Future Battery Tech with Electra Vehicles Founder Fabrizio Martini & Autonomous Solutions with Mobileye's Prof. Shai Shalev-Schwartz

MotorTrend

MotorTrend's Ed Loh & Jonny Lieberman take the show on the road to CES in Las Vegas where they chat with Electra Founder

Transcript

00:00:00 (dramatic music)

00:00:02 - Hi there and welcome to The Inevitable.

00:00:18 This is Motor Trends podcast about the future of mobility,

00:00:21 the future of the car, where we're going,

00:00:22 how we're gonna get there and new,

00:00:25 this vodcast is brought to you

00:00:27 by the all electric Nissan Ariya,

00:00:30 inspired by the future, designed for the now.

00:00:33 - And that's great.

00:00:35 My name is Ed Lowe, I'm the head of editorial at Motor Trend

00:00:38 and you just heard our very first sponsor read.

00:00:40 Because we have a new,

00:00:42 all sorts of new stuff going on this year.

00:00:45 You saw it in the intro, we have a new sponsor.

00:00:47 - Thank you Nissan.

00:00:48 - Thank you Nissan, thank you Nissan Ariya.

00:00:50 - Do we get free GTR?

00:00:51 Anyways.

00:00:52 - Yes, that's what I alluded to

00:00:54 in changing up the graphics.

00:00:57 But this is gonna be great.

00:00:58 We're very excited to welcome our new sponsor.

00:01:00 You're gonna hear that read quite a bit going forward.

00:01:02 And we have some new features in the podcast.

00:01:05 We are gonna do, give away some swag.

00:01:08 We'll tell you about it in a little bit.

00:01:09 - Look at this, we'll give you this cup.

00:01:10 Look at this, we got hats.

00:01:11 - Hats, we'll tell you about it in a little bit.

00:01:13 How we're gonna do that.

00:01:14 - I'll mail you Ed's shoes if you want.

00:01:16 Just, but you gotta do some stuff for us

00:01:18 and we'll get to that.

00:01:19 - Yes, we'll get to that.

00:01:19 And we are, you know, we want you guys

00:01:22 to be invested in this, so we're gonna answer

00:01:24 some of your questions more directly.

00:01:25 Whenever we go live.

00:01:27 - Is this our new segment, question of the week?

00:01:28 - Question of the week.

00:01:30 You drop us, you've been dropping us questions

00:01:32 on our Instagram.

00:01:33 You occasionally drop questions in our YouTube channel.

00:01:36 Keep doing that, let us know.

00:01:38 Send us your questions.

00:01:39 And we have one actually.

00:01:41 We seeded the field a little bit.

00:01:43 And our pal, Dr. Bobby Bones, Bob Cho,

00:01:46 who's actually the chief of staff

00:01:48 at the Shriners Children's Los Angeles Hospital.

00:01:51 - Like the most impressive--

00:01:53 - Human on earth.

00:01:54 - Magnanimous person I know.

00:01:55 I actually feel like just guilty

00:01:56 for being such a horrible man whenever I'm near him.

00:01:58 - Yes, Bob Cho asks us a question.

00:02:01 - And a GTR.

00:02:01 - Yes, he has a former GTR.

00:02:04 - Which has nothing to do with the Nissan thing.

00:02:07 - Anyways, Bob Cho asks us a question.

00:02:08 When do you think we will go to a single charging standard?

00:02:12 VW doesn't seem to want to give up the CCS plug

00:02:14 and I've seen more broken Electrify America stations

00:02:17 than working ones.

00:02:18 So it's a great question, as a lot of people know.

00:02:22 And we've covered this on Motor Trend.

00:02:23 - But we could also do an entire show just on this question.

00:02:26 - Yes.

00:02:27 - But to keep it brief, Volkswagen announced,

00:02:28 since Bob asked that question, that them, Audi, and Scout,

00:02:31 the upcoming all electric truck and SUV brand

00:02:33 from Volkswagen, they're going to NACS,

00:02:36 the Tesla Charging Standard.

00:02:38 - It's called North American Charging Standard.

00:02:40 And like over a dozen manufacturers

00:02:43 have decided to abandon the CCS plug, the SAE CCS plug.

00:02:48 The round one, if you guys know what we're talking about,

00:02:51 that looks kind of like a ray gun

00:02:52 and it's got a little clip on the top,

00:02:54 in favor of Tesla's North American Charging Standard,

00:02:57 which has no moving parts, it's slimmer,

00:02:59 it's generally considered to be the gold standard

00:03:02 for charging, at least in North America.

00:03:04 To his question, when do we think we will all get there?

00:03:08 That's a good question.

00:03:10 There are some limitations to NACS, as we understand it.

00:03:12 They're not as fast or as high output as the top end of CCS.

00:03:17 - So CCS, in theory, can go up to 350 kilowatts

00:03:21 into the vehicle, although you never really see that.

00:03:24 But NACS, there is a new one that's coming

00:03:27 for the Cybertruck and I saw some new stations installed.

00:03:30 Yeah, so when?

00:03:32 2027, I think is a good guess,

00:03:34 when you'll stop seeing a lot of CCS stuff sold.

00:03:37 But it's weird, 'cause like currently at my house,

00:03:40 you know, I drive a lot of different cars,

00:03:41 I have a CCS charger for my truck,

00:03:44 and I've been thinking like,

00:03:46 boy, I'm going to start getting press cars that are,

00:03:48 you know, I won't be able to charge at home.

