Interview with FSD Skeptic Dan O'Dowd | Tesla Motors Club Podcast #53
01:56:03 Share

Hey there, welcome to another Tesla Motors Club podcast. My name is Lewis. I'm Doug. And I'm Mike. On today's episode, we have a very special guest. Dan O'Dowd is going to be joining us from the Dawn Project. Many of you may know him as an FSD critic, but I'm excited to talk to him about Tesla, FSD, and other things, technology. And we're really excited to have him here. So why don't we just get the show started? Because, you know, that's kind of a big deal.

episode 53 starts now 53 go all righty that was a little crazy you got how you guys doing it's a monday night we don't normally do them on mondays it is very discombobulated i'm a little out of it that's true it's already dark outside i'm on the east coast i'm like oh you're doing this now

Yeah, I feel you. I say we jump in. I just want to talk to Dan about all kinds of things. He's got a really interesting perspective. I know those in the chat, if you have any questions, feel free to post them and we'll see what we can do. I'm not going to ask him, you know, ridiculous things, so keep it civil. But let's introduce Dan. There he is. Welcome to the show, Dan. So great to have you here. Hello. Good to be here.

I just wanted to, for those that don't know, because not everyone in our audience is familiar with the Dawn Project, could you maybe give a little bit of background or a little explanation of the Dawn Project, what it is, why you started it, that kind of stuff? Sure. Well, it goes back a lot longer than the Dawn Project, but basically, I've been working for almost 50 years on basically the problem of how do we make software that's safe

for systems that people's lives depend on. And that covers a lot of different ground. How do you make software that never fails, that it always works? When thousands or even millions of lives are on the line, it has to work all the time. It can't work most of the time. You can't reboot it from time to time when it does the wrong thing.

And then there's also the security side, which is if you build something that people's lives depend on and somebody finds a way to hack into it, they can do some damage. They can kill people or cause serious dislocations and problems. And

This wasn't a problem 10, 12 years ago. This wasn't even a problem. Problem for the future. But now we're in the future. People are connecting everything together and we're running everything with software. The most important critical systems now have software running in them. And then when they hook them up to the Internet.

Now you ask 99.9% of the people, can you build something connected to the internet that can't be hacked? They'll say no, impossible. And yet here we are doing it every day. We're connecting up the power grid, water treatment plants, cars, every, you know, trains, every mode of transportation, everything we depend on.

is being run with software and most of it's been connected up to the internet and is pretty easily hackable because people just downloaded software off the internet that you can go to a not even black market, gray market place and buy vulnerabilities. You can literally buy a vulnerability in a major product and then that machine or anything that runs that software is completely vulnerable. You can get in at any time you want and do anything you want to it.

So that's really what's been happening. I've been working for most of my career doing military work and building the core security system for military systems.

And that is the most important thing. So nuclear weapon systems. I've done the operating system for the B-1B, B-52, the B-2 bomber, the F-35 fighter, the F-16, F-22, Euro fighter, done Airbus A380, Boeing 787. We're doing all the FBI laptops for FBI agents in the field.

Just large numbers of these are the systems that cannot fail. And there is a very small, tight-knit group of people who build this software. And we have a completely different set of principles of developing software than is used in any other systems.

The aircraft is the one that's sort of intermediate in terms of a very well-defined process and lots of testing and lots of design and lots of analysis and lots of review to keep it from failing while the airplane is flying and that would be catastrophic. So that's been my background, working on all these systems, building the core security system for all these military systems.

But of the last 10 years has been this mechanism of turning everything over to software and everything onto the Internet. And all of a sudden we come out and it's now the civilian systems, which are the greatest weapons on the planet. If you could gain control over the Tesla, I can't remember what it was a year or two, a couple of years ago. Elon Musk was at a conference and somebody asked him, what about security? What worries you? And he says, well, a fleet wide attack would really be a problem.

somebody gained control of all of the Teslas at the same time. Well, they all run the same software, so you find out a way to get into one. You pretty much found a way to get into all of them. And they're all on the internet, so you can connect up to all of them and send them commands. He said, well, he happened to be at a governor's conference in Rhode Island. He said, Rhode Island. He said, well, somebody could send all the Teslas to Rhode Island. Oh, my God. That would probably be the end of Tesla if somebody did that. I've always sort of wondered whether

He missed the point or did he carefully deflect us? Because that's not what somebody could do if they controlled all the Teslas. That would be the worst thing, driving them to all the Rhode Island and create a traffic jam. Somebody gained control of all the Tesla, they could accelerate them to 100 miles an hour and drive the wrong way down the street. All three or four million of them at the same time. That would be catastrophic.

And yet nobody seems to be very worried about it. Hopefully he's really talking about that as an issue. Dan, realistically, if the car got a command to drive 100 miles an hour or whatever, the driver could step on the brake and control the steering, right? No, no. The brake is just a computer game input.

If the software decides to ignore the break, it can do what it wants. But the break is hydraulic, right? Yeah. So there's a hardware. I mean, you can't be told to ignore the break. I mean, it can add the break. It can send a signal to depress the break, but there's a break booster. It's a mechanical connection. These days, this is not mechanical anymore. No, it is mechanical. I mean, you can software command

the mechanical brake to stop, but you can't make it non-stop if someone's stepping on the brake. If you're in autopilot and it is braking, you can feel the brake pedal get loose below your foot because it's actually mechanically being depressive. But if you step on the brake, the car stops. Same with the steering. It's still a rack and pinion steer. There is an electrical assist, but you still have a shaft going to a

that then adjusts the direction of the wheels, right? So there's no, there's still a mechanical override, right? The car isn't going to go. That I know of, but maybe you're right. I'm not, I don't know about Teslas, but other cars have gone past that point and they are,

It's just a, it's a computer game input. It's not a mechanical connection. It's driving the car when there's no one. There has been talk of moving to steer by wire is what you're talking about, but as it is right now, you mechanically are in control. So if you got some command to veer to the left, you can

You can hold on the steering wheel. That's part of the reason the steering wheel is turning as the car is driving itself, because it's all mechanically connected there. But to be fair to Dan's point, this is something that Tesla and other companies have looked at doing, and it is kind of the direction they're trying to go, right? Like in the grand scheme of things, that's kind of the whole robo taxi concept that Tesla wants to do is effectively...

eliminate you steer by wire and then you know if there aren't protections in place um somebody could take control and cause problems right with tesla i don't think it's because they want robots i i think with tesla it's the same thing they've done for years with the model 3 in particular it's cost reduction if i can read if i can pull out the reservoir the brake fluid the piping all of that and replace it with a mcu and some wires it's cheaper

Now, Dan's got a point, though, that if you do that, now you've opened up a vector for a threat. It's really hard to hack a mechanical contrivance.

It's not so hard to hack something that's computer controlled. And we keep doing that. We've taken the governors off of generators. There used to be a hardware governor and now we write software governors and people hack the software governors on some of the power equipment. And you just run it up past red line until it breaks down, catches on fire or fan blades go crazy.

So Dan, I've got a question for you regarding public acceptance of threats and threat analysis, like you've just said. So you say that your car or some device can be hacked and by some threat actor, and it's going to be a bad day. I actually wrote a piece about this back in 2012 and the publisher refused to publish it because it was considered inflammatory and fear mongering. Is that still the case? Is that something you run into when you try to talk to people about this?

Well, yes. I mean, the public sort of in general gets it and is kind of scared about it. The problem is that the companies don't. The companies say, if we conform to the government standards, whatever those are, we see no reason to do any better. All that is spending money for no purpose. And that's very deeply ingrained. I've spoken to lots and lots of people and you point out somebody could do this and they basically figure out

you know, risk of doing business. I can't afford to make my operation inefficient. I can't afford to spend more time and energy making it secure because no one even really knows what that means. No one knows how to make it secure. And so people just put it on the bottom of the level and say, you know, we'll throw in some security software that we downloaded off the internet and that'll make it secure. So let's keep with Tesla, for example. Tesla's got something that I think is rather interesting

well put and well designed for their car. And as they do over there are updates and they're really quite good at it. Unlike other cars where they try to do it over there update and they either bricks the car or they can't do it at all, or you have to bring it into the dealer. So I would think in Tesla's credit that they're saying, well, okay, we find a threat. We find a hole, we find a security flaw. We can send you a patch pretty rapidly and take care of that.

you know i mean i've been on the assembly lines where the where the controllers were 15 years out of date and they wouldn't update it and i don't see tesla being in that same position i see them a lot more current and and they have the ability to stay current well everybody's doing off over their updates now i mean that's standard for car companies at least ones i've worked with um that that's that's a pretty standard thing but

But the fact that you keep having to patch it means those problems were there before. And if somebody knew about them, they could be exploiting them anytime for years and years and years. People say, oh, it's secure now because we know of no vulnerability. We've patched all the vulnerabilities. But look back a year from now and you'll see what were all the vulnerabilities that were discovered and patched over that year.

Today, all those problems are in the product right now. There are dozens or hundreds of vulnerabilities in there that somebody may know about. You don't know. A hacker doesn't report it to Microsoft or to somebody. They'll say, oh, please fix this bug. They cherish it and use it or they sell it. And there is an entire market of buying vulnerabilities that you pay $100,000, $200,000 for.

And you can hack into Windows, you could hack into Linux. A little more to get into Apple these days, but you go to NSO Group in Israel and they can hack anybody's iPhone and turn on the microphones and the cameras all the time. And you have no idea that's happened. And it's been done again and again and again to those systems.

sure it's definitely you bring up a lot of good points on like the broader problem in technology in general right like people bring amazon alexis and other things in their house and now they've got a microphone there we've got everything all your appliances are on the internet um a lot of those things though as much as they're terrible for privacy they aren't really a risk of harming a person generally but there are systems where it could cause harm such as like if your car breaks you

I was thinking about what you said earlier, and I agree with Doug. Currently, Tesla's mechanically, you can't force to accelerate over the driver, but you could forcibly break the car, right? You could say, oh, they're going 80 miles an hour. I'm going to slam on the brakes.

which could be deadly and harmful and things like that. So there are definitely vulnerabilities out there. Do you specifically worry about cars more so than like power plants and energy grids and nuclear plants, all that kind of stuff? Or do you think it's, is this, is Tesla basically an avenue to get your broader message out? Or you specifically, are you more worried about full self-driving, autonomous driving, those types of things?

The two biggest threats that I see in terms of catastrophic results are the power grid and self-driving cars. They're not just that they can cause great catastrophe, but that they're so easily programmed. They're weapons of mass destruction. You could just do it in Philadelphia. I don't know how many people would die, but you don't have to...

to destroy everything. Same thing for power grid. If you can bring down a power grid, not bring it down, but destroy it by over charting, running it past its maximums until the generators are destroyed. If you destroyed all the generators in Detroit, we would have to evacuate Detroit for a year at least before you could get new generators. You can't just order a new generator overnight and have it shipped by Amazon.

