Interview with Amit Mehta on the Future of Auto Tech

Hey everyone, this is Ian Peterman and I'm sitting here with Amit from North American Lighting. And welcome, really good to— happy to have you here. Excited to talk about things you're working on.

Thanks, Ian, pleasure to be here.

Yeah, and we've talked, talked a little bit before and you guys are doing some really cool things with Making light bulbs for cars, headlights that are way better than a headlight. So, right, I'll let you, I'll let you dive into that.

Yeah, thanks for the introduction. I think it's a little more in-depth than that, but yeah. Um, so I work for a company called North American Lighting, which is a fully owned subsidiary of the Koito Group. It's the largest tier 1 manufacturer of headlamps and rear lamps in the world. We have about 21% market share. So 1 out of every 5 vehicles you see on the road, that's us. Our biggest customers include Toyota, Honda, Nissan, Ford, GM, and all of their subsidiaries, and many other OEMs around the world. So what we've really been doing is creating headlamps for the past 100 years or so, 'cause that's how old our company is. But now we see this movement into ADAS, and autonomous driving. So ADAS being level 0 to 2, where the car helps you brake and steer, or brake or steer, and then eventually where you get to the point where you have level 3 to 5, where the car automatically drives by itself. Of course, with some human intervention at level 3 and level 4. But we've been really concentrating on how can we add more value to the headlamps. I think you're, you're kind of familiar with The vehicle moving away from what it is today, it's more of a product. It's also internal combustion engines into something more sustainable, such as electric vehicles, something that's better for the planet in the long run. And as you look at that vehicle and it changes from a discrete model where it takes you from point A to point B, and you think a little bit more about it, how we're moving into electronic vehicles, Think of all the components inside the engine that are needed to make a vehicle run today. And now think of all those components gone because now you're using an electric vehicle, which is just battery operated. I like the word. Right, exactly. So now think about that, but think about what you will need on a car no matter what, and that is lighting and sensors. And that's where we're making a major play since we already have the largest team in the world, 23,000 employees plus, and that really specializes in optics and manufacturing and scale. Now we're using that asset, using those resources to bring the correct technology that's needed to make a car from a product into a solution.

Yeah. Well, and there's, like you said, we're simplifying, the engine space is evaporating. It can be, as Tesla's done, you can end up with a front storage area because there's Nothing there. There's no engine compartment anymore. But yeah, we still have sensors.

We still need—

I mean, more, more and more. Like, we at the level we're at right now, you know, just looking at like a Tesla level of autonomy and driving, that's a certain level of sensors. So are you looking at like beyond? You know, you mentioned the different grades of autonomous driving.

Yeah.

You know, what are you like planning out? Like, you're gonna be able to put into your headlight enough technology to make it so that full autonomous, it'll have the sensors? Like, what, what does that look like?

Yeah, if you think how a human operates, right, we, we have our eyes and we use those as our main sensors to see what's ahead of us. And then we, we actually capture that image, but we have to learn what that image is, correct? So the first time you and I ever saw a kangaroo, we're probably like, weren't like, that's a kangaroo. We slowly learn what a kangaroo is, right? So we were able to build that neural network, network inside our head. So our eyes, which will be located in the headlamps, help us determine what's in front of us. We see a reflection or an image that comes back to our eyes and then gets transported to our brain, which actually determines what exactly that is. So that's pretty much how the car will work. We'll have the headlights, which of course will have the functions of headlights. Such as the LED lighting, the turn signal, the high beam, the low beam, et cetera. Then it'll have the sensors built in. LiDAR, radar, and camera are the major sensors used today, which will help the brain of the vehicle, which will be the GPU, kind of determine what it should do. Should it— is that a kangaroo or is that not a kangaroo? So it's kind of the same thing as using how the brain or the human body works and transporting that into a vehicle.

Right. And as part of your headlight and building that up, are you also a part of figuring out how to process it? Are you kind of, are you looking at that whole process channel or are you guys just focused on making the sensor package work, you know, in the form factor, which is pretty small?

Yeah.

You know, what kind of are you looking at in that whole chain of, you know, you say all the way from see it to actually process it and do something?