00:03:50 That kind of like makes EVs terrible

00:03:52 if you can't charge them at home.

00:03:53 So when I put in the charger,

00:03:55 I actually had two wires drop,

00:03:57 'cause I wanted to put in two chargers eventually,

00:03:59 'cause I figured one day I'd own two EVs.

00:04:01 But now what I think I'm going to do is an NACS charger

00:04:04 and a CCS charger and keep it that way

00:04:06 until it all converts over to NACS.

00:04:08 - Yeah, there's a lot.

00:04:10 So Johnny's talking about it's probably like level,

00:04:13 so it's higher level thinking.

00:04:17 You're going to see a lot of this transition happening

00:04:19 basically now.

00:04:20 - That's the nicest thing you've ever said about me.

00:04:23 - Hang on.

00:04:23 All the OEMs, Ford led the way with the announcement

00:04:27 of going to North American Charging System, NACS,

00:04:29 GM has followed suit.

00:04:30 It's all happening literally right now.

00:04:32 Like February, March is when they start to roll out

00:04:34 these adapters that allow you to use an NACS charger,

00:04:39 one of the Tesla chargers with your current

00:04:42 like F-150 Lightning or Mustang Mach-E.

00:04:45 Basically in the next 12 to 18 months,

00:04:47 all these car companies are going to get this thing

00:04:49 sorted.

00:04:50 Will everybody be on one single charging standard?

00:04:53 I don't think so.

00:04:54 - But this is just VHS Betamax 2024 edition.

00:04:59 Like, you know, the one's winning out.

00:05:00 And again, the other thing, I think it's just,

00:05:03 it's such a big thing.

00:05:04 We should probably devote an entire show to it,

00:05:06 but the alliance of GM, Stellantis,

00:05:11 and remember Stellantis is much bigger than we think it is

00:05:13 in the US and they're doing, making a lot of moves.

00:05:16 You'll find out in a second, but with our CES guests,

00:05:19 Volvo, BMW, Hyundai, Kia, Genesis,

00:05:23 and I think there's one more,

00:05:23 but they're building a competitor to the Tesla charging,

00:05:27 a supercharger network.

00:05:28 And Tesla right now, I think they're like 18,000.

00:05:32 - 22,000.

00:05:33 - Oh, they added a lot more.

00:05:35 22,000 chargers in North America.

00:05:38 Yeah, it's either 17, 18, or 22.

00:05:40 But this alliance is going to be like 30,000 chargers.

00:05:45 So once that happens,

00:05:46 and that should be happening real soon,

00:05:48 just game on for EVs.

00:05:50 - Right.

00:05:51 - Yeah, it's going to be a big change.

00:05:53 - Good question.

00:05:54 We want to encourage that.

00:05:55 So please ask us more questions.

00:05:58 You can shoot them to Johnny and I directly

00:06:00 in our Instagram.

00:06:01 So I'm @lowdown, L-O-H-D-O-W-N.

00:06:03 It's @johnnylierman.

00:06:05 - Yeah, just Google me.

00:06:06 - Or drop them in the YouTube comments.

00:06:08 So just send us a question and we'll pick it.

00:06:10 - We're too lazy to make an inevitable email.

00:06:12 So if you write to motortrend@motortrend.com,

00:06:14 we will get your question that way.

00:06:16 And if it's a good question,

00:06:17 maybe we'll send you something like this.

00:06:19 - Send you some swag, send you a t-shirt,

00:06:22 belt buckle, you name it.

00:06:23 Okay. - All right.

00:06:24 - So we covered it off.

00:06:25 We got a new sponsor.

00:06:26 We got some new segments.

00:06:28 We got some new audio visual stuff.

00:06:30 And we have some great conversations

00:06:32 about the future of mobility coming live and direct

00:06:36 from CES.

00:06:38 So Johnny and I went to CES,

00:06:40 which used to be called the Consumer Electronic Show,

00:06:43 but actually it apparently doesn't mean anything anymore.

00:06:45 But it's all about tech.

00:06:46 No, CES is not an acronym anymore.

00:06:47 - No, I know.

00:06:48 It's funny.

00:06:49 It doesn't mean anything anymore,

00:06:50 but it's the biggest trade show there is.

00:06:51 - Biggest trade show in tech in the world,

00:06:55 arguably in the world,

00:06:56 happened in Vegas just about a month ago.

00:06:59 So these are all pretty hot and fresh.

00:07:01 And we did something different here.

00:07:03 Everybody's super busy.

00:07:04 It was hard to get this caliber of expert,

00:07:08 executive PhDs, professors, and this and that.

00:07:10 - For our normal two hour, normal, geez.

00:07:13 For our normal two hour deep dive interview,

00:07:15 tell us everything about yourself.

00:07:17 So we did a series of like 30 to 45 minute interviews

00:07:20 and we're bundling them together.

00:07:21 Today you get two.

00:07:22 - Exactly.

00:07:23 - And we're starting off with-

00:07:24 - Fabrizio Martini.

00:07:28 - A person who's much smarter than me and Ed,

00:07:29 who, well, he was sending stuff to Venus

00:07:32 and now he has his own battery consultancy company.

00:07:36 - Yes, Fabrizio Martini of electric vehicles.

00:07:40 We'll be chatting with him.

00:07:42 And then we're gonna follow up with a guy,

00:07:44 Professor Shai Shalev Schwartz from Mobileye.

00:07:48 So we'll go first with Martini, always start with Martini.