And so you'd be out of power. No one would have any power. You'd have to go to Cleveland or you'd have to go to Chicago. You took out one city. Same thing for the self-driving cars. You could take out one city at a time. Now, to make a long story short, once upon a time, Harry Truman played this game. He dropped a bomb on Hiroshima, wiped out the entire city and the entire population. Then he dropped another one on Nagasaki and wiped that one out. And then the Japanese surrendered because...

It's the perfect weapon. You just destroy one city and then destroy another city and destroy another city and the other guy has to, if they don't surrender their entire nation is destroyed and you know, a few months if they get around to destroying everything, you give up, you could you could conquer the United States of America intact. If you could

take down the power grid selectively in places or send all the electric cars or all the Teslas going crazy in a given city, you'd kill hundreds of thousands of people. And you could just keep doing it. It's basically, I'm going to keep doing this until you surrender. There's really no choice.

Those are the ones that really have to stop because they're targetable weapon systems. They're not just catastrophic weapon systems. You can take out one person. You just want to assassinate somebody, just take over their car, drive it into a bridge. It's easy. Take out a city, take out any group you want. Just go online, buy a list of people's names and their VINs and kill the ones you want to kill. Okay.

This building this system, this is insane. We're building this system and everybody's running as fast as they can to build it. And I'm saying we've got to slow down. If you want to put something like a car on the Internet, you really need to use software that nobody can have. I mean, certainly that's a possible vector, but I don't know that that's novel. Like if you wanted to target an individual, you could.

put bombs on drones or I mean that if that's what you're talking about and certainly I think the power grid is comparatively a much bigger thing than individual vehicles isn't it in terms of uh you're talking about a threat well individual vehicles as you say I mean somebody could get a gun and track the guy down they can probably you can probably shoot an ordinary citizen um but

Right. But I'm saying that's the point, is it from an assassination tool up to just wipe a country off the map. I mean, if you touch out all the power generators, imagine what would happen if all the power generators in the United States failed or were destroyed. You wouldn't even know. You'd be sitting at home and saying, well, the lights are out. It happens every now and then, and now they'll get it back online. Next day, you'd be, well, that's a problem.

I wonder when it's going to get fixed. Let's turn on the TV. Well, we don't have a TV. You don't have anything. Now, you know, maybe some guy down the street has solar panels on their roof and they can, you know, and they can get a signal out. Hey, San Francisco, are you there? Nope. You know, Philadelphia, you're not there either. Nobody's there. I can't find anybody. I can't get in touch with anybody. What's going on? Well, we could get in our car and we could drive to find out what's going on other places. Problem is that

What are you going to do? If you want to charge up, if you find somebody with some solar cells, you can't go that far. You wouldn't even know what happened. There's nobody coming to rescue you. You'll just be there. It will be days. It will be weeks. It will be months. Okay, you feel pretty good. You've got solar panels on your roof, so you're going to be good, even though there's a power failure. No, no, no. The Proud Boys will be there in a few weeks, and you'll be outside, and they'll have your house when they show up with their stuff.

Yeah, I mean, I think it's a fair argument to say that there are critical infrastructure things that are dangerous and probably shouldn't be on the internet. I guess I'm – what's funny is I generally consider myself somewhat pessimistic about a lot of things. But I would feel that in this case, I feel fairly comfortable that, one, it's extremely hard to –

to execute some kind of coordinated attack to that scale. And two, there are ways in which we can protect it. Now, while I would hope the software and things would be much better, and I fully support your fight for that, especially as an engineer, yes, if say, for example, another foreign government decided they wanted to take us out, which most foreign governments never would for many geopolitical reasons,

But say a bad actor like, you know, wanted to do it. I mean, we can actually cut the Internet connection to the US, right? Like we can act like government has the ability to say there's no external communication between like it's possible to do, but it would be very difficult to like to do all of it.

Interestingly, I did used to work in electrical systems and power distribution. So I agree with you, it would be fairly easy to take down power plants. And if you did, like our whole grid doesn't even have redundancies. There are many issues there. So it's kind of surprising that we haven't had those types of problems already. But I'm hopeful that things will improve. Some years ago earlier in my career, I worked in medical devices and I did work on software for medical stuff used in operating rooms on patients that was hooked up to the network.

in the hospital and I was kind of opposed to it. I was just like, I don't wanna do this, like those embedded systems. So, okay, you're not running Windows or something where it's easier to break in, but it's not impossible to break in. And yeah, I was definitely uncomfortable about it, but they're generally in those cases where we were regulated by FDA and other things, we had hardware interconnects, we had, you know, it's right there, someone can pull the plug or there are ways, there's buttons to disconnect. We have those kinds of things.

We don't really have those kind of things in cars and some of the other areas that you're talking about. Do you recommend that? Do you think that would be a fix of, hey, there's a button or a way to shut off my cell signal on my car or other types of things in case you felt like something was going wrong?

Yes, we've actually advocated such a thing at some length without any success that we could just order the car companies to put in a button on the dash or something. You can get panic button when you hit it, you're disconnected and you're back to driving yourself and you're not hooked up to the Internet.

And, you know, it might work. It would be a partial fix. It doesn't fix it because this happens very fast. If you put one of these things in place, it would all be over in minutes. And people are not going to see what's going on. Will they react? Maybe. But what do we need to do? The answer is really easy. We need to have software that you can't hack. It's pretty simple. Before you do it, you can't put your dangerous device on the Internet until you demonstrate, prove it can't be hacked.

And right now, I hear there's another SolarWinds problem. You know, like 90% of everybody had that single piece of software in their security system. Somebody got into their IT systems and programmed in a back door in and then let it get distributed to basically every company building a security device on the planet, just uncritically downloaded the new release, slapped it in and ran it. They didn't.

test it. They didn't even opposition test it. They didn't do anything. They just downloaded a piece of software from a company that had already been hacked once. Now, again, the processes are terrible. There are vastly superior processes that we follow when we build really important systems that millions of lives depend on, but nobody's following them in the commercial area. I have been advocating this. I've been working on this for years and years and made virtually no progress in

talking to people and just saying, we have to do this. This is crazy. They're like, I'm following the government regulations. That's all that's necessary. I'm curious, Dan, not to disagree with any of your points about software and attack vectors and all that, but I am curious how you relate this to the Tesla FSD. You seem to specifically call out FSD and you seem to be skeptical about that in particular.

But I don't hear much about other vendors driving software. Explain to me why you call out FSD and what the rationale behind that is. Well, I mean, one in general is that it is a system that is, you know, that is on the Internet and that is, you know, a software system that millions of people's lives depend on. So it sort of fits the criteria.

But the real reason is, or the fuller reason is, I know I get a lot of pushback on this, and I just love to debate it. We can right here. It is the most incompetent, stupidly developed software I have ever seen. This is an amazing piece of junk.

Nobody, the people who built the Yugo would hang their head in shame to put a piece of software out this bad. Now, I'm going to trigger everybody. Now, let's talk about it. I am appalled at the quality that

that I see this thing can't go 10 miles without running into something. Okay. I just want it said, I just want it said right now that everybody views me as the curmudgeon, but I think you've just taken the crown. I'm here and I'm telling you, I am serious. What company would put a product out? This is a product that your life depends on.

And it grabs the steering wheel, not infrequently, and turns it to the left or turns it to the right when there's a car there or there's a curb there or there's a truck there.

unbelievable you call it beta i mean they do call it a beta i mean it's not a beta okay that's good people say we get beta this product is not beta let me explain the engineering process to anybody who doesn't all right you build a product you prototype it in the lab you think it's working pretty well and you send it to alpha site alpha

which is your own employees take the product home or whatever and use it and tell you all the problems and tells you bugs and how it's hard to use and all that kind of stuff. That's the alpha testing process. You keep doing alpha testing until you can't find bugs anymore.

You keep doing it until the employees are saying, you know, this thing's working pretty well. As long as the employees are finding so many bugs that they overwhelm the developers, they can't fix them that fast, you stay in the alpha process of your own employees and not the outside world. When you finally get it to where the employees are saying, I like it, it's good, it doesn't break down on me all the time, then you send it to beta sites.

This product, you cannot use it for 10 minutes and not find problems. I will make a caveat. On the freeway, way better. But in town, in real driving, in real cities with stoplights and pedestrians and other cars and having to turn right and all those things. I mean, you know, Elon went out for a drive, live streamed it, and at minute 20, it tried to run a red light.

All right. 20 minutes, not 10 minutes. No, sorry. Can I interject? Can I interject for a minute? Do you know how far Waymo goes without running a red light? Like thousands of miles, thousands of miles.

Yeah. So I would say it was farther ahead than Tesla FSD is today. I would say that Waymo is obviously a very different product. It's geofenced in an area where they think it can work well. LiDAR. And also they have some hardware differences that may, at least at this point, may be better than just cameras alone.

I guess, though, I mean, I have my own complaints about FSD. First of all, calling it FSD, I think that's...

A bit of a misnomer. I mean, it obviously isn't yet full self-driving. It's a lie. It's a lie. It's not a misnomer. It's not a mistake. It's a lie. And the reason they won't, and it makes, misleads people, not so much the people who buy FSD. I think most of them kind of know what they're getting, but it's the rest of the world thinks that this product can drive itself because it's called full self-driving when it's not. I mean,

I would say even the full part, right? When you say full, you're like, okay, that's really self-driving now as opposed to just self-driving or as if autopilot was enough. I mean, my sense is the reason they say full self-driving because autopilot, the promise was that that would be the, what,

The implication is it can drive itself. Yeah. But no, now it's self-driving, but not just self-driving, it's full-drive driving. And they call it self-driving beta. I get what you're saying about the normal engineering practices in terms of labeling, but FSD beta, just like FSD, is just another marketing term, I'm pretty sure. I mean, no one... Right, but people don't know that. They don't know it's a lie. Right.

But I think the actual drivers do know that. I use it regularly. I'll use it every day. I do like it. I like it a lot. Do you use it on the freeway or on the city driving? I use it everywhere. Now, do I...

I make a distinction. It's not bad on the freeway. I'll tell. I mean, I will say it. It's not that it's not just not bad on the freeway. It's amazing. But I've had my car since 2018. One of the things I think that makes Tesla special and great is that I have a car that's very different than the car I bought four or five years ago.

And it is continually improving. I've seen autopilot get better. Autopilot's kind of not gotten much update because all the emphasis is on FSD, if you make that distinction. But I've seen that get a lot better. But frankly, regardless of what they call it, because I do agree with you, I think FSD is not the right thing to be calling it. I think even calling it autopilot was aspirational. Yeah.