That's a good question. So when you think of the vehicle itself in the future, it's gonna have 3 areas. It's gonna have perception, where you actually see the image, then planning, what do you do based on seeing the image, and control, how do you control the vehicle to make those decisions? What we're really trying to play in right now is the perception stack. So we're providing the sensors And hopefully some sort of neural network that determines that's a cat, that's a kangaroo, that's a car, that's a bus, to the OEMs, so the original equipment manufacturers such as Toyota, Honda, Ford, GM, and they can control the next steps, which are planning and control.

Got it. So you guys are really just focused on the seeing part, making sure what is it identifying, giving the information in a way that's that's really usable by someone else. And then, you know, they, they can worry about what, what do you do when you see the kangaroo? And that's right, that's right, that's theirs and determined on, you know, what they're trying to look for.

Right.

And you mentioned kind of you're, you're building, so it feels like you're building a bit of a tech platform almost because headlights are so integral, we already need them, and now you're just You're leveling that up. You're saying, hey, we have lights, let's add some technology. Let's bring, bring a 100-year-old part into the current-day technology realm. And is that something that you're building into a bit of a platform? I know you talked a little bit about startups and working with them. So how does that, how does that work for you, like, moving forward with— Sure.

Yeah, so sure, I wouldn't say that the headlights are 100 years old. I mean, you're not bringing new technology into a 100-year-old item. It's progressively and incrementally improved over the years. And to show you understand how that's happened, the 80 years from 1915 to 2000 or so, the 85 years since Koito was established, the headlight was pretty simple. It was basically 15 components. You put a bulb, a reflector, and some wire harnesses and some other components all together in a housing and package it onto a car. But then something amazing happened. It's the invention of LED lighting for automotive applications. And what that allowed us to do is make an image out of the vehicle. So I'm pretty sure if you're driving in LA that you can easily recognize a BMW at night, you can recognize a Porsche at night, you can recognize an Audi at night, Mercedes, because those images have been created by these LED lightings. So that particular 115-component part is now 160 components in that same form factor. Because when you think of LED lighting, it gets very hot. You need heat sinks, you need fans, you need a whole bunch of screws, you need a lot of things in there to make it work correctly.

Right.

And now you can actually drive a car that suits your image instead of just 2 bulbs out there. So our form has always been around safety, and then we moved into branding, 'cause now you can brand your vehicle, right? You really look at the car and you say, oh, that's my Porsche, that's my Lexus, that's my Mercedes.

Right.

But now we're looking at how do we go to the next step? And the next step really is incorporating beautifully functional sensors inside the lighting. And the only way we can get there, 'cause there is a lot of momentum around ADAS and autonomous driving, is work with these startups that have come out of Silicon Valley or around the world actually to develop this technology and bring it inside the lighting. And some of the startups we've worked with or working with right now is— first one is a startup called Septon. They make a LiDAR technology solution. In February, we invested $50 million into their Series C. So we, we think that their technology is very good and it provides a very beneficial solution for the future. Another area that we invested about $25 million in last year is a company called Brightway Vision. It's based in Israel. It's a gated camera technology, so it takes slices of an image in edge-case scenarios such as adverse weather, fog on the Golden Gate Bridge, for example. Sometimes you can't see going from point A to point B on the Golden Gate Bridge what's ahead of you, but this camera allows you to see that. So these types of sensors and solutions will be required to go from where we are today to a fully Level 5 autonomous solution. And our goal has always been safety. So our value proposition is always working with safe technologies that can provide safety to not only our customers, but our customers' customers, which in the end result is you.

Right. The actual drivers.

Exactly.

Now, you mentioned that, you know, the headlight hasn't— didn't change very much for 85 years, and then it had this leap with LED technology. And now you're approaching this next leap here. Has— at least when I think of like LiDAR systems and all that, I don't think of something that fits inside of a light bulb. So is that— are we just At that point now where form factor has been a primary limiting component to this? Like, what is making now the time you're like, all right, let's put money into this, we can actually do it?

It has. It's been a big limiting concern in the past 4 or 5 years. But as you work with these startups and as you work with these OEMs and tier 1s, you kind of work together to work on the major problems. What are the major problems? Number one, it's a C-SWAP problem always. So it's cost, it's weight, it's size, and it's performance. And we're continuously working on those 4 areas to make sure that we can bring scalable automotive technology solutions to the mass market.

