Are Tesla's Promises of Self-Driving Cars Risky for Investors?

Tesla (TSLA 15.31%) has taken a challenging path toward autonomous driving that is a risky bet on its unique technology. In this Backstage Pass clip from "The AI/ML Show" recorded on Jan. 12, Motley Fool contributors Toby Bordelon and Trevor Jennewine discuss the strengths and weaknesses of Tesla's approach to self-driving vehicles. 

Toby Bordelon: I have a Tesla. I have a Model 3 that I got in 2018, so it's almost three years now. When I first got that, that was before they had the current full self-driving chip. Until they gave that and then it went to free to customers, they would upgrade the car to anyone who bought the full self-driving package, which was substantially cheaper back then. Do you have any sense if they said whether or not they're going to do that for the second-level chip or is it just you keep the first one because we think that's good enough?

Trevor Jennewine: I haven't seen anything where they said that they will upgraded like they did in the past. I'm hesitant to make a guess on which way that will fall. Musk did say recently that he believes that the FSD 1 chip will still be, I think you said 200-300 percent better than a human driver, were the figures that you use. He still thinks it will support full autonomy. But like I mentioned, they said the same thing about the previous chip they had in the car and I don't even think Nvidia's chip was meant for full self-driving. I don't think it was meant to support full autonomy. We'll see if it turns out that the chip won't handle full self-driving, I would assume they would replace it. I know you're a Tesla owner and I was hoping you'd shed some insight in that way.

Toby Bordelon: Yeah, it's an interesting challenge, I guess I would say because they're selling full autonomy right now, they're claiming that what you buy will do that. They put themselves on a mission where they can't really release a new version even with new hardware, and then say, OK well, I guess it sucks you bought them beforehand, but this is the real thing, because you sold to those people, you sold them this promise that eventually you're going to get that, where you have a hardware, you need to do it. They have to make good on that. If they don't, I think the solution is, if they need to upgrade the chip, which is doable pretty easily, or refund the money.

Now, the risks here for them too if it turns out this camera based system doesn't actually work and they can't ever get there with that and what do you do. Do you go back to all your customers and say, oh, those cars you bought that were full autonomous ready, I guess not. That's a business risk. There is not a lot of technological risk, but it's a business risk because they've put themselves in a position where they saying this will work eventually. It's just the software issue.

Trevor Jennewine: No, absolutely. I think you're right on and Tesla's approach is, they've taken their own path and they've built their own chips for the data centers, they build their own in card chips. They're taking the vision-only approach, so I think that's it's definitely a risk and I would hate to be in charge of the company if they realized that it wasn't going to work, I think that would be a PR nightmare.

That being said, when you think about it in the fact that you are only using your eyes to drive and you're able to do it. It seems like eventually we will be able to get software that has that decision-making capacity to do it with just cameras. If you can do it as a human, it seems like the software could eventually do it.

Toby Bordelon: Intuitively, it makes sense and there is this idea, that a lot of other companies are working toward a world where we have autonomy with a lot of different hardware, not just in the car but in the streets, in the street signs. With repeaters and city just stuff external to the vehicle. In some cases, depending on other vehicles also being outfitted with sensors and transmitters that can talk to each other. That's an ultimate long-term vision I think.

But Tesla's approach is saying, "You know what? We're taking world as it is," and right now, highways and roads are designed to be driven with looking at the signs, looking at traffic signals, looking at what's around you. This is an interesting approach, but it does make some sense. It doesn't make some sense. If you can make it work, you can get there without reliance on anyone else, without reliance on governments to upgrade things, without reliance on third-parties. You can just build the car, and it works. In theory, we'll see if they can do that.

I will say from my personal experience, I just recently got the full self-driving beta. There's issues with it. A lot of false negatives, a lot of slowdowns, a lot of don't quite know what to do with this. A big issue being those flashing lights in school zones, but they're off at night or when schools aren't in session, but the car doesn't quite know when school's not in session, it sees what it looks like a stoplight that's not lit up, so it doesn't know what to do. It starts to slowdown down what are you doing on like you shouldn't be slowing like that. You're going to get me hit. There are some issues with that I think needs some work. Needs some things to be worked out.

Trevor Jennewine: Yeah. I've read some issues with like the phantom braking that they're talking about and I think if I were to sum everything up, I'd say that Tesla's approach is more difficult, it's definitely risky. If they get it right, they're going to be miles ahead of everybody else. But if they get it wrong, it's going to be a disaster.

Toby Bordelon: Yeah. I think that's a fair assessment. We felt like the issues I'm seeing so far, I don't think they are related to cameras and certainly it's more of a software processing issue. That's promising in a sense that that can be in theory dealt with.