But what I mean, what is it? It's like fancy cruise control, really, right? And would you blame your cruise control if it drove past a school bus? I'm not expecting my cruise control to stop at a school bus. But yes, I agree. But would you expect it to turn right for no reason at all?

No, but that doesn't really happen. I don't, I mean, I don't have that experience. I will say when I first had autopilot back in 2018, I wasn't using it on the highway. At that time they said it's only for the highway, but they didn't do anything to prevent you from using it on sort of neighborhood streets. And I was on a four lane road with the double yellow in the middle. And I did try to like drive me into oncoming traffic, but obviously I had my hands on the wheel and didn't let it. That's what I did.

That was four years ago. It hasn't done that since. But again, I am expected to be engaged. I'm expected to be watching it. It really is only a level two system. It's not a level four system like Waymo. Does it aspire to become a level four system? Yeah, but it just isn't there yet. I guess it's matching the expectation to what the reality is.

Again, the labeling, the calling it FSD. I don't think that's correct. But I said it several times here. I say that Elon lies, but he lies in a very special way in that often like he talks about things in the present as if they're true. Things that he expects to be true in the future, he talks about them as if they're true now. And a lot of those things do end up becoming true. Will we be on Mars in two years? Probably not. But those are the kind of things

that he has goals for. You can argue about his methods and there's always a question of the ends justifying the means. But he does often have a pretty good track record of getting to the ends that people would like. It might not be something you would like, but I would like to be able to take a nap and let the car take me where it's going. It just isn't there now. And I think they should be clear about letting people know it's not there now. I feel like it's maybe related to investors and

and and just the general perception that they want people to believe that's happening soon and i i think uh i think elon definitely believes that it's coming soon he's been saying next year for the last i don't know what six years or something eight or nine years yeah yeah for a while anyway if you believe your own delusions are you lying that's the seinfeld defense right the uh right that's right it's not alive i believe it so

Well, if you believe it, no, I mean, if you, if you do, if, if you have no reasonable reason to believe something, I mean, then you can't just say whatever you want and say, ah, I believed it. Cause then everybody will just say that every time that it's not the proper standard. It's a reasonable basis for your belief. I mean, the company is definitely invested in it. They,

or working on their robo taxi, even though the software isn't there yet. Another thing I don't like is that they essentially sell it to you and sell it to you again. You know, you pay full price for this thing and then you buy it again and you never quite got the software or even the hardware that was promised.

Yeah, you never got what you were promised. They said it was going to be the end of the year. People say, well, they get it eventually. No, they don't. They haven't gotten it. And some people bought it back in 2016 and they've waited seven years and they haven't got what they paid for after seven years. And yeah, and I mean, these things are these things lies. The painted video is is a lie. I'm sorry.

He told his engineers to develop a video to show how great full self-driving was going to be someday. And he told them, I know it doesn't do all those things. Just fill in the gaps where, you know, the things it can't do.

And they built that video and they sent it to him and he says, there's all these jump cuts, make it look like one continuous run. We have all the emails for this. He told them to do that. And then he said, yeah, I want you to put something on the front of a big black thing, a disclaimer like warning. The disclaimer says the the the car, this car, the driver, the driver is there only for legal reasons. He's not.

touching the steering wheel and if the car is driving itself. And then he put out a tweet saying, here's self-driving. Here's our description. Here's self-driving. There's no indication this was a concept at all. He put it out as the fact of what it did. We ran that route a few months ago.

It can't do that today. We ran it four times. It failed every single time. It still can't even do that route from that place. We found the place where it started, the Tesla headquarters. It still can't park itself. The whole thing was a total fake and Elon Musk ordered the whole thing. And we have everything in writing and in testimony and lawsuits now. That was seven years ago. Faked the whole thing from beginning to end. I will agree that that doesn't look too good. Exactly.

He called a solar roof meeting of the press. He had six houses with silver roofs on them. None of them worked. It was just, they were prototype concepts. They weren't even wired up. And he said, here's your arrangement for these art accidents.

Sure. Some of it, I don't know if you give some of it a pass for just sort of showmanship. Like when Steve Jobs did a tech demo or Bill Gates did everything really work. I know from personal experience that back in 2009, when they did the launch of the Model S, our sense was this is the actual car. And they described it as there's a battery in the floor and there's just storage in the frunk.

And they acted as the car that ran and we got test rides in that this was the real Model S. And I discovered that the car was actually just like a converted, I think it was a Mercedes CLS that was reskinned and they just put a Tesla Roadster drivetrain in it. And you knew because if you opened the armrest, the center console armrest, you could find that the little display, the little Roadster VDS hiding in there. Yeah.

And the battery was in the frunk. So that's why they didn't want anybody looking under the hood. So do I feel lied to? I don't know. I mean, it was still drove amazingly and I was super impressed and everybody was impressed and they eventually delivered the car. And that was a tricky time. At that point, they didn't even have a factory yet. They're talking about having some factory in LA as opposed to doing the new me plant and the shit they eventually did.

So I guess the point here is they eventually did get it. It's hard to know when something that hasn't quite been done yet will be done. I believe it's possible. It's in the realm of possible outcomes. And I think they will reach it. The question is just when. And so it's hard to predict actually when it will be done. That last 10% or whatever is always the hardest. But I think the product as it is right now,

the fi despite it being called FSD is pretty, pretty amazing. I drove from from DC to Austin and I wouldn't have done that drive without FSD. That would have been really taxing. You're on the freeway. Okay. Now let's talk about FSD as not as on the freeway, but as it's, as driving in the city, I fail to understand why it improves anything.

You get in this car and the moment you get in it and you let it start going, you take your hands off the wheel, you take your feet off the pedals, you say go, you give it a destination, you say go. At that moment, at any moment, it might decide to jerk the wheel to the left or the right

and run into something i mean this is not something that happens once every year it happens every day or two days or whatever while you're driving it means you have to be watching all the time you have to be watching to the left making sure there's nobody in the blind spot watching to the right make sure there's nobody in the blind spot right checking everybody in front of you because it could at any moment slam on the brakes or turn to the right or turn to the left okay so i will tell you my use case

What's better about that than just turning the wheel yourself? I'll tell you. So a lot of the things you described are part of the driving process, right? You're looking around, you're seeing what's around you. And what I find with autopilot or FSD is

you know, sort of three major things, right? You have, you're regulating the accelerator. So you're regulating to the vehicle in front of you. So that's some feedback loop that your, that your brain is dealing with. You're staying in the lane. That's another thing that you have to keep up with. And then the other thing is your navigation, like where you're going. When is my next turn? What lane do I need to be in for that turn? So in a long drive, all those things can be a bit taxing. And so what I find

with the Tesla, that drive is not as nearly as taxing because I don't really have to worry about, am I in the right lane for the right turn? But what I am paying attention to is, okay, the car is changing lanes now. So I guess I need to be in that lane and I'll pay attention to how it changes lanes. And so that's mostly highway, but also in suburbs or whatever, it's the same thing. It's okay, this is where I need to go. It's taking me to the hardware store or whatever. It's found the

The route, which is probably done by Google maps, I assume. And but it's, oh, it's getting in the right lane. Oh, I need to turn left here. Okay. Yeah. All right. Great. And I'll pay attention to that. So I'm basically just monitoring it. It's a much less stressful, less taxing experience. I mean, I'm ready for it. My hands are sore on the wheel. I don't really experience.

crazy jerks too often. And it makes the experience nice enough that I appreciate it. Now, I'm not like some drivers, I will intervene as soon as I feel like it's

doing something that doesn't make sense to me. If I get to a stop sign, I'm not going to wait for it to creep up. If there's nobody there, I just tap the accelerator and it'll go. It's getting better. There's things it can't do. Like it doesn't know, as far as I can tell now, it doesn't know no turn on red. So I have to take over for that. You don't understand any sign except stop signs. I think it understands yields. I think it... We've seen it do that all the time, especially on...

Uh, roundabouts, it just dies. So I don't know what, I don't know what version of software you're on, but roundabouts, they used to be bad, but I have no problems with roundabouts now. No problems with yields.

So, again, it's... Come to Santa Barbara. We give free rounds out to people around Santa Barbara. And if you don't walk out of there feeling a little nauseous or scared or whatever, it's just amazing. It does the crazy things.

I guess I'm continuously impressed with what it does and the choices it makes. I think it's definitely maybe because I saw it when it was worse. Okay. It's a lot better now. And so I feel that the slope of that trajectory is positive.

And it may be asymptotic in terms of getting to where it needs to go, but I feel like it will get there and I'm very happy with it. But I see it as fancy cruise control. So I'm not getting mad at it for not knowing things and I'm impressed when it does know things. I don't often have it saying, okay, kill you now. I don't have that. I don't have that experience.

These days, in the last couple of years, I haven't really had that experience. So just for the record, Dan, I don't agree with Doug. This is some of the back and forth we have a lot. I run FSD too. I agree with both of you that it's pretty decent on the highway. Although I will say on my last road trip, it wasn't.

suffered mightily from the mirage effect. So when you've got that water looking mirage on the highway, the car would immediately slam on the brakes. It just didn't understand what that was. It didn't like big shiny trucks next to me either. So I learned to avoid those. But my use case in town, which makes me very critical of it is simple. I want to leave my house and

I want to leave my neighborhood. I want to go two miles to the high school and I want to drop off my daughter. The car is absolutely incapable of any of the above. I happen to have a roundabout in my housing track and the car, my kids and I laugh about it because the car literally goes into a cyber meltdown at the roundabout. It will get me to the high school, but once it sees these hordes of kids crossing the street everywhere except the crosswalk, it just kind of throws up its hands and says, I'm done. I can't deal with this.

So we're back to the whole point about full self-driving, what the words mean and what's really happening in the real world. There seems to be a disconnect. Oh, D-Robin? Yeah. Auto-summon? That's a party trick, dude. I mean, just, yeah. Yeah.

Good luck with getting that fixed anytime soon. I've never used that feature. I've never even tried to use that feature. To save its life. It's amazing how bad it is at parking. And everybody solved that. Ford's got it. BMW's got it. Toyota. Everybody solved parking and they can't park. It's terrible. That was one of the first things I tried when I got my Tesla was to auto park it. It couldn't do it. And I just kind of shrugged and said, oh, that's awesome. Right. Well, we've discussed that.

why we think that is, right? Most vendors, the parking is a compartmentalized feature that was bought by some third party and it works because it has to work. Whereas Tesla, it's this sort of deprioritized, but it's also part of the overall design

FSD thing using the same sensors. Yeah, right. Similar with the windshield wipers, which I still have them come on randomly every now and then. And you can live with that, except it's forced on auto if you're using FSD. And most manufacturers have solved that. I mean, they didn't solve it themselves. They bought it from Bosch or whoever.