00:07:53 And then you'll get a chance to hear from the-

00:07:57 - Professor Schwartz.

00:07:58 - Professor Schwartz of Mobileye.

00:08:00 All right, so welcome Fabrizio Martini.

00:08:03 Right, did I get it right?

00:08:04 - That's a good name.

00:08:04 - Fabrizio Martini sounds made up.

00:08:07 - It does.

00:08:08 - That's really good.

00:08:09 - You sound like a cartoon character's name

00:08:10 from a Pixar movie.

00:08:11 - Yeah, I was gonna say the bad guy in "Cars 4."

00:08:14 - It's easy to introduce myself

00:08:15 because people pick up my last name very easily.

00:08:17 - Mr. Martini.

00:08:18 - Mr. Martini.

00:08:19 - Like a drink.

00:08:20 - Do you like martinis?

00:08:21 - I do like martinis.

00:08:22 - Whoa, whoa, whoa.

00:08:24 Gin or vodka?

00:08:24 - I think the traditional one.

00:08:28 - It's gin.

00:08:29 - Gin.

00:08:30 - Yes, okay, great.

00:08:31 - You know, Ed, if people knew more about us,

00:08:33 you and I should never talk about martinis at all.

00:08:34 - Yes, we should never talk about martinis.

00:08:35 - So on that note-

00:08:36 - Don't drink and drive.

00:08:37 - Yeah, yeah.

00:08:38 Unless the car drives itself.

00:08:39 On that note, so-

00:08:41 - Electric vehicles.

00:08:42 - Yeah, tell us everything.

00:08:45 As far as I know, you guys are like,

00:08:47 if I'm gonna build a new battery,

00:08:51 I use your services to help me optimize my battery.

00:08:54 Is that-

00:08:56 - We can help with that.

00:08:57 We can help with the design of new battery packs,

00:08:59 new architectures.

00:09:01 We can help with the analytics.

00:09:03 We have predictive analytics tools,

00:09:05 and we have an advanced control strategy that we call EVI.

00:09:08 The project started from a NASA project.

00:09:10 I was a principal investigator working

00:09:12 for the exploration of Venus.

00:09:14 We are trying to explore the surface of Venus,

00:09:16 and they need an electric rover.

00:09:18 So they wanted to see which batteries

00:09:20 could actually be deployed on Venus,

00:09:22 and they wanted to be able to predict failure

00:09:24 up to one year in advance.

00:09:26 So any thermal runaways, any potential exposure issues,

00:09:29 with prediction.

00:09:30 - And Venus is tricky 'cause the surface temperature

00:09:32 is like 800 degrees Fahrenheit, right?

00:09:35 It's extremely hot.

00:09:36 - It goes very low and very high.

00:09:38 - It does go low, oh, okay.

00:09:38 - It does go low, yeah.

00:09:40 - Absolutely. - When it's not facing

00:09:41 the sun. - Okay.

00:09:42 - 'Cause Venus is second planet from the sun,

00:09:45 but it's the hottest planet. - Big, fat atmosphere.

00:09:48 - It's hotter than Mercury. - Mercury, yeah.

00:09:51 - Because Mercury has an atmosphere,

00:09:53 Venus basically does not.

00:09:55 - Correct, so it was very challenging,

00:09:57 but we were able to explore battery manufacture

00:10:00 all over the world, from North America, Europe,

00:10:03 China, South Korea, many other areas.

00:10:06 But what we have seen is that the software stuck.

00:10:09 What was really making a big difference from the hardware.

00:10:11 So whatever battery we were picking,

00:10:13 the software was making a big difference.

00:10:16 Was increasing energy density, that means more mileage.

00:10:19 It was increasing the power, that means more horsepower.

00:10:22 And also we were able to predict failure.

00:10:24 Not 12 months in advance, like the NASA folks wanted,

00:10:26 but up to three months in advance,

00:10:28 we can predict any failures of any chemistries.

00:10:30 - Okay, okay, wait, wait, wait.

00:10:31 - Let's start with-- - Hang on, let's start first.

00:10:32 - Oh, whoa. - This is crazy.

00:10:33 So you're the principal investigator on a NASA project

00:10:37 of exploring Venus in an electric rover.

00:10:40 - Correct.

00:10:41 - What was the name of the project?

00:10:42 - It's still ongoing, it's a Venus exploration project.

00:10:45 Although lately, Mars and lately the moon exploration,

00:10:50 we want to send the first woman to the moon by 2026,

00:10:54 so it's upcoming for sure.

00:10:55 So the Venus program has started to place

00:10:58 a bit less relevant, but it's still there.

00:11:01 There are still developers working on it.

00:11:03 Great relationship with NASA.

00:11:04 - Plus, I think Russia's going to Venus, right?

00:11:06 - Everybody's going everywhere in space.

00:11:08 Okay, so then, okay, so how long have you been at NASA?

00:11:11 - Two years. - Two years?

00:11:12 And what's your background?

00:11:15 Do you have PhD?

00:11:16 - I did a master and a study PhD at Northeastern University.

00:11:21 I'm originally from Italy, where I did my--

00:11:23 - No. - Yes, you cannot say it, right?

00:11:25 Little Italian accent.

00:11:27 So I'm originally from Italy.

00:11:28 I moved to United States about 15 years ago,

00:11:31 and I did a master and a study PhD

00:11:33 at Northeastern University in Boston.

00:11:35 - In what field?