Right.

Have you guys—

is power at all an issue? That's a very good question. Especially now that we're moving away from engines Everything's run off the same battery now.

That's a good question, and that falls in performance. So we understand that power is going to be a big concern of ours. We've been squeezed on the power budget over and over every year. I imagine as we add more and more technology, yes, we're going to continue to be squeezed on this power, but that's solutions that we're continuing to work on to make sure we actually deliver to our OEMs what they want.

Have you, just in working through this, have you seen, you know, if— is power or size been a bigger hurdle for you to jump through in order to get to where you have a working product?

Today it's size. Size is a big issue because as you kind of come down in the size of a product, you give up— you have trade-offs, right? You might have a field of view trade-off, you might have a A limitation in the optics that you can provide. Whereas power, you can work on going from, let's say you start off using an FPGA, after you have a good solution, you can go down to an ASIC, which is half a watt down from 5 or 10 watts. So these are kind of solutions that are easier to solve than the form factor issue for today.

Interesting. And is that, you mentioned a couple of startups that you guys are working with, is that kind of some of their secret sauces? Is it shrinking down the technology or—

Yes.

Okay.

It's not really their secret sauce. So when you first come out with a sample or prototype, it's usually very big. You just want to have a proof of concept and say, okay, this is what I have and I want to present this to the world. And then someone will see it, whether it's an OEM, a car manufacturer, or a tier 1, and say, you know what, our engineers can figure this out. We understand that the solution you've provided works. But let's go to our A sample, let's go to our B sample, let's see what we can do to cut down the cost, cut down the size, increase the performance, and cut down the weight of this product.

Right. And is that something that you guys, like, is that kind of your expertise that you bring to these startups in a way? Is like, you got a great idea, the core technology is there, now let's make it, make it something that someone can actually install on a car.

Right. So that's—

And afford it.

Exactly. The 2 most expensive components of an automotive product is qualification and validation. So it's not really the technology, because a lot of companies spend a lot of money on R&D. It's to bring safe, reliable, validated components to the mass market in a scalable technology. So it's kind of, for example, it's easy to make money on a Porsche when you sell it. Because it's a very high-margin, expensive car. But when you try to do it with a Corolla or Camry, it's very difficult because your process has to be fully controlled for 1 million, 2 million vehicles, and you have a finite amount of time to deliver those vehicles. So that's what we bring. We bring scalability, we bring time to market, and we really bring the qualification and validation to the startups.

Right. You've mentioned safety a few times, and I know it's a huge part of it. So what When you're building these sensor packages, and you mentioned a camera that can allow us to see through fog and actually get what could be even better than human vision, I would imagine, if it can actually slice through.

Yes. Yeah. You can see past the fog. Yeah.

Right. So we're at that point, we start getting into something that's safer than possibly a person. And is that where, is that Is that where you're really aiming at, where you're going to be able to supersede human ability to drive and have an autonomous vehicle that can really do better than a human at driving?

You would think that, you know, humans are kind of very good at driving already. So if you take out the number of drunk drivers or drivers that text, the accident rate is actually very low.

Okay.

Right. Humans are actually very, very good at driving, but there's things that we've allowed them to do to get in the car that, that they do. And it's just, you know, it's unfortunate, but it happens. Will the computer become better than a human? That's a good question. I don't think it's— I think the question more is, will the computer become exponentially better than the human, because if a computer were in— were to get into an accident that a human would get into as well, the human has no empathy for the computer, but the human would have empathy for another human being. So in order for autonomous driving to really take off, it's not about the regulations, it's not about the technology, it's going to be the consumer perception of these autonomous robotic vehicles. controlling their lives, taking them everywhere they want to go. And it's very similar to a plane, right? But you have a— obviously you have a pilot there, but 90% of the time, except for takeoff and landing, that plane is on autopilot.

Right.

So it's pretty much the same situation, but now you can see the pilot not doing anything. So that becomes a little scary, right? You're sitting in the back seat and this guy's sleeping and you're like, okay, I don't really feel comfortable. So it's going to be a perception-based solution where We need to make sure we educate, we inform the consumers and the mass market that this is a safe, reliable vehicle over countless validation and countless qualification miles.