And they have a unit, a specific unit that does just that to detect if there's rain. And instead, Tesla is using the same cameras that they use for autopilot. And there's some neural net that is dedicated to or has a piece of it that says run the windshield wipers when there's rain. And it's just not prioritized to filter out false positives. So it comes on every now and then. So it's

It's not a safety issue, but it's an annoyance and they don't see it as a huge priority. Like same thing with auto parking. Occasionally I see the option, but often I'll just do it because it's too slow. It's not worth it. It's embarrassing if there are other people driving by.

But I still enjoy the car. It's still the best car I've ever had. I do have faith that Tesla will fix these things eventually. So I don't know. I'm not an unhappy customer, but we have Tesla Motors Club. Our goal really is for the customer, right? To advocate for the customer. And so I do wish that they would get those things fixed sooner than later.

So I need to ask, so Dan, you own Teslas, you drive Teslas, obviously. Do you enjoy Teslas as a product beyond the FSD and the internet software and all that kind of side of it? But like, do you enjoy from the, how it drives and that kind of, those aspects of it? Well, I love my Roadsters. They're the most fun in the world. That's all I've driven for 13 years is, I have two Tesla Roadsters.

Right now, Mandy's got a regen brake problem. So I drive candy all the time. I usually switch them off about once a month. But every now and then one has a problem. And I send it in and drive the other one for a while. So I always have one available. I always have one to do. My wife got a, in 2012, we got a Model S. And she's still driving that today. That's 2012.

11 years. That's still her daily driver is the Model S. It's our family car. It's what we do every day. Okay. That's awesome. And would you buy or would you be consider buying the new Roadster if it ever gets developed? Well, speaking of misrepresentations, you had $250,000 in 2017 to reserve one. Mm-hmm.

Maybe if it had been five, I might have put down a reservation, but I'm glad I didn't because everybody who paid $250,000 six or seven years ago has not seen what they thought they were going to get. They saw one. I saw it drive out of the semi and everything. Oh, wow. And here we are six years later. At this point, I'm not waiting, you know.

Yeah, I didn't wait either. I ended up buying a Model S Plaid because I went, they're not going to release that car. Otherwise, I would have loved the Roadster, the new Roadster. I thought it would have been cool. But yeah, I mean, will they ever develop it? Maybe. I don't know. We'll see. I think it was $50,000. It wasn't like the cost is $250,000, but the reservation, I think, was $50,000. But it's still a lot, though, to give as a free interest-free loan to the company. Yeah.

When you don't mean to, when you thought you were going to get one a year away, not seven. Well, that's how the company started. Everyone that bought the first Roadsters back in 2008, a good number of those, like the first hundred or so, paid full price and waited for their cars. And that money was used to float the company, obviously, right? It's not a new thing. Yeah, it's not a new thing. It's a new thing. It's a new thing. We've been doing it since the beginning.

But I would say, and I've said before, the new Roadster is another deprioritized

product. Their moneymakers right now are the three and the Y, and they need to do their cost savings on those things and get those out and get the newer versions out to refresh them since they've been out for a little while now. And then the semi, to me, is the big thing that actually makes an environmental impact. So I think that the company has those priorities in the right spot, and the roaster will happen when it happens. But partly because of how expensive it is, they only sell so many of them. So it won't make that big deal in terms of the company's bottom line. So that's obviously going to be deprioritized.

Right. So, Dan, you mentioned Waymo earlier. Do you have positive opinions of how Waymo and Cruise are doing their self-driving? Do you think their approaches are better? Or are you worried about them as well? Is it just a general safety concern? Well, until a few weeks ago, I basically commented only to a limited extent on that because we had no experience with those cars. We can see what is being reported.

We can see what other people say, but we decided to get to learn about it. A lot of people gave us a hard time, said, why are you looking at Waymo? Why are you looking at Cruise? You're only looking at Tesla. So we decided to go to San Francisco and take our Model 3 up there. And we picked, I didn't, I was not there, but two of my guys were. They went and they

went and they took a cruise. We would take a Waymo. We've been on the waiting list since last November to get a Waymo, to get an invitation for Waymo. I don't know. We keep asking them why they- Call up Sergey and Larry. Get them to give you some access. So we will do Waymo if they let us on. They just, we couldn't. We called them. We tried again and again, and they said, you're on the waiting list. Okay. But cruise said, fine.

So we went out there and drove around on a cruise. We drove, you know, a bunch of like a whole evening on a cruise. Not one problem. I mean, not a single problem, not a single stop, not a single got hit the accelerator, nobody in the car. It just did exactly what it was supposed to do.

You didn't have a hipster jump out and put a cone on the roof? We filmed the whole thing, but then we decided, I wonder what would happen if we set the Tesla out on the same route. Now, Omar is always out there giving his zero intervention drives. And so we said, let's go find out how Tesla does against Tesla.

About cruise, so we gave it exactly the same route that it took us on the, uh, on, uh. On the cruise, and we just programmed into the Tesla and started it. Couldn't do it. It couldn't do the route. Several failures tries it tried to crash twice.

while we were driving it, just trying to reproduce its behavior. It does not work. There's this rumor that it's programmed, like it's geofenced for San Francisco. It knows San Francisco really well, like the roads have been programmed into it, so it really knows to do it well. We went out there with our car, and it screwed up just as bad in San Francisco as it does in Santa Barbara, and we tried it in L.A. as well.

We also took good old Omar Holmar's. We took his zero intervention videos and we mapped out his route and we decided to try his route. And guess what?

We ran it and it failed every time. We didn't get a clean, one clean run on his route. I think we did three or four times. Ran a stop sign, ran a red light. You know, the run of the mill things that you get on a 10 minute drive through San Francisco. It's not better than it is other places. - But in your car, you put tape over the cameras, right?

You can't put tape over the cameras. It won't work. We tried it. I mean, let me tell you. Oh, that's not true. If you put tape over the rear view camera, it'll still run. It'll still run FSD. But every other camera, if you put tape over it, it will not run. It says, I can't see. I'm not going anywhere. You know, I can't see where I'm going. Do you think Elon knows your VIN and they gave you the bad software? No, I'm just kidding. Yeah.

So I guess my question then is, so you had a good positive experience with Cruise. So do you see a path forward for self-driving and autonomous driving in general? Do you think that's a viable technology that's achievable in the not in the let's say by the end of this decade? We won't go crazy into ever. But do you think it's a viable path that people should be working on this and that it can happen?

Yes, I think it's actually not that far away. Okay. I think Cruz and Waymo. And if you go to China, there's like five or six companies doing this and they're doing a tougher job. You know, they have a tougher job in Shanghai and places like that, which are crazy, crazy cities. All the evidence is that there are.

probably about 10 companies who are probably only a year or two away from really having a thing. I mean, in cruise and cruise and Waymo have actual driverless road taxis. They are geofenced, which means they've been tested in certain areas and made sure that they really work. That's what geofencing means. And they say you can't use them in places that we don't know if they work. Well, the geofencing is expanding and cruise just announced

Nashville, Atlanta, Miami, Seattle, Raleigh. They got like six new cities they say they're going to be in in the next few months. Houston, Dallas also. Those are all, they're all there. And we have, and in China, the same thing is true. There's people in multiple cities. It's just a matter of experience. You've just got to do those things. Those guys will succeed. They have a product that, I mean, no one has died. They haven't killed anybody.

They've been doing this for a long time. They're being very careful, unlike Tesla, which is reckless. They're very careful. They follow it step by step. It's taken them forever. It's going to take longer before it gets done. But I believe they will succeed. From all the evidence that we've seen, they have a product that's really quite good. It is not actually better than a human yet.

and they need, you know, but they're working on it and it is not on any freeways, but they are now testing it on freeways and will presumably be opening up various freeways as soon as they complete their testing.

whatever that qualification they need to do. So what about other systems that are in normal customers' hands like Blue Cruise? I keep wanting to call it Blue's Clues, but Blue Cruise and sort of Mobileye based systems. Do you have any opinion on that? And does your company have any software related to those types of systems?

Well, so the interestingly, we've not put out any reports yet because we just haven't done it yet. But we have tested both the Ford and GM systems against Tesla and to see how they work. And they all work pretty well. Again, on the freeway, they work pretty well. And of course, the Ford and GM stuff is limited to the freeways. They're not allowed to operate off those areas. But

They work pretty well, but there is a huge difference. Driver monitoring.

Now, the entire basis of the argument that FSD is not basically a murder machine is that there is a competent driver behind the wheel stopping it from doing insane things. It does insane things very frequently that threaten other people. But there's a driver there who is competent and their attention is on the road and they're there to stop it when it tries to kill somebody. That's their job. It's to hit the brakes or grab the steering wheel when it's about to do something crazy.

But that assumes the entire safety case is based upon there being a driver who is attentive and ready to stop the car if it does something dumb. The basis of that is driver monitoring. We know it doesn't happen because the first fatality that ever occurred, it was ruled that the driver had not looked at the road for seven seconds when he crashed into that truck. And he was distracted. We don't know exactly what it was. There was a rumor he was watching a movie. That was not true.

But he was distracted in some way. And so the NTSB came back and said, you've got to insist that you have to build a system and make sure the driver is not taking their eyes off the road. You have to have a driver monitoring system. And that was, and Tesla agreed to put that in and they did put it in. But we tested the driver monitoring system. You should look at our videos.

We thought I asked a question. My first question that I asked our team was, so that's great, but what about it? What about if I'm wearing sunglasses? I have these mirrored sunglasses. You can't see my eyes. How do you know they aren't? How do you know they're not closed? How do you know my eyes aren't closed behind the glasses? So we said, let's go out and try it. They went out and tried it with some dark sunglasses and and you could do pretty much anything you wanted.

If you looked away, it was okay. If you turned around, looked out the window, that was okay. So we ran a series of tests where we showed that even if you close your eyes with the sunglasses on, it can't tell. So we realized sunglasses are a total defeat for the driver monitoring system.

we've complained about that before on the show we wish that they had used ir cameras in the car like some other companies because then it can see through the sunglasses so it is definitely a limitation that i think they should probably fix in future models of putting ir cameras or sunglasses don't defeat the system it's a fair argument i would say that was in a model last

that first fatality that was using the Mobileye based system. All it really was back then was cruise control with lane keeping. So there shouldn't have been any more expectation from the driver, but the driver was overly confident in the system. - Called autopilot, I think.