00:11:36 - The bachelor was in vehicles, vehicles dynamics,

00:11:39 and then when I moved to US, it was 2008, 2009,

00:11:42 it was a big push in electrification

00:11:44 and batteries development.

00:11:45 The USA government invested 500 million plus--

00:11:49 - That was the end of the Obama administration.

00:11:51 - Exactly, you got it.

00:11:52 So I was lucky enough to start to work

00:11:55 for the Department of Energy first,

00:11:56 Department of Defense after,

00:11:58 and then NASA for the last two years.

00:12:00 And then when I decided to transfer the software stock

00:12:03 to a private business called Electra Vehicles,

00:12:06 and we started about six, seven years ago,

00:12:08 and now we are a team of 55 people.

00:12:10 We have great partners like BlackBerry, Stellantis,

00:12:12 and many others, and we're bringing innovation

00:12:16 for the design, analytics, and controls of battery packs.

00:12:20 - So, you know, there's a lot of, let's say,

00:12:24 mental barriers to Americans specifically

00:12:28 buying new electric vehicles, and I call it FUD,

00:12:32 which is fear, uncertainty, doubt, you know.

00:12:33 And one of them is like, "Fire!"

00:12:35 But like--

00:12:36 - Thermal runaway, thermal runaway.

00:12:37 - Thermal runaway, but I think going with that,

00:12:39 you know, is being able to predict failure.

00:12:40 Like, how do you do that?

00:12:44 What are the markers you look for?

00:12:46 - Machine learning and artificial intelligence.

00:12:49 Those are keywords that can help us with charge EPT, great,

00:12:53 but can also help us to predict failure.

00:12:55 We have demonstrated with the Californian companies

00:12:57 that we're able to predict five out of five failures

00:13:00 of a set of 55 vehicles.

00:13:05 We're able to predict up to three months in advance

00:13:08 of potential thermal runaways or so on,

00:13:10 thanks to machine learning.

00:13:11 - And so, in other words, there's a bunch of variables,

00:13:15 and the machine says, "When X, Y, and Z happen,

00:13:18 "three months from now, something bad goes with the battery."

00:13:21 And so you can predictively say, "Is that basically it?"

00:13:24 - Correct, so our algorithm, we call,

00:13:26 what the innovation is called the EVI.

00:13:29 It's a brain for batteries.

00:13:30 It's a very advanced software that leverage a lot of data

00:13:33 coming out from the battery, the vehicle,

00:13:35 the driver, and the environment.

00:13:37 It's a 360 degrees data inputs.

00:13:39 In our patents, we have over 130 inputs.

00:13:42 Of course, the main one are voltage, current,

00:13:44 and temperature, as everybody knows,

00:13:46 but there are way more related to the driver.

00:13:48 How do you drive your vehicle?

00:13:49 Are you aggressive driver or not, and so on?

00:13:52 Where do you drive the vehicle?

00:13:53 Is it in Boston, in Palo Alto, in Milan, Italy,

00:13:56 or in Seoul, Korea, and so on?

00:13:58 So it's, in other words, environmental factors

00:14:00 playing heat and humidity and whatever.

00:14:03 >> How you charge the vehicle.

00:14:05 Are you doing a lot of fast charge

00:14:06 that could damage the battery heavily?

00:14:08 Are you doing a lot of high acceleration

00:14:10 that could damage the battery?

00:14:11 The machine only capture all of these instances,

00:14:14 and then we leverage a digital twin

00:14:16 of the battery in the cloud.

00:14:17 We are partnered with AWS,

00:14:19 and we create a digital twin of the battery.

00:14:22 >> Oh, that's wild. >> It's an exact replica.

00:14:24 By the way, we do it Electra, we do it differently.

00:14:26 We do a physics-based model plus a data-driven model,

00:14:30 and we combine the two.

00:14:31 The industry a few years ago moved all to physics-based,

00:14:34 and then a couple of years ago moved all to data-driven.

00:14:37 The solution is both combined.

00:14:39 >> Here's what, this is great.

00:14:41 >> That's cool. >> This is wild.

00:14:41 >> I like the virtual battery in the cloud.

00:14:43 >> So you have--

00:14:44 >> Real quick, though, you can do

00:14:45 an infinite number of those.

00:14:46 Like for every vehicle with a battery,

00:14:49 you can have its virtual clone in the cloud

00:14:52 and be monitoring the battery.

00:14:53 >> And custom, you can have your own digital twin

00:14:56 of your own battery pack.

00:14:57 You can have your own head, and I can have my own as well.

00:15:00 So each individual, we can start

00:15:02 from the same battery pack, we can all buy

00:15:03 from the same vendor, same model,

00:15:05 but then the algorithm becomes unique.

00:15:07 So you, driver, have your own singular digital twin

00:15:11 of your batteries. >> Yeah.

00:15:12 >> Wow.

00:15:13 >> Johnny loves to rapidly discharge his batteries

00:15:15 all day long 'cause he's like a super lead foot,

00:15:17 and then I'm way more cautious.

00:15:18 Okay, wait, hold on, that's a lot.

00:15:20 I'm trying to keep, I'm barely holding on,

00:15:22 so I know our audience is probably like,

00:15:23 what are they talking about?

00:15:25 So you're talking, from a data-driven perspective,

00:15:29 you're talking about all the variables in your data set.

00:15:32 This is voltage, time, temperature, location,

00:15:35 like hundreds of potential variables, right?

00:15:38 Thousands?