That's so interesting that the human safety is actually higher than we all think. If you remove those 2 elements, which are, you know, really not safe to do while driving, So that's interesting because it kind of makes that bar a lot higher. And that's not really— that's not really actually even the conversation of can you be as safe as a person. It's can you as a person be okay sitting in a car that drives itself and accept that and feel safe?

It's similar to bridges, right? So when Carnegie first built the first bridge, nobody would actually walk over the bridge. So he had to go rent an elephant, which was quite expensive back then, and show them that the elephant walking over the bridge was safe. And then they finally took the bridges. So we'd have to— we have to prove it, that it actually works over and over again before the humans will actually get in the car. Right.

Yeah, I remember that story. I'd forgotten about it. That's a great analogy.

Yeah.

Um, is there— when you guys are working on this technology, do you have a What is your roadmap in looking down? Do you have an end goal with it, or is it just— do you guys just see this as a continual improvement for eternity and there's no road marks to say, okay, we've accomplished a vehicle that our system can help?

Yeah, there's a lot of things going on here. So first of all, we're looking at electric market, then we're looking at autonomous market. Then we're looking at connected market, then we're looking at shared market, and all these things are going on in parallel. It's not that one and then we finish that one, then we go to the next one. So what we're really trying to do is we're trying to compete in 4 different markets and see how lighting provides value in those 4 markets. If we look at the autonomous area, we provide value because at night some of those sensors such as camera need the visible light to actually see. what's ahead of it. If you look at the electric space, we're providing value because we were needed. As you remove all the components of an internal combustion engine, you're just going to have basically the lighting components and the battery, the inverter, and the drivetrain. So we provide some sort of solution that you're going to need those sensors and the lighting to see at night as well. If you look at the shared solution and the connected solutions, We don't know how we play in those areas yet because the business models are developing. So we're trying to think, okay, can we provide a subscription-based model for lighting? Such as, do you move into— if you're in the hills in Europe and you want to buy a technology, subscribe to a technology called ADB, it's already in the market. It's called Adaptive Driving Beam. It's always in high beam mode, but when it sees you coming because of a camera technology, it can dim just your area to low beam mode. So it selectively dims a very thin fraction of the full scene into low beam and keeps everything else as high beam. So that gives you an extra almost 70, 80 meters of visibility.

Oh, wow.

And right now you can't get it in the United States, but you can get it everywhere else in the world just because regulations haven't approved it. But if you think about it, in Europe, you don't need it every day. What if it was a subscription-based model? So you can use it sometimes and not use it sometimes. It would just require flashing the software and turning that on in a vehicle. So we're looking at how we can play in different areas, and we don't know what the full roadmap is. We, we want to make sure that we account for every single risk that could take place, and that, that's basically how we're approaching the next 10 to 15 years.

Yeah, well, I, I—

that makes a lot of sense, especially in the, the shared space. And even when you say shared, one that popped into mind is, you know, basically ride-hail companies who are looking to develop cars and the additional level of technology that they would probably want embedded into their car almost could end up being— they need more connectivity, right? They need more because now you have a driver service, or if even looking at taxis now before we go full autonomous, right? There's additional layers that might be Might be worth having that seem like could fit into this, this form because you're already there, you already have a headlight. And I think that's probably one of the strong suits that you have is we already have this form, we already have this product, it's going— it has to be installed, it's got to be there no matter what. And it's easier to kind of add on and augment that, especially since we're losing engines.

Right.

It seems like such an easy place to pack it in and already has power going to it, like it's already Well, well situated to go that route. In your kind of what you're building, is it— are you looking outside of just that area as well for technology, or are you just staying purely there? Are you doing like, you know, side markers, turn signals, like every kind of lighting, internal lighting areas, or are you sticking to just the front and back?