Yeah, it was called autopilot back then. And and once you're here, take your phone call. It was autopilot 1.0 using mobile eye, but all it really was was lane keeping. You know, there is some driver monitoring now there's a camera. And as we said, we complain that it doesn't have an infrared illuminator that could, you know, see through sunglasses. My experience with it is, yeah, if you have your sunglasses in it and you keep your head forward, it it

complains to you less as someone who's gotten plenty of dings for like driver attentiveness, basically because I'm messing with the screen to change what supercharger stop I'm going to go out and yell at me to pay attention. But you can look around. It gives you a moment to do that, but it's fairly strict about it, I feel. But it doesn't work as well at night either because again, you're not illuminated. And I think the irony of it, though, is that it wouldn't complain to me

If I didn't have autopilot or FSD engaged, so if I disengage that and I'm driving, but then I'm also messing to decide which supercharger I'm going to charge at, it's not going to complain to me. But if FSD is on, which should be the safer situation, and I believe is the safer situation, that's when it's complaining, right? That's when it's yelling at me and beeping at me.

while the system is running. So in theory, maybe it depends. People complain about seatbelts, right? Maybe it should be monitoring you the whole time. I've seen systems that do things like let you know if you appear

appearing drowsy supposedly that's something tesla will do soon but i do think it should have better driver moderating systems like a an infrared system maybe something in a different position as well and an ir illuminator like just the led because watch our videos our guy look turns to the right completely to the right and looks out the window for five minutes and it doesn't say a thing he looked at five minutes five minutes doesn't see a thing

We started wondering why that was. So we decided to put one of our mannequins. We're famous for running over mannequins, but we decided to give them their chance to fight back. And we put a mannequin in the driver's seat and it wasn't a problem. It drove the car with a mannequin in the driver's seat. So we thought, wow, well, that's a kind of a human-like sized person that's there. So let's try something else. So we tried a big teddy bear and it drove the car.

Then we tried a balloon that we put a face on it. We drew eyes and nose and mouth on a balloon and we put it in the driver's seat and it drove the car. So then we took it all out and we just said, well, we'll have no driver at all. And it goes fine. It drives fine with nobody in the driver's seat. I mean, you must have had to lock the steering wheel, put a weight in the seat. That's right. We put a steering wheel weight on the steering wheel. Okay. Well, that won't work anymore.

Really? Yeah. Yeah. You can't, you can't get away with the weight on the steering wheel anymore. I think it's just looking at it. If you have a constant torque, I'm sure you could engineer something with some water or something that adds some noise to the torque.

But right now, if you put a weight on it, it'll kick you out and it can detect that you're violating that. But I mean, the point is, right, there are certain interlocks and anyone can come up with ways to defeat those interlocks. And so it's... There was nobody in the car. The driver monitoring system is so incompetent that it can't even detect when there's nobody there or when there's a balloon there. Yeah. You've got to be able to do that. Ford and GM system.

are impossible to beat. Our same guys who beat the Tesla so many different times and so many different ways, it was just terrible. Yeah, I guess they wouldn't... They know when there's not a driver there. They know everything.

I guess I wouldn't fault Tesla that much there. It's definitely a different design philosophy. I mean, those systems that you're talking about, the early ones with that are just really annoying, right? Because you have to sit in a certain way. And if you move your head in a way that their system can't detect, then it won't let you use your super cruise control or whatever. My guys don't have a problem with it. They drove hundreds of miles on GM and Ford systems and said,

As far as they're concerned, they work their job. They do what they're supposed to do. And Tesla does not. We know what the algorithm is, actually. We figured it out. What it does is it looks for something that looks like human eyes.

When it finds them, it sees if they're looking in the right direction. If so, it says it will beep you if it sees you're not looking straight. But if it can't find a set of human eyes approximately where they're supposed to be, it just says, okay. It faults to on rather than off when it can't see something that it...

things are humanized. Yeah, and I would agree with that. That's what it does. I don't know that that's the wrong thing. Absolutely. It's the wrong thing. What's the point of a driver monitoring system that doesn't monitor whether there's a driver there or not? Well, I will say that driver monitoring

was an afterthought with Tesla. Again, they're thinking they're in the future, right? I'm not even joking here. Why would you need driver monitoring when the car just drives itself? And I think that was the Tesla idea. Again, they're living in the future, but you need to deal with the reality of now, which is okay. Yeah, we do need some driver monitoring. And that was in a sense a retrofit. That's what came with the Model 3. And it took a while for that to move back into the Model S and the X.

and they're still developing it. But like I said, it sounds to me that you're using a slightly older system, like the weight on the steering wheel doesn't work anymore. Don't ask me how I know. It's 10 feet away and he says it works. You can't use it on the freeway anymore, but you can still use it in the city.

They just tested it in the last few days with the latest version. So I just, I want to ask a question real quick, Dan. So you mentioned earlier software development improvements. There's ways to do it much better. You disagree with the ways they're doing it. I'm just curious, is your software, Green Hill software, are you guys doing any vehicle software? I know you did a lot of aerospace and aircraft. Do you guys have products or software that you're doing in cars? Like, are you working towards fixing this?

Oh, well, in some, yeah, we have, we work with almost every car OEM. We do something with almost everybody. Okay. So are people using integrity or your OS based system for running cars as well, or your embedded system, or is it?

Well, it's all over the map. For some, we're the primary operating system in the car. For some, we're on only certain systems, like maybe self-driving systems or maybe entertainment systems versus the actual drivetrain systems. Every car company, every car is bid out to multiple vendors individually.

And each of the various components are built out to multiple vendors. And we sometimes win and sometimes we lose. We're not on everything, but we're on a lot of things. Almost every company uses us at least for something. Okay. So I guess the reason I was wondering, because obviously you have a lot of knowledge in the space, but some of your critics will say things like, oh, there's business reasons or there's competitor reasons of why you're critical of this. Do you think those are fair assessments? Well, I have an analogy.

So people say, because I sell software to GM, I'm a competitor to Tesla. Well, but Pirelli sells to GM. Are they a competitor to Tesla? I mean, if you sell a component that's not a car, but a piece of a car to somebody who makes cars, you compete with the other people who make those pieces. People who other they compete, Pirelli competes with the other companies.

tire makers, we compete with the other people who are out there making software for people to do development of the car software. But we don't compete with Tesla. They don't sell that product. We don't sell anything they sell. They don't sell anything we sell. It's not a competitor. We aren't a competitor. They're a potential customer of ours, which they're not. Well, they might be somewhere, but

Not directly that you're aware of, really, is they're not necessarily a direct customer of yours. Well, we don't make self-driving software, for instance, or cars or anything, and they aren't out there in the market selling operating systems to companies to build cars, right? Sure. Our product lines don't overlap. That's what competitor means.

If there's some interest, of course. And if Pirelli, would Pirelli go out there and start trashing Tesla because they didn't buy Pirelli tires? No, because then they'd never sell a tire to them. Why would you do that? Why would you attack a potential customer just for not buying your product? Doesn't make any sense. You just insured that you spent money ensuring you're never going to sell anything to those guys.

Sure. I guess what I'm wondering is what's your end goal for the Dawn Project or these types of, you know, obviously you're raising awareness, you're trying to voice these concerns. Do you think that it's something where you got your company could be a vendor to Tesla and helping them improve these things or do a better job? Or do you think that...

that, you know, like what's your end goal? I guess is what I'm trying with these types of things. With respect to full self-driving, it's been one simple goal all along to get that product off the market until it is tested thoroughly, until their own testers, their own test drivers can't find any material nor bugs. That's it. Keep your own guys out there pushing it and finding problems and fixing them and pushing it and finding problems and fixing it. When you get to a point where

You just can't. It's just running out of bugs. You're just running out of bugs because it's working so great with your test drivers. That's when it should go on the road. Nothing should go on the road that has bugs that your customer, that are safety critical bugs that your customer is going to see and you know about them and you've seen them and you don't bother to fix them. That's wrong. We keep reporting bugs and they keep not fixing them.

I'm going to flip this on its head slightly and I want to ask your opinion about the bugs. So with FSD, there's a delineation that's taken place. You've got the time of FSD with ultrasonic sensors and radar, and then you have FSD without because Elon made a decree that says everything's going to be Tesla vision. That's the way it's going to work. Would you consider that a self-induced bug?

where you literally remove sensors and say, yep, we're going to concentrate on a single way to see the world? That's a very technical question, right? You don't think it's a general answer.

Can you build a car without LIDAR? I don't know that. I mean, I don't know. Maybe you can, but my way of stating it is all the people who have self-driving cars, I mean, ones that are actually self-driving that don't have anybody in the car telling it what to do, every single company that has a product like that, of which there are probably at least 10, all use LIDAR.

All the people who say that's crazy and we shouldn't use LIDAR or radar, their cars can't drive themselves. That's not a proof. I admit it. It's not a proof. It's just a trend. Oh, and they say it's necessary. They say we use it and it's necessary and we've succeeded. And the guy who says it's crazy, we'll never do it, hasn't succeeded yet. Will he ever succeed? Maybe. I don't know. So as I said, I don't know the details of

of lidar i've actually taking out radar i remember i just read that chapter in the in the isaacson biography and it was he confirms that it was pri it was availability based it's that the chips that they needed for it had stopped becoming available and they would either have to stop shipping cars or stop shipping cars with the radar in them so he chose drop the radar is radar necessary i don't know i'm told

the problem with both lidar and vision is that when it's raining and when it's you've got dust storms and you've got other obstructed things your eyes and the lidar which is just light after all are affected whereas radar is relatively immune to that and can see like a large object you know in front of you uh and tell you i've done a little research on radar

And the research I've seen is it is a very low resolution device. It's not very accurate. So, but I don't know. I'm not an expert on radar or LIDAR. So whether they're necessary or not, it's the people who have it have been much more successful than the people who don't. So there was a question back there about...

My goal is in general to get software that lives can depend on that's crappy, that has bugs in it out of the market. Don't allow people to sell it. And if they sell it, they should have to recall it and fix it before they're allowed to put it back in the market. That's the general goal. This is an example of that. If people ask me, well, you hate Tesla, you want to shut them down. What I've said is simple from day one. If tomorrow...

They said, gosh, we get it. We're going to put it back in the lab. We're going to put it back on. We're going to put it back in the hands of our professional drivers. And we're going to drive it around until those guys can't find any more bugs in it before we give it back to civilians, to ordinary people. I'd shut up.

And I go bother somebody else who's selling another piece of crap software that shouldn't be out there. So you've mentioned before as well, I've heard you talk about how you have a methodology for developing software that's alternative to say Agile and

And I agree with you that Agile is terrible, but we won't get into that. That's for a different type of show. So what I guess what I'm wondering is maybe the altruism behind the message, right? So a lot of these systems are things of processes out there that are not working so well that you say that you have a better way to do that.

"Do you have any intention of releasing or publishing your methodology so other people could follow it versus paying to have a course?" Like I actually looked, I was curious. I was like, "Hey, I would love to learn more about it." And I was like, "Oh, I gotta sign up for a course. It's $4,000 a person to learn all these details about some of this stuff."