00:15:40 >> Well, the data inputs are about 100, 130, and so on.

00:15:43 >> Okay.

00:15:44 >> And then the uniqueness are infinite

00:15:46 because you can have infinite current,

00:15:48 temperature, voltages, and so on.

00:15:48 >> Okay, and that's where machine learning comes in

00:15:51 because you need a brute force approach

00:15:53 from a machine that can basically see

00:15:55 all of these changes in this data set

00:15:58 and then make a prediction.

00:15:59 >> Absolutely.

00:16:00 >> All right, so that's accurate, that's the data portion.

00:16:02 What's the physics model?

00:16:04 >> Physics model is basically the idea of,

00:16:06 you take a battery, and then you do a full set

00:16:10 of experiments under a design of experiments,

00:16:13 so you fully characterize the battery

00:16:15 by discharge curves at different current,

00:16:17 at different temperature, so you have an exact replica

00:16:20 of the physical object in digital.

00:16:22 >> Right.

00:16:23 >> Oh, okay, so but an actual physical test

00:16:25 on a physical model requiring a lab and--

00:16:27 >> Correct.

00:16:28 >> We have a couple of partners.

00:16:29 >> They're like, Lucid has a battery research place

00:16:34 at their headquarters, and they use these,

00:16:36 they kind of look like ovens.

00:16:38 >> Sure.

00:16:38 >> And you put a battery cell in there,

00:16:40 run electricity through it, you put in a race

00:16:43 'cause they do Formula E, so here's what the throttle

00:16:47 will look like through the race,

00:16:49 hit the change the humidity, change the temperature,

00:16:52 change the salt, same kind of testing.

00:16:54 >> That's exactly it.

00:16:55 So you want to produce as many scenarios as possible,

00:16:59 but physics-based model cannot cover all the scenarios.

00:17:02 The machine learning, data-driven, can,

00:17:04 because the machine learning start to improve

00:17:06 while you take off, you go to work, you go to the gym,

00:17:09 you go to pick up your kids, start to learn more and more.

00:17:12 That's why at Electra, we have developed

00:17:14 digital twin of batteries of multiple chemistries,

00:17:17 whether it's NMC, LFP, LTO, lithium metal solid state,

00:17:22 digital twin of those batteries,

00:17:24 based on physics-based and machine learning.

00:17:25 And that's extremely powerful, because it can give

00:17:27 a big advantage to the driver and the user experience.

00:17:30 >> Let's, can we pause right there and talk about

00:17:33 the state, current state of automotive battery tech.

00:17:36 Like, you know, everybody, lithium ion,

00:17:38 but now everyone's moving to lithium ion phosphate.

00:17:42 >> Let's do that one second, 'cause I actually

00:17:43 have a question about what he was saying about,

00:17:44 so just so I understand it.

00:17:47 The physics-based model sounds, I mean, I get it,

00:17:51 and it's kind of, it's old school, 'cause it's,

00:17:53 you're actually using the experiment and,

00:17:56 but it sounds really expensive, right?

00:17:58 Is that, I mean, from a cost perspective, right?

00:18:00 If you have a good data-driven model,

00:18:03 machine learning model, you can do,

00:18:05 you can do infinite number of tests,

00:18:07 and without needing a lab, needing physical experiments.

00:18:12 So why, why then do you need the,

00:18:14 why is it better for you to combine the two?

00:18:16 >> You just got to make sure it lines up with reality.

00:18:17 >> It's a very good question.

00:18:18 So the industry actually, initially,

00:18:20 had focused only on physics-based model,

00:18:22 and then they realized what you just mentioned,

00:18:24 that was extremely expensive.

00:18:25 So then they moved to data-driven model,

00:18:28 and they say, okay, let's collect as much data as possible.

00:18:30 But what they were missing is that the physics-based

00:18:34 give you a full understanding of the basic principle

00:18:39 on how the chemistry works in different scenarios,

00:18:42 and you cannot really cover those with machine learning,

00:18:45 because maybe sometimes you don't experience

00:18:47 a certain temperature,

00:18:48 you don't experience a certain current,

00:18:50 you don't experience certain scenarios.

00:18:52 So it's like a map, and in the map,

00:18:54 you start to cover areas with the physics-based,

00:18:57 but then you start to see some spots,

00:18:59 and those spots are being covered by machine learning.

00:19:01 So the combo is really nice,

00:19:04 and when you apply machine learning,

00:19:05 you don't have to do thousands of thousands

00:19:07 of physical-based tests that are going to spend

00:19:10 millions of dollars on.

00:19:11 You can run the set of tests,

00:19:13 and then you can collect as a design of experiments

00:19:15 specific for each chemistry.

00:19:17 You run through those, and you have enough data

00:19:19 to have your basic physics-based model.

00:19:21 And then machine learning can be from day one,

00:19:23 when you buy the vehicle, to end of life.

00:19:25 >> Okay, so now, and before we get to this,

00:19:27 here's a good question on battery,

00:19:28 the state of battery tech, but now I get it.

00:19:31 So your shirt, we were talking about before we went on,

00:19:33 on this sleeve, you have your partners, right?

00:19:36 >> Partner partners, yes.

00:19:37 >> So you're a partner, so you've got AWS.

00:19:39 >> Thank you sponsors.

00:19:40 >> Is that Qualcomm IC?

00:19:42 >> Texas Instruments, NXP, and Stellantis.

00:19:44 So-- >> Blackberry.

00:19:45 >> Blackberry, yes.