Yeah, we don't do internal lighting, we just do exterior automotive lighting. But we're not just looking at, you know, the solutions are going to change, lighting is going to change, it's going to evolve and adapt. When we go into the autonomous driving mode, there's going to be a requirement for lighting to play a bigger role because there's going to be a time when there's a hybrid mix of regular driven vehicles, so L0 to L2, and autonomous vehicles. And the only way for the consumer, the driver, the pedestrian, the bicyclist to know at night what mode the vehicle is in is through lighting. So we need to provide that information, that communication to cyclists, to pedestrians, to other drivers to make sure they know which vehicles are in autonomous driving mode or not. And that's where lighting is gonna really play a solution. It's really gonna play that role in other areas too, because as you get into autonomous driving modes, there are a lot of cases out there That people have adverse health reactions in a vehicle, such as a stroke, a heart attack. And the only way for others to know that person is actually having this happen to them in the vehicle is if there's a camera inside the vehicle that monitors the vital signs of a driver or passenger and then relays that to the outward world and lets them know that this person is having a minor or major health scare, and we're gonna flash these lights and take this car to the hospital autonomously. So these are solutions, some of the solutions that autonomous lighting will play in the future.

Yeah, that's, that's like a whole realm of communication, you know, is the vehicle being allowed to communicate with people outside of it, which We definitely don't have now, except for turning, right? You can, you have your brake light, you have your turn signal, but there's no other real communication method to let somebody know what's happening or what's needed built in. So that's, it seems like a very interesting kind of, kind of road to go down. And I've talked to a couple people that are on the kind of the beginning and start of building smarter cities and having those more connected cities with being able to communicate. So I would imagine that that kind of information would be valuable to have inside, especially a city where you have a heart attack and you're driving. That's a user, right? Right now it's, well, you're probably going to be in a pretty bad wreck because the vehicle's not doing anything and nobody knows. And so this is Like a huge— I think it'd be a really great, great piece to have for people.

Yeah, another area that we play in right now, it's not a huge business for us, but we have 30% market share of infrastructure lighting in Japan. So that includes traffic lights and streetlights. So we're using interchangeable technologies such as LiDAR, radar, and camera and putting in our real estate to kind of develop solutions for cities to let them optimize traffic, to provide data to cities to say, should we build a mall here? Should we build housing here? To let them know of any safety alerts that they should be aware of. So this is some solutions that we're actually playing into today in Japan.

Interesting. Are you guys seeing that going outside of Japan? Are you building it as a platform? Is it kind of— what is your goals with that?

Our goal is to take it outside Japan, but the one thing you have to remember between automotive business and infrastructure lighting business is the scalability. So when we provide a product to Toyota, it's easily scalable across the world because it's one company who owns all these vehicles and they can easily sell it around the world.

Yeah.

But when you think of, you actually think of infrastructure lighting, it's city by city. So city of San Jose versus the city of Cupertino versus the city of Palo Alto, it's very difficult to make those connections and make sure that you can can scale that solution. So that's an issue. There's no standard for what solution will be used in different cities.

That's so interesting, which I guess you would notice if you drive through, you know, lights, traffic lights aren't all the same everywhere. Lighting isn't all the same. So interesting.

Interesting.

Yeah.

Well, I guess, is there anything else you'd like to kind of add just to To wrap up here.

Yeah, I think one more thing. If anybody's really interested in the mobility aspect and learning about technical solutions, I host a meetup once a month called Sunnyvale Sensor Fusion. We have 1,500 members. A new startup presents every month. They present their technology on a technical standpoint that's used for either autonomous driving, ADAS, or some sort of smart city solution. So if you ever like to see that, Please look up Sunnyvale Sensor Fusion Meetup on the Meetup site.

Wonderful.

That's a great— that's a great place. Uh, you guys get 1,500 people, so you probably get a lot, a lot of people that want to pitch to you there.

Yeah, we have a lot of pitches, but we're mostly interested in the— we leverage the community of Silicon Valley and the world now with, with everything online to kind of get these PhDs, these experts to ask the right questions during the meetup. So we can vet the startup very easily during our meetup because we have these really high intellect people who really understand the technology. So that's kind of the solution for us. And we used to have beer and pizza and just talk about technology openly, but unfortunately now we're stuck in our houses.

Well, hopefully soon we can have beer and pizza again and be able to talk about it.

That's right.

Perfect. Well, I really appreciate your time. This was great, great learning about what you guys are doing. And so there's so much we can go with the technology that you guys are developing. So it's really exciting to see that part of the future of autonomous vehicles.

Thank you. It was my pleasure to be here today.

Awesome.