And so I was just wondering, are you planning on publishing those types of things or your operating systems, obviously proprietary microkernel closed source. It's not for free. Obviously you're entitled to make money. I don't want to say that you shouldn't make money off your software. I definitely make my living from that, but do you have any intention of releasing things so that other people could do a better job or do you just want them to come and buy your products?

I have published some pieces of that. Some of it's on the website. There's much more. It's mostly a time issue, but having time to write up a lot of the stuff is, I would say, obvious. To me, it's obvious, but it's not obvious, you know, up front about how to, I mean, well, for example, we have the 90-50 rule, very important rule.

When you say, when you're a programmer, you're a developer, and you're developing this product, and you're working on it really hard, and you've got a specification, and it's coming along really well, and a lot of features are working, and you feel that you're at 90%. It's like you could finish stuff any time. There's some bugs I've got to fix, maybe a couple features I've got to add, but pretty much any time I could just drive this thing to completion and ship it, at least beta, ship it out, get people to use it.

You're at the halfway through your schedule. That's when that happens. You need to recognize that moment. It's a really important moment. I've seen it to myself dozens of times. It's a feeling that you feel you're on top of it, right? You're not climbing up the hill anymore. You're finally in control and you know what you're doing and you're going to finish this thing. And it's not a slog. It's more downhill from here.

But you're only at the halfway through the schedule and you need to understand that. That's what gets you in trouble. If you think you're 90% of the way there, you tell everybody it's going to be out in two months and it's not because you're not that close.

That's that's one of the rules. It's called 90 50. And I felt it. I discovered it 30, 40 years ago. And and it was just an observation. And then I started, well, is that really always true? It's always worked that way. And yes, pretty much. And now you write it down, because if you know that I keep asking people, are we at 90 50? And tell me why you think we're at 90 percent.

And if I'm not convinced that they really are, because they're lying about it, I say, okay, you're not at $950,000 or you are. Okay, great. Now we know how long you've been working on it. Now we can guess where it's actually going to be. It's actually going to ship. And I can make good business decisions on when I should announce it, knowing when it really will ship. Whereas often sales and marketing people think that engineers are gigantic liars. Do you know that?

Because it's going to be three months. And then the guys tell all their customers is three months. And then at three months, they say, well, it's going to be a little longer. Another three months. They're being lied to. It's just that's one of the things they don't know. They don't know how to estimate the project. They don't understand this. There are simple rules, rules of thumb, which are really very effective. That if you don't know them, you keep making bad decisions when you know them.

you make your schedules you get things done when they need to get done it also tempers your uh ambition as well because when you realize that you say maybe i should not do all the things i want to do and get you know get it out in in a shorter period of time uh sure because you you're just mis-estimating it's hard to run a business when everybody's lying to you about when things are going to get done and you know elon musk suffers from that to a certain extent um you know sure

He has, at least in some cases, thought things were going to get done a lot faster than they did. Then he got everybody in a tizzy over the fact that it wasn't even close. Sure. Yeah. So I agree with everything you just said. I've never heard the 90-50. I've always described it as the first 90% is easy. The second 90% is hard because it's basically the same thing. You're not nearly as done as you think. Is FSB at 90-50?

No, I don't think so. But I would say... We've been working on it eight, nine years. That's right. That's how far away it is. Sure. I mean...

Yeah. In my experience, I've done a lot of machine learning and data science stuff. I actually, I'm an architect over a data science team now at my current day job. And what I would say is the data scientists that work on these ML problems are actually worse than software engineers when it comes to estimating. Because generally, cause they're doing research, they're trying to solve a novel problem or doing something. So I would say somebody needs to coin a 90 50 rule, but it's going to be something like 95 or something like that for data scientists, because those things tend to take a lot longer.

And the problem is FSD is a novel problem with a much larger search space of solutions. And yeah, anyway, we could talk about, I could ask you lots of software things, but I don't want to waste your time. One other personal question for me, I'm just curious. So you guys do a lot of embedded safety critical systems.

I'm assuming aerospace you do C, C++ and Ada. Have you guys been playing with Rust at all? 'Cause Rust is the language I mostly develop in at home for embedded and also desktop. So I'm just curious, that's a safety critical more modern language. Is that something you guys have invested in? We have some capability, but no, not any substantial number of customers. They're almost all C, C++ and Ada. That's our product line basically.

Sure. That we provide the whole set of tools, debuggers and compilers and everything for C, C++ and Ada for basically every processor out there, every embedded processor out there.

Sure. Okay. I was just curious because as somebody that also cares a lot about safety, that's one of those, it defeats a lot of classes of bugs in C and C++ with its memory safety guarantees. Not that it's still bug-free software. You still have to write software intelligently. But anyway, I was just curious. All right. You can enhance C in certain ways.

both with the compiler and with your debugging system and with your development protocols to get a lot of those benefits by just doing things a different way. You could do them badly, but choose to do them better. - Correct, agreed. Okay, cool. Let's see, I don't know what other questions we have, Doug. - One thing I wanted to ask since we've been talking about other systems, do you have an opinion about OpenPilot or Comma AI? Have you seen those systems? Do you have any opinion about those systems? - Yes.

We have actually studied it. We've got our comma AIs and we got it all set up and we got some cars and we installed it in the cars and we wrote some software and downloaded it into the comma AI to modify its behavior.

And it is hard for me to believe that those idiots expose the canvas to every idiot on the planet. You don't even have to break anything. You just go into the rear view mirror and go like that. You don't have to break anything. And the canvas is exposed with connectors that you can unconnect. You just open the connectors up. You slot yourself into the canvas. Holy moly. It's something like that.

Who thought of that? That was crazy. Well, I think that's designed as an... It's unencrypted. I think that's expected to be an open project, right? They expect their users to be hacking into it, right? I don't think it's even designed to be... Why? I couldn't believe they would expose the CAN bus

I mean, it's a product for DIY hackers, tinkers. It's not a product for a regular person. I'm talking about Ford and GM and Toyota. We did it to all the cars. Oh, gotcha. Why did they expose the canvas in an unencrypted fashion so anybody...

could change the braking on the car and inject braking messages and steering messages. It's crazy. We did it all. Because they're now a security company. Yeah, I agree with you. Oh, I see. And that's how common AI works, right? It's using the existing hardware in the car. Yeah. And you can reverse engineer, and they reverse engineered the messages and everything. We tried all that stuff out. Yeah. Yeah.

And it doesn't do much. I mean, it is not much of a product itself as it is open pilot and whatever. It doesn't do very much and it does it pretty poorly. I would not even, I mean, I would not even think of using that as a, as a mechanism. So are you going to put out a Superbowl ad against Kamei OpenPilot? I don't know. We haven't, we haven't, we just did this a couple of months ago where we bought a device and did it. My criticism is of the car manufacturers for exposing that bug.

That's what they shouldn't do to everybody. You can just buy this product, right? It tells you exactly how to do it. Go up to the thing, click on this, open the thing up, pull out this thing, pull apart this thing, stick your thing in and you're on the canvas. That's the real canvas. That's not a fake canvas. That's the real one with the real commands on it. And you're sitting in the bus.

I actually like that I can do that for myself, but I agree with you that most people should probably never do that. And it's a horrible idea, but I've played around. We drove the car. We made the car steer differently just for fun to see that we could do it. So speaking of videos, you mentioned Tesla's video from a few years ago and you felt that was, let's say, inaccurate. It's still on the website today and it still can't do it.

Okay. So you put out a, I guess it was a commercial and I guess it showed in the DC area where I was at the time. So I think I saw it during the Superbowl. When I look at that, I have a lot of questions about that video. So why don't we look at it? Let's see. We'll pull it up. Full self-driving will run down a child in a school crosswalk, swerve into oncoming traffic, hit a baby in a stroller, go straight past stop school buses.

ignore do not enter signs and even drive on the wrong side of the road Tesla's full self-driving is endangering the public with deceptive marketing and woefully inept engineering 90 agree that this should be banned immediately why does NHTSA allow Tesla full self-driving okay how many takes did you need to get the head to roll off hundreds of takes of all of those

We've got hundreds of them. We've run it on again and again. We've got ones where the head flies like 50 feet. We have ones where the dummy, I mean, rapid unintended acceleration, its arms come off and its head comes off and it goes in a different direction. We've done that 100 times with different roads and different speeds and different sizes of mannequins. And with dogs, dog mannequins, not real dogs, dog mannequins.

We've done all of it. Okay. But so looking at that video, it certainly has a goal, right? The tone of the voice, the drama of children and heads that come off when you hit them.

Accurate is that video really like when I'm watching it, I'm trying to look at the screen to see what mode is the car actually in. I see a lot of warning signs, but I can't quite tell what the warnings are. There's one part where I think you say that the car will drive in opposing traffic or go out of lane. But when you're looking at it, it's the other car that's crossing the line, not the Tesla. So I feel like there's some showmanship in that video that maybe treads on accuracy of it.

No, they're absolutely all reproducible. The warning message is that it says supercharging unavailable. That's what the warning message says. It's hard to see it on some of the versions, but our car always says that because we've never signed it up for supercharging because we don't want to put a credit card on our car. I'm trying to put a card to our car.

Every one of those tests is reproducible. We've reproduced it many times. We have invited, I've invited Elon Musk. I've invited Tesla Engineering. I've invited NHTSA. I've invited Whole Mars. I've invited Tesla Boomer Mama. I've invited Dirty Tesla. I invited Ross Gerber. I've invited everybody to come. I will show you every one of those tests. They're absolutely legitimate. They happen. Those aren't one out of a hundreds, right? I mean,

Of course, people putting together found the most interesting and dramatic issue. It's right. But that will happen again and again. There's no... They're reproducible tests. And we do know the exact speed. They're all running whatever the latest version was at the time, which of course has changed. But...

But even today, every time a new version comes out, we go out immediately and rerun our tests on it to make sure. Did they fix anything in the last year? Running over kids in crosswalks? Nope. Going past school buses? Nope. It does not know what a do not enter sign is. As far as it's concerned, it's just did nothing. It doesn't know what road closed signs mean. It doesn't know what right turn on red means. We reported these as bugs. All they had to do is shut us up was fix them.

They never fixed them. It's a year. How can you know that your product would run over a child on a crosswalk for an entire year and not fix that problem? I don't understand. That is outrageous. Again, we're talking lower than you go. There just isn't a level of hell that low that you would do it. If this happened to Toyota, they would put 100 engineers on it the next day.

If this happens to GM, they put 100 engineers. Holy crap, I ran over that kid. We've got to do something to stop it. How do we stop it? He says, let's work on parking or summon or God knows what they're working on. You know, they literally do this. It's in the ludicrous book and I've seen it myself.