00:19:46 Who's booth we're in, thank you Blackberry.

00:19:47 >> Yes.

00:19:48 >> I get why a company like Stellantis,

00:19:52 even though everyone's trying to go vertical,

00:19:55 they're not going to want to do what you're doing,

00:19:57 'cause this is highly specialized, right?

00:19:59 Or at some point, is it, and maybe I'm jumping to the end,

00:20:02 is like, are you waiting to be acquired,

00:20:04 as like a big company, to buy out what you do, or?

00:20:07 >> So the reason why it's extremely important

00:20:09 for OEMs to work company, technology provider,

00:20:12 like electric vehicles, is because we have been training

00:20:14 our algorithm and our nets for the last six, seven years.

00:20:18 With not only a specific brand and model,

00:20:21 but multiple brands.

00:20:22 We work with the two wheelers, three wheelers,

00:20:24 passenger cars, trucks, delivery trucks, and more,

00:20:28 and so on. >> Buses.

00:20:29 >> Drones, buses, and so on.

00:20:31 So the network effect is something that you only,

00:20:34 you cannot buy time with the network effect.

00:20:36 You need to run the algorithm over and over.

00:20:38 So when a company like Stellantis,

00:20:40 or other are interested in this,

00:20:42 they want to jumpstart the competition,

00:20:44 and instead of start building data set

00:20:46 and training algorithm, it will take two, three, five,

00:20:48 ten years. >> They partner with you

00:20:49 'cause you have all the data. >> They partner with us.

00:20:51 And they immediately have solution

00:20:52 after a couple of weeks of data inputs,

00:20:54 they have beautiful output, and we should talk about

00:20:56 which are the output, because maybe this is what

00:20:58 the people are interested in.

00:20:59 >> So then, so you just buy everybody's battery

00:21:02 and start building, testing?

00:21:05 >> Electra has characterized over 330 batteries

00:21:07 from over 80 manufacturers in the world.

00:21:10 So we have done, the first three, four years

00:21:13 of the company, we have characterized

00:21:15 most of the batteries around the world.

00:21:16 And of course, lately, we have been working

00:21:18 with the NREACH, NMC Chemistry, Lithium Metal,

00:21:21 some solid state, to continue to build up

00:21:23 that battery database.

00:21:24 But we have over 330 batteries fully modeled.

00:21:27 >> How do you-- >> I was just going to say,

00:21:28 it's fast 'cause when I was up looking at

00:21:30 Lucid's battery testing department,

00:21:32 yeah, I mean, they just literally had battery cells

00:21:34 from Sony, in fact, the Lucid Air, my understanding is,

00:21:38 that they were, I think it was Sony,

00:21:40 and they said, "We want to use this,"

00:21:42 and they said, "No, we didn't design that for automotive."

00:21:45 And Lucid said, "This is the perfect battery for us."

00:21:48 And they said, "All right, but you got to take

00:21:50 "the word Sony off of it.

00:21:51 "We'll sell it to you wholesale, undercover."

00:21:53 >> I'm sure they're really happy

00:21:54 that you're telling everybody.

00:21:56 >> Whatever, and then Sony was like,

00:21:57 "Could you put the name back on?"

00:21:59 And they're like, "No."

00:22:00 >> So then do you-- >> But it's fascinating,

00:22:02 'cause it's just batteries.

00:22:03 >> It's a dumb question.

00:22:05 Do you buy these batteries, just go out to the manufacturer,

00:22:08 or do they want to give them to you?

00:22:10 I mean, you're creating a deep database

00:22:15 that you might eventually sell back to them

00:22:17 using their technology, right?

00:22:19 >> Right, exactly, you're right.

00:22:20 So yes, we do buy the batteries.

00:22:23 We receive the batteries from our partners.

00:22:26 We receive batteries from OEMs, and a mix of those,

00:22:29 and we have two laboratories, two partners,

00:22:31 one in Italy, one in North America,

00:22:32 to run all the tests for us,

00:22:33 because we are a software company.

00:22:35 We are a pure software, so we license our--

00:22:37 >> So in other words, what you're saying,

00:22:38 let's say, just pick on Sony again,

00:22:40 brand X makes a car with a bunch of Sony batteries in a pack,

00:22:46 and so does brand Y, but it's like,

00:22:48 different configuration, so in other words,

00:22:52 both might be the same brand of battery,

00:22:54 but because of the way they're put together,

00:22:57 different voltages, different results.

00:23:01 >> It changes performance, for sure.

00:23:02 The architecture changes performance.

00:23:05 Electra has developed a solution called N-Power

00:23:07 that helps also to design the next generation batteries.

00:23:10 So we work with a couple of battery manufacturers,

00:23:12 saying exactly this, we love these kind of specs,

00:23:15 help us to build that, and what we do,

00:23:17 we leverage machine learning and AI,

00:23:19 we set off running thousands of tests,

00:23:23 we actually run a few tens of tests,

00:23:25 and simulate most of the others.

00:23:26 So ML can also help us to cut down 98% of R&D time,

00:23:31 so instead of 10 years of development,

00:23:34 we do it in six months, one year,

00:23:36 one and a half year maximum, for developing new chemistries.

00:23:39 We must leverage machine learning and AI

00:23:41 as much as possible.

00:23:42 Great for charge EPT,

00:23:43 but can be deployed for electric vehicles.

00:23:46 >> So then, you must have pretty good competitive,

00:23:51 do partners reach out for competitive analysis?

