all that, the perspective changes on what is expected of Tesla.

They set this bar themselves and a lot of people believed them for a long time.
I think at some poit the Level 5 systems will require a hardware update to fully function and not just software. Any comments of this?
 
I think at some poit the Level 5 systems will require a hardware update to fully function and not just software. Any comments of this?

I can see you're very new here, but many people here have been saying this for a very long time... since the AP2 announcement in 2016 in fact. I agree 100%, the current hardware (including HW3) is not ever going to be L5 autonomous. There are many posts in many threads discussing this, and my own personal heavily biased summary is this:

On Tesla's side:
Fanboys with very short memories who believe Elon can work magic, and possibly that HW3 will be powered by unicorn farts.

On the other side:
Every expert in the field that is not affiliated with Tesla*, and many well-reasoned arguments.

*And even Tesla has precious few experts in the field who are still willing to be affiliated with them.
 
On Tesla's side:
Fanboys with very short memories who believe Elon can work magic, and possibly that HW3 will be powered by unicorn farts.

For the record, I do not believe that AP3 will give us L5 autonomy.

But maybe if Tesla fanboys believe Musk "can do magic" as you put it, is because we've seen Tesla repeatedly beat the odds and defy these experts time and time again. These experts who said Tesla would never succeed with the Model S, would never succeed with the Model X, would definitely never mass produce the Model 3, would never sell enough to make a profit etc... All wrong. And when Tesla and Mobileye broke up, the experts said that Tesla's AP was done. Certainly, for a long time, that seemed to be true. Yet, we've seen Tesla come back and now EAP is starting to surpass AP1 and put out some really good features like NOA with AP3 coming soon. So yeah, Tesla fanboys will think "if Tesla can beat the odds before, maybe they can beat the odds again."

There is another big factor too. Don't underestimate that Tesla fanboys see Tesla as the underdog fighting against the evil legacy auto maker/Big Oil conglomerate. People love rooting for the underdog against the supervillain. So yeah, Tesla fanboys are going to root for Tesla because they hate the groups attacking Tesla. They are certainly not going to give any credence to experts who are against electric cars and want us to continue driving dirty gas cars.
 
  • Like
Reactions: Richard34212
For the record, I do not believe that AP3 will give us L5 autonomy.

Great, then you agree that Elon is a big fat liar head, right? A fraud who lied and continues to lie to this day in order to sell cars? Because he is still out there talking about our cars turning into robo-taxis and generating revenue for us.

Somehow I suspect you are not going to agree with that statement... but you have to reconcile his statements with reality somehow.

But maybe if Tesla fanboys believe Musk "can do magic" as you put it, is because we've seen Tesla repeatedly beat the odds and defy these experts time and time again. These experts who said Tesla would never succeed with the Model S, would never succeed with the Model X, would definitely never mass produce the Model 3, would never sell enough to make a profit etc... All wrong.

These are very different experts, and very different situations. What they did with EVs in general was a classic disruption of an established industry and we saw a classic response -- i.e., coordinated pushback from established players -- against it. And there were many knowledgeable people on both sides of that conversation -- it wasn't just Elon vs. everybody who know what they were talking about. It was Elon and a bunch of people who knew what they were talking about vs a bunch of other people who knew what they were talking about.

But AVs are different from EVs. AVs do not exist yet in any form. They've never been done. AVs will in fact be a disruption to all sorts of things, and there are a few established players pushing back against the AV industry, but mostly the reaction to the AV industry as a whole is more of a deer-in-headlights thing without a lot of specific action. I'm talking about Tesla's approach in particular, which gets significant pushback and absolutely no support from anybody knowledgeable about AVs, even those who are very bullish on AVs in general. Tesla is standing completely on their own. This is very, very unlike EVs, where they had all the EV enthusiasts and experts in their corner as they fought the entrenched oil and auto interests.

Back to EVs, it was very clear to anybody with half a brain (that was not clouded by bias) that EV technology was well within reach and achievable in a reasonable timeframe. Elon's timeframe was seen as aggressive and in fact for the most part he didn't meet his own timeframe goals even on EVs, but at least it was clear that what he was trying to do was well within the realm of possibility. People had actually made EVs. They were just too expensive.