He brings the Tesla engineers in and he shows them his route to work and how it screws up. And that's what they fix. They fix his list of bugs in his order, not the ones that, well, his kid didn't get run over in a crosswalk. So that's not on the list. To be fair, I don't think the car is seeking out kids to run over. Yeah.

I see that kid. You can, we can look at the video from the inside. You can watch it. And sometimes at the last second, I mean, 10th of a second, maybe it sees the kid. Sometimes it even slows down a bit because it sees the kid. It's way, way too late. It does not see that kid. Yeah. Um,

We've also talked previously about the placement of the cameras and that a camera in the nose may be better for that sort of thing. It's all sorts of things you can do.

The thing is though, the driver's in control. I personally would like them to get the parking working better and the smart summon, because those are things that I use. I'm not planning on running over any kids anytime soon. And I do have my hand on the wheel when we're going through areas with crosswalks or people. So again, to me, it's more of an issue of how the system is sold.

To me, it's just fancy cruise control and I wouldn't expect cruise control to not run over a kid. I'm the one in control. It's only a level two system. But automated emergency braking should. Why does it not stop? Why does it not have the emergency braking that they advertise that about the NCAP tests and whatever, how it won't run over the kids? Why don't see the kids? I've seen tests where it does work.

But I don't know. Personally, your opinion is the software shouldn't be sold or be available at all. I am very happy to use the software and I understand its limitations. Perhaps if its limitations were better advertised. And again, I don't think calling it self-driving or full self-driving

Like when it actually is available, we'll have to be super duper extra, super full self-driving because I had so many modifiers on top of it to say it's the real thing. If you want to be cynical, no, that's not what they're going to do. They're going to put it only in their own robo taxis.

that they custom built. They're now custom building a robot taxi device. How are the people who bought Model 3s and were told at Autonomy Day in 2019 that they were going to be getting $30,000 a year starting in 2020 and their car would be worth $200,000. That's what he told people. Well, it is an appreciating asset. That's what they said. Check out what you're getting for your car now.

Yeah, that again, totally outrageous. What happened in Autonomy Day? He told people that they were going to get all of those things. They've never got them. It's how he sold the car. And he's always been out there pushing this self-driving. When it doesn't work, he knows it doesn't work. On Autonomy Day, he came out and said, we now know, because I just read the chapter in Isaacson,

He told Grimes that they were about to go bankrupt and he told his cousin they were about to go bankrupt until he thought of what to do on autonomy day when he came out and he said the cars will go a million miles with annual maintenance. They'll cost 18 cents a mile instead of 62 cents a mile for your average car.

And oh, and so and you're going to get $30,000 a year when you buy my car from from us when we make it a robo taxi. They put that whole that that whole story together on one day and and and people believed it. He knew on that day because he talked to his cousin and he worked with the two. It's right in the book.

He knew it didn't do traffic lights. He told them to do traffic lights on in March of 2022 when he decided to hold that meeting on April 22nd of 2019. He told them get traffic lights done.

in less than a month. They didn't succeed. He knew that when he walked into that room and told investors that Tesla was going to be, that your car was going to go from $38,000 to $200,000 and $30,000 a year in your pocket.

He knew that it didn't do traffic lights on that day. And we now know that from the, from the I-60 book. It's not, it's a little optimistic. It's a completely made up story. That's why they tell you there may be forward looking statements and not make interesting decisions. He told, he said for sure 1 million robo-tanks by next year. He believes it. I have a new theory about Elon Musk. Well, try this one out.

Okay. Elon Musk words are just sounds that he makes to get people to do his bidding. That's all it is. He puts these sounds together and then people do what he wants.

That's it. That's how he works. That's not unreasonable, right? What I'll say is I'm glad that you're saying this and you are wealthy enough to handle any losses that Elon throws your way. I'm going to give no. I will not be being that direct. But no, I agree with you. A lot of people buy into his BS to say it nicely. That's probably my biggest...

annoyance is while I'm more optimistic about how good FSD is now and works in my experience seems to be better I agree with you I don't think I've had a single drive with zero interventions like I don't know anybody that says that I don't know what they're talking about it's not been my experience here in Austin um but it does work 90 of the time and that's really enjoyable it's just you need to be ready for that 10 where it doesn't work so I definitely enjoy it but

But I also, it's not to the level that Elon convinces people and sells people on. And he walks away with billions of dollars from people buying into that hype, which is clearly- $40 billion was his income in 2022. I mean, that's a cheat code for life that I wish I had-

I would say willing to say those things, I guess. I would say that certainly with something like FSD, right? That's one of those things, as Lewis put it, where there's a large possible solution space and will take a while to get to, but not to defend Elon because he certainly has his personality quirks to put it lightly. I do think though, there is something to that

Not that I would want to be managed that way, but to that kind of management that actually makes things happen. Here are things that are very hard. And by saying, oh, it's going to be done two months from now, the people working on it work really extra hard. And eventually, and it may take four months or may take four years, but eventually those things do happen. I mean, in terms of reusable rockets or profitable...

a profitable EV company. Um, and there are plenty of bodies along the way, but, uh, I think he ends up getting there. So this is one of those unsolved problems. So you don't necessarily know when it will be solved and the way they're trying to solve it. But, um,

I'm optimistic about it. And again, I'm happy with the product. I know you don't like this model, but I tend to think it being out in the wild makes the turnover faster, makes it improve faster. And again, just using the product, I see constant improvement and it's constantly getting better and better. It's agile. Yeah, exactly. It's the agile system, right? Cycles through. When do they put in the Do Not Enter signs?

You're an engineering manager and you've got a project. You've got to decide where do we do various things for self-driving cars.

Presumably one of those things somewhere on somebody's list is we ought to do do not enter signs. Like they put them up, but you don't drive the wrong way down one way streets or so you don't drive into crazy construction, right? You end up in a hole somewhere. That's why they put them up. It doesn't know what they mean. So it will gladly go anywhere. There's a do not enter sign that it thinks that it can get through. And that's what it does. They haven't programmed it yet. Where is that on the schedule? Should that be before releasing it to 400,000 customers or afterwards?

I mean, again, somebody is driving it, right? So if it's doing most of my drives and if I saw a do not enter sign, I would stop the car because that's how it's used now. That's not what he's selling. He's selling full self-driving and they haven't even gotten around to putting the code in or the training in to do do not enter signs. Why? How many more things are there I don't know about that they haven't even started on yet?

And they push it out. And also, there's a reason why you do this alpha site beta set thing is that they haven't finished development. Technically, they are not even alpha site. They haven't finished development, let alone alpha, let alone beta. And because they keep changing it in major ways to implement these new things,

The public is seeing a product and and you, I know you focus on the person who buys it and I get that. That they signed away their lives. I mean, you sign that paper or if you ever read it. You should see what it says. It says on a perfectly nice day, it might swerve out of control into another lane. Roughly there's that's in there somewhere roughly says that, but.

But it's not you signed it. You signed it. But the people who are not Tesla owners or FSD owners did not sign anything. You're testing a product for the benefit of a private corporation on public roads and putting people at risk. That's the problem, the bigger problem, that they never signed up. You might have signed up and you might be competent and capable, and you might screw up and it might hurt you, but you could also hurt other people.

And yeah, didn't sign up. And just because somebody wants to make a trillion dollars is not a justification for why they got to test it on the public without without permission and without compensation. And there's no point in it. It still has so many problems and so many bugs. They need to fix those problems.

that we know about, that I found, that the other YouTubers have found, that they found with their own books. Isn't it on a list somewhere to get around to all these things going past a school bus with the lights flashing? Why isn't that on a list somewhere?

It's just as a school bus, see the lights flashing and say, oh, let's stop. Wait for the flashing lights to go away and then we'll continue. It's not a very hard problem. Why can't they train it to do that? What are they training it to do other than that? I don't know. Well, they've been training it to

take curves better. When you pass a semi-truck, it'll give you a little bit of space. The way it accelerates, the way it decelerates, there are just a lot of little comfort things that I've just noticed that do a lot better now. And it may be prioritized for Elon's drive, but

It's the stuff that we're experiencing, I guess, 90% of the time. Those are the things that it's making better. And a school bus or do not enter are less frequent interactions. And so I would think that there are less priorities because...

because I'm still driving the car. Those are the things that need to be solved before I can go to sleep in the back of the car, but we're a ways from that. So I guess the real issue that we keep coming back to is the definition of scope of the product. Eventually they wanted to be able to do all those things. They can't do those now. They don't do a good job of

saying that it doesn't do those things now but i think most drivers do understand that so i'm fine with the product as it is if this was the product that it was and it's all it ever would be i would have it i wouldn't pay 15k for it i'd probably pay 3k for it maybe so if they just did what they call autopilot and they made it a little better that's a product i would buy and be happy with and again it's just fancier cruise control right so when will it do those things

I don't know. I'm not in a rush for those particular things because those aren't things I interact with on a daily basis. I would like it to recalculate routes a little faster, be better about getting over to the exit when there's one exit next to the other, better about changing lanes. But it's pretty good, but changing lanes, but sometimes it's not as aggressive as I would need it to be.

So those are the kind of things that I have to take over. The school buses. I mean, I'm going to beat their heads against that thing. I'm going to push their face in the mud until they fix it. And when they finally fix it, I'm going to have to say crap. No, I can't say that. I mean, I don't think it'd be hard for them to do. It's just not for us. Then do Elon. Please put an intern on school buses so Dan doesn't have to make Super Bowl ads. He probably doesn't go past the school zone on his way to wherever he has to go.

What you need to do, Dan, Dan, what you need to do is you need to get a school bus and you need to drive around where Elon is. I think that's your goal. That's what you do. Get a school bus. It's Gulfstream 650. He's never driving around. How am I going to find him? He does drive. There are ways, I'm sure. Nobody else would do this. Nobody else would do this.

Would not fix those problems. I mean, people say he has a thin skin. I disagree. I think he's got a skin thick as an elephant because I can go out there and show his product running over little children in a Super Bowl commercial and he doesn't do anything. Actually, it's not true. He did put a tweet out. He said, I'm glad now people will know that we have a self-driving car.

I mean, the searches did go up for Tesla. Let me ask you regarding that. I think Elon did give you guys a cease and desist at some point. What's the status of that? Did you cease? Did you desist? Well, sadly, it cost me real money because I had put together a campaign on that original. That was the what we call Will Springs. It's the first one where he showed the mannequins get run over. And I had bought a bunch of TV time for that. And and I

And then the day before my campaign was supposed to end, I just bought some, I can't remember what it was, one week or something of time.

for that commercial. And I got the cease and desist order on like the day before it was supposed to stop. And so like, I had to go out and buy another week of commercial time just to show that I didn't stop running the ads because he set the cease and desist letter on the day right before it was, I was going to stop anyway. So I had to spend like $2 million more to like continue the ad running for another week or two.