00:23:54 Like, yes, they go, okay.

00:23:58 >> But with Electra started very early.

00:24:00 If you look at our competitors,

00:24:01 Electra was one of the first ones to start

00:24:03 six, seven years ago.

00:24:04 That was also supported by USA government,

00:24:07 that invest a lot of money back in 2010,

00:24:10 and also in 2017, 18, into electrification

00:24:12 and battery development.

00:24:14 And so we have a competitive advantage

00:24:17 of that time to market.

00:24:19 >> Your first packs must have been Tesla,

00:24:21 and the Panasonic packs, and Nissan Leafs,

00:24:24 and that and so.

00:24:24 >> Initially, we actually had the Tesla Model S,

00:24:27 in 2017, 18, we ran a lot of tests in Arizona,

00:24:32 many areas to collect data.

00:24:33 But then we started to add the Stellantis Fiat 500,

00:24:36 the Hyundai Kona, Hyundai IONIQ 5,

00:24:40 and many other vehicles to collect data.

00:24:42 We actually have drivers right now,

00:24:43 that while we have in this chart,

00:24:45 they are driving and collecting data for us.

00:24:47 So we have a network of battery testing facilities,

00:24:51 and drivers that collect data for us,

00:24:53 and feed all in this case, AWS Cloud,

00:24:55 to feed the data lake.

00:24:57 >> Interesting.

00:24:58 >> Data lake.

00:24:58 >> So, but, I do want to talk about the state of,

00:25:01 yeah, the state of automotive battery tech.

00:25:04 Like, where are we at?

00:25:05 >> We're in Vegas, okay?

00:25:06 If you're going to put a bet on battery chemistry,

00:25:10 and we want it to hit, what's the one?

00:25:14 What do you believe in?

00:25:15 >> There is not going to be one.

00:25:16 There are going to be several.

00:25:17 >> Give me your top three.

00:25:18 The trifecta.

00:25:19 >> Well, they continue to develop.

00:25:22 We are just at the beginning of battery development.

00:25:24 But definitely, NMC, LFP,

00:25:28 and the next one would be NITU metal.

00:25:31 So I think these three.

00:25:32 There's also sodium ion, quite picking up,

00:25:34 due to cost and safety.

00:25:36 But NMC and LFP would be the biggest one,

00:25:39 and then the other will start to come up more and more.

00:25:41 >> What about, we had a guy on the podcast,

00:25:44 from RNext Energy, they are doing the dual chemistry battery

00:25:49 they call it, I think, Gemini?

00:25:51 >> Yeah, one for acceleration, one for steady state.

00:25:53 >> Good idea, bad idea?

00:25:55 >> So in the NASA project,

00:25:56 the design was actually three batteries working together,

00:25:59 because of the temperature range

00:26:00 that you were mentioning, Johnny.

00:26:02 Is that, what we have developed there,

00:26:05 Electra has developed an algorithm that can power split.

00:26:08 So leverage two or three chemistries,

00:26:10 and what we have showcased is that

00:26:11 combining two chemistries with an ML algorithm,

00:26:14 so instead of having a switch on/off,

00:26:16 so one battery or the other,

00:26:18 we have developed an algorithm that utilize both batteries

00:26:22 as needed, and mix and match the two.

00:26:24 We actually have demonstrated on a small battery pack,

00:26:26 we didn't do a large vehicle demo,

00:26:29 but we have algorithm that we license to OEMs

00:26:32 to run two batteries.

00:26:33 So that's another aspect, because I mentioned to you,

00:26:35 design, analytics, and controls,

00:26:37 under the controls umbrella,

00:26:39 there is smart charging we can talk about,

00:26:41 but there is also power split of two chemistries.

00:26:43 So we can, we could collaborate with them.

00:26:46 >> So you mentioned briefly solid state batteries,

00:26:49 and Toyota's made a big promise that they're coming in 2025.

00:26:53 Where's this technology at, and for people listening,

00:26:56 like what's the real advantage of a solid state battery?

00:26:59 >> Well do you believe that we're going to see solid state

00:27:02 in that time frame, or is it?

00:27:04 >> No, it's going to be very challenging.

00:27:06 >> Why, why?

00:27:07 >> So one of our, the manufacturing issues,

00:27:10 but one of our technical advisor,

00:27:12 Claudio Capilla, Dr. Claudio Capilla,

00:27:14 is one of the inventor of the solid state technology

00:27:18 at Toyota, 20 years ago, 25 years ago.

00:27:21 And there are a lot of limitation

00:27:23 in solid, full solid state batteries,

00:27:25 in particular in the manufacturing aspect.

00:27:27 The ceramic layer that you need to put

00:27:29 between anode and cathode,

00:27:31 it's very easy to build at a laboratory scale,

00:27:34 or a pilot scale, but when you start to do mass production,

00:27:37 it's very hard for at the edges,

00:27:39 they can break, they can have edges problem and so on.

00:27:43 This problem has not been solved yet

00:27:46 on the science aspect, so it's not

00:27:48 an engineering optimization, it's science,

00:27:50 and it's not been resolved yet.

00:27:53 It will take time, but there are other chemistry

00:27:57 like enriched NMC or lithium metal

00:28:01 that are picking up way faster,

00:28:03 but they need a software layer, like the one we provide,

00:28:06 in order to make sure that to predict failure,

00:28:10 because those batteries can fail rapidly.

00:28:13 >> More rapidly than lithium ion?