Conversely, nobody has ever succeeded at making an AV, at any cost. And people have been trying for a long time -- a lot longer than Elon has been trying. And yet, Elon thinks that not only can Tesla make an AV, but they already know exactly what hardware will be required to do it. And -- good news! -- that hardware happens to be inexpensive and commonly available to put in cars right now today, and without making them look like ugly laboratory projects! Isn't that convenient! Maybe, just maybe, a little too convenient, don't you think? Wouldn't you like to, I don't know, have even a single knowledgeable, reputable, independent expert validating this idea? Just one?
 
  • Disagree
  • Like
Reactions: jepicken and OPRCE
Indeed. This is not about 2016. Elon doubled down on autonomous ride sharing on Teslas sold today just days ago.

How does one reconcile that and not believing in at least Level 4 capability... one can’t. Do you believe @diplomat33 Teslas of today will reach Level 4?

The fact that Elon did double down on autonomous ride sharing again would suggest to me that he does believe it in his mind. Presumably, the progress with AP3 has made him confident enough that it is going to happen. Whether he is right or not, is a different matter.

There are several possibilities:
1) AP3 progress is lackluster and Musk is outright lying to help the company.
2) AP3 + the new NN really are amazing. Maybe he witnessed a demo with AP3 where the car really did do L4 autonomous driving on a particular route?
3) AP3 is good but maybe not THAT good but Musk is extrapolating that ride sharing is still possible if they just keep at it and push harder. Maybe it's an important goal that he has for the company so he is pushing hard to try to make it happen even though it is not really realistic. Musk does come across to me as someone who believes his own goals and believes in pushing the limits.

Personally, I go with #3 as the most likely. I can't say #1 because I don't know that he is outright lying. That implies motivations that I am not privy to. #2 seems very unlikely based on what we know. So #3 is left by process of elimination and #3 is the most realistic of the 3 options IMO.

To answer your question, I think AP3 will probably do L3 autonomy. I think L4 is a huge stretch on the current hardware. Although the increased processing power of AP3 should help somewhat. Ideally, I would love to see Tesla's get a rear radar, 360 degree LIDAR and a better computer than AP3. That would complement the existing cameras nicely. If Tesla's got that added to the existing hardware, I would be a lot more confident in L4.

If I had to guess, I would say that Tesla will probably get some cool L3 features with AP3 but they won't make it to L4. Musk will stubbornly try to make it happen for a couple more years until he eventually caves and goes the full suite like I mentioned above (cameras, radar and LIDAR and more processing power).

Frankly, we just need to wait for April 22. We should get some answers one way or the other.

Great, then you agree that Elon is a big fat liar head, right? A fraud who lied and continues to lie to this day in order to sell cars? Because he is still out there talking about our cars turning into robo-taxis and generating revenue for us.

Somehow I suspect you are not going to agree with that statement... but you have to reconcile his statements with reality somehow.

No because I don't like throwing around such serious accusations like calling someone a fraud and a liar, especially when I have never met Elon Musk. Furthermore, being a fraudulent liar implies malevolent and criminal intent which goes well beyond just the fact that the statements are wrong. I am just not going to go there. I can judge if his statements are factual or not, yes, but I don't have the right or the ability to make a judgment about the secret motives or intent in his mind.

I reconcile his statements with reality this way: He can be wrong without being a fraud or a liar. Maybe he really does believe that he is right and can't see the other side of the argument? Maybe he is in denial about the hardware not being good enough so he is pushing hard to make FSD happen on the current hardware? Those would be my guesses.

Frankly, I don't really care if Musk's promises of L5 robo-taxis are completely off. I love my Model 3 and I love Musk's vision of the future (electric cars, renewable energy, space exploration). It's the future I want. And the legacy auto makers all suck! They have no interest in that future. So I am willing to support Tesla despite their mistakes because I support the direction that they are moving in.