You should read my response to the letter. One of my more creative things I've done, it's pretty funny. But I called him, let's see what I said. He was hiding behind his lawyer's skirts.

I did read that. Happened to have a female lawyer at the time. His general counsel was female at the time. And so I thought that was fun. And I think I said I hood as widow feelings. Oh, and about being a free speech absolutist, too. I worked that in somewhere.

So it seemed like there was a lot of talk like three months ago when you did the Gerber drive and that's a while ago. So why are we talking about this now? Your team reached out to us. Is there any reason you're just trying to raise awareness about it again? Or like, is there some particular thing with the timing? What's that? No, it's constant. We're doing this constantly. We've got, we have this whole, we just did our, we just finished our driver monitoring segment, right? We did that for a month or two.

We have in the can the stuff on the driver monitoring with the Ford and the GM before we just did on Tesla. But then we decided to try that. So we've got some stuff written up for that. We've got video, right? We haven't decided when we're going to ship it. We've got several other projects going. And we're sort of continuously trying to keep it interesting, keep people listening and talking about this subject because he hasn't.

The thing they should do is fix the damn bugs and they're not doing it. They seem desperate. They did not follow up on their on suing us. I tried my best to, you know, to insult him, I guess, with my response.

When is the public going to learn that this is what he's doing? I don't think they will support this. I don't think they do. They just don't know because all he has to do is send out one tweet and 150 million people see it. And then every journalist on the planet repeats it and nobody else got to say anything.

And, you know, he's not quite where Trump was, but, you know, he dominates like every field now. You know, like, oh, I'm going to invade another country. I got to call Elon Musk and ask his permission to invade something. Yeah.

Why are you going? He's met more. He has G650 to more heads of state in the last couple of months than Biden. Right. Well, they all want a Tesla plant in their country. I got a new term for him. I will try this out. He's a bond biller.

Right. They always have these gigantic international organizations. Right. They've got factories everywhere and people everywhere. And they all have rockets like they all have rockets, like all the bond builds up. That's because space is cool. But no, I agree. We've got a complete. He's like, you know, it doesn't matter. He can do anything he wants. The government can't touch him.

no matter what he does, he's beyond the law, he's extraterritorial now.

Yeah, but I mean, that's, it's not just him though, right? There are other, I mean, arguments to be made for other folks in the same boat, Jeff Bezos and you know, others. No, no, no. They're nowhere near the same. They're like the Bonds. Elon accepted the Bond villain persona a long time ago. You can find photos of him stroking a cat and he's been Blofeld for a while. He's when he was launching the Falcon one that was on a private island, right? So he's like, Oh, I got a private island and my rocket. So he was totally you only live twice.

So that's not new. I think he enjoys that moniker because it is kind of cool, you know, for, for nerds of our generation. You can get people to work 80 hours a week and do exactly what you tell them to do all the time. You can do a lot. You can accomplish a lot.

Yeah, I love my Tesla. I'm a principal engineer. My wife's a principal engineer. Neither one of us would probably ever work at Tesla. Even though I love the technology in the space, I don't agree with the working conditions and how they do their software releases and some of that stuff. So I definitely feel you on some of that. I thought Hank Scorpio was a pretty good manager, though. He seemed to be pretty kind for a- Hank Scorpio? Yeah. Is it Simpsons reference, right? Yeah, yeah, sure. Yeah, yeah. You get a hammock in a hammock district.

Yeah. We got, we got it before we wrap up. We had a couple questions here that were directed at you that I want to try to get through sort of rapid fire style. Uh, here's one Ken Spencer asks, well, the recent F 35 runaway plane, a result of your software. No, that was a lie. It was made up by Walmart's blog and, uh, and, and, uh, uh,

Warren Redlich and I forgot the other guy's name. Corbin Williams. They made the whole thing up. They published it all over the internet, everywhere they could. They later on tried to back away from it. We've got all the tweets where they said, well, no, we didn't lie. We were just mocking you by saying that your plane crashed. We've also been told I crashed another F-35, a 787.

I've killed 30 million, those are millions of people. Right, the 737 Max, that was me too. We didn't do any of those things. - So you're admitting that here? - No. We didn't do anything. And they just put this stuff out. I'm blood on my hands, murderer, sick fuck, whatever. That's what they say about me all the time. - That's a bit much.

They ran out of arguments when they found out that their product actually does do all those terrible things I said it would do. And the only thing left to do is to try to bury me. They don't have any arguments. I said, come and watch. I will show you in front of the world. I will live stream it right now. And they won't come. The only guy dumb enough to come was Ross Gerber.

He took me up. He said, "This is great. It's going to be great. We're going to show Dan O'Dowd. It really is great." And then he disengaged three times, one on the freeway. The second time when he tried to go behind a garbage truck that was crossing the two-lane road and decided it was going to go behind it, he had to grab the wheel, accelerate around it to avoid getting his nice plaid mash.

And then about like less than a minute later, we blew past the stop sign at 35 miles an hour. This was one hour of driving, less than an hour of driving. Okay. All right. Rapid fire. That's how it goes. Nope. Then he believed all the propaganda. That poor guy actually believed all the propaganda about what it did and how great it was. And, and he decided to show me he was in the car. He drove. It wasn't my car. It was his car.

It did all those things in front of the world. Yeah. I will say not to go on too much about Gerber, but I did watch some video he put out about that incident and he described the Tesla as a level three system, but it's not a level three system. And right in the video, he says level three system is a system where

The driver doesn't need to be aware of the environment, but can intervene when necessary. I mean, the current software, you definitely need to be aware of the environment. All right. So the next question we had here is, do Cruiser Waymo use a GHS? I assume that Greenhill. Yeah. Does it, does Cruiser Waymo use a GHS platform to run their self-driving software? I don't know the details of what they do. I know we work with

GM on various projects. I don't know about Waymo. I mean, I don't know what we run on. It's not a big customer if it's a customer at all. Okay. The last one, we sort of talked about cruise a bit, but Hugo asks about those cruise cars in San Francisco that keep blocking traffic and getting stuck. What's wrong with that system? Well, the funny part is, of course, as I told you, we went out and tried the cruise vehicles.

And they work well. I mean, again, we only ran, I think it was three times, three or four times we took the vehicle or we took the cruise. So that's not proving anything, but there are several hundred cruise vehicles and they're driving hundreds of miles a day around San Francisco. So actually, if you start calculating out how many miles it is between failures, it's something, what's that, 300 miles,

300, 200 cars times at least 100 miles a day. They were basically running through continuously 30,000 miles a day, something like that. So if one of them gets stuck every day, they only get stuck once every 30,000 miles or something like that.

Well, the test can't go 10 in a city like San Francisco without screwing up and they can go 10,000. Okay. But to bring it back to cruise, I think some of the issues they're talking about is sometimes they cluster, like we'll have a bunch of them and they'll basically...

be waiting for the other one to move and they just clog up an intersection. And something similar happened in Austin. I think the failure they related it to was that they all needed cellular internet and it was after some sporting events. There's so many cell phones in one place that it overloaded the system and that the cars just got stuck where they were because they rely on the internet.

It is not perfect. Absolutely. But when you count up the number of errors they make per thousand miles, it's way less than FSD. It's not 10, not it's like 100 times or 1000 times, not three times or 50%. It's an immense difference.

Because there's running hundreds of cars. Yes, they foul up. Yeah, we know that. They need to fix those bugs. They've got bugs. They actually fix the bugs. When they get a bug, they go nuts. When one of these bugs comes out and it does something like that, the whole place goes nuts. And they say, we've got to fix this right away. And they rush engineers in to figure out what went wrong. Yeah, right.

Exactly. And they fix it. That's the difference. Tesla doesn't do that. When one of those problems happens, they just continue doing something different, not fixing a serious bug. So and so it does happen. But the actual rate at which it's happening, if you do the calculations, is not is way better than Tesla. Thousand times better, not 10 times better.

And they're trying to make it better. It needs to be better than that. Tesla needs to get all the way up to there to not be good enough. And they're not even close and they're not getting there very rapidly either.

Okay. Again, I don't exactly agree with you. I think Tesla does seem to be, I mean, they're not solving the problem you want them to solve, but they are iterating, I think, fairly quickly. And they do seem to respond to the problems that actual customers alert them to. Other than running over children and school buses. Yeah, but we're not. Speed limits when they say, when children present, 25 miles an hour, it blows through at 40. Yeah.

It's why not fix those problems? I'm going to keep doing this until somebody at Tesla finally says, you know, we probably ought to fix those horrible bugs. Then I'd shut up. I'd be done.

I suppose you could sit in the corner and have your mannequins ready and just toss them in the road as people drive by and see what happens. It's a real world test. Yeah, that goes. But we don't do that. We give the car plenty of time when we're in the car and you're in the car with us and watch it. We say, see, you see the kid, right? You could stop right now, right? One second later, you could stop right now, right? One second later, you could stop right now, right? Boom, you hit the mannequin. Yeah.

He'd had like four or five seconds that it could clearly see the mannequin. It could have stopped and it didn't. The human driver would have stopped every time. That's the problem. Why doesn't it see them? Why doesn't it?

Maybe the software is so advanced, it knows that they're not really people. And that's, no, I'm just kidding. Sorry. That was a bad joke. Many people have made that argument. Many people have made that. Well, that's a ridiculous argument. So hopefully. It can test the blood pressure with its cameras. It doesn't see the eyes, right? So then it doesn't think it's a person. There you go.

Oh, man. I just want to say it was great having you on the show. I appreciate your time. We all do. Again, we don't necessarily agree with you, but it's still good to have these discussions and debate. And I agree with you with many things, just not all things, obviously. Hopefully our viewers and listeners, which will have hopefully many...

We'll appreciate some of these things. And again, thank you for being here. Thanks to all our viewers, subscribers, followers, all that kind of stuff. Keep the love coming. Keep the hate coming. It fuels us. And...

Dan, do you have any last words or sites or things you want to plug as far as like where people should find out more? We can put some links in the description and stuff. We've got a ton of videos on dawnproject.com. It's not with the V-A-W-N project.com. There's tons of videos, our videos. There's tons of YouTube videos that we've collected that show many things, lots of fun things you can learn there if you really want to learn about all this stuff. So thanks a lot for having me and...

You know, I'm working on spreading the word and get them to fix the bugs. Please, if you have some influence over them, please just fix the bugs and I'll shut up. Great. And I think I'll have to add the disclaimer. The opinions of the guests don't necessarily reflect the thoughts and opinions of Tess the Motors Club or its hosts. That's always true in both directions. Don't sue us. All right. We'll see you next time. This was episode 53. Thank you all.

Laters.