00:28:15 >> Yeah, because they're pushed to the limit.

00:28:17 >> And you said lithium metal was one?

00:28:20 >> Yeah, lithium metal.

00:28:20 >> So describe that, what is a lithium metal battery?

00:28:23 >> It's one of the two electrode,

00:28:25 is a pure metal foil.

00:28:28 >> Made of?

00:28:30 >> Lithium.

00:28:32 >> Oh, so solid lithium.

00:28:34 Okay.

00:28:35 >> And that's--

00:28:36 >> Increase energy density and mileage.

00:28:38 >> And advantageous for us in the US,

00:28:41 as we are rapidly behind China,

00:28:44 and China has locked down all of the resources

00:28:46 for battery production, but we've discovered--

00:28:49 >> Yeah, America's nothing but lithium.

00:28:51 >> Yeah, apparently--

00:28:52 >> The whole country's made of lithium.

00:28:53 >> Apparently the whole country's made of lithium.

00:28:54 So maybe that provides a little bit of a strategic advantage

00:28:59 for us as we're building up our battery.

00:29:02 >> But I'm still convinced, definitely the material

00:29:06 challenge will be definitely a global thing

00:29:09 to discuss and so on.

00:29:10 But I still believe that the software stack,

00:29:12 MLAI stack is what make it a big difference.

00:29:15 That's why I put all my saving, all my life,

00:29:18 everything and the team as well into Electra.

00:29:20 >> Let's talk about that.

00:29:21 So I still have no idea, so your degree is,

00:29:25 are you a computer scientist?

00:29:26 Are you a coder?

00:29:27 What are your degrees in?

00:29:29 You like math?

00:29:30 >> My degree is in mechanical engineering.

00:29:32 >> What?

00:29:33 >> But I have a great team of PhDs, over 37 PhDs,

00:29:37 helping me, they're great minds.

00:29:38 The team is what really make a difference.

00:29:40 And we are a full remote company.

00:29:42 So we have people from all over North America

00:29:45 and all over Europe and those people are very talented.

00:29:48 They can go to work for Meta, they can go to work

00:29:51 for Tesla, AWS and the reason why they work for Electra,

00:29:54 there are two.

00:29:54 One is because they believe in the mission and vision

00:29:57 and values we have.

00:29:59 We have a very long term vision on how software defined

00:30:02 vehicles can really make a big paradigm shift

00:30:05 to the inevitable EV race.

00:30:08 >> Look at this guy, look at this guy,

00:30:09 he's tired of this kid.

00:30:10 >> He's good.

00:30:11 >> By great name by the way, he will be inevitable.

00:30:15 But also the reason that--

00:30:17 >> That needs constant validation.

00:30:19 >> Thank you, thank you.

00:30:19 >> Absolutely, you're sad.

00:30:21 But the fact that they can work remotely,

00:30:24 they love that because we have people from Atlanta,

00:30:26 Chicago, Spain, Portugal, Italy, anywhere in the world

00:30:31 and they love that flexibility.

00:30:32 >> I'm still blown away because this is so funny.

00:30:35 Right now in this shift from electromechanical vehicles

00:30:38 to software defined vehicles, it seems like everybody,

00:30:41 they're all mechanical engineers that are finding

00:30:43 these companies, yeah, well when I went to school,

00:30:45 I didn't really want to study CS or electrical engineering.

00:30:48 But how much of your day is immersed in code,

00:30:53 in that side of it?

00:30:56 You had to learn all this stuff?

00:30:57 I mean at NASA, you went to NASA as a mechanical engineer

00:31:01 specializing in batteries?

00:31:04 >> Maybe.

00:31:05 >> Hypercapacitors is what I heard.

00:31:06 Ultracapacitors.

00:31:07 >> Good job, yeah, supercapacitor was definitely

00:31:09 one of the chemistries where we're deploying

00:31:11 for the high power stock.

00:31:12 >> Also not, yes, but also not mechanical engineering.

00:31:14 Like where did you pick up all this physics?

00:31:16 >> No, the master and PhD was on basically

00:31:19 material science and chemistry.

00:31:20 >> Got it, okay.

00:31:21 >> But then we are seeing that every time we add

00:31:23 in a software stock, we are improving the performance.

00:31:26 So you add like, you know, an energy density

00:31:31 and then with software you're able to increase

00:31:32 that 10, 15, 20% extra energy.

00:31:35 So you're able to squeeze more energy out

00:31:37 fast charge a battery thanks to software stock

00:31:41 can also happen.

00:31:42 So the software makes a big, big difference.

00:31:44 And then I started to hire great minds

00:31:46 that are in the team that are all like

00:31:48 ML/AI expert, data scientist, and so on.

00:31:51 We are mainly driven by that.

00:31:53 >> Okay, this is fascinating 'cause again,

00:31:56 he said it not twice, let me break it down for you

00:31:59 'cause it's something I'm barely holding onto.

00:32:01 Through all of your testing, virtual machine learning

00:32:06 in the cloud, through physical testing,

00:32:09 you're able to analyze all these batteries

00:32:11 and basically say that if you have a battery

00:32:14 of a certain potential, the software that manages it

00:32:19 on top of it can provide a measurable,

00:32:24 a substantial advantage if it's well written,

00:32:27 if it's well executed.

00:32:29 And I think this is the challenge of the industry

00:32:33 where there's not a lot of people who understand software

00:32:37 at that level where they can execute

00:32:39 and take advantage of this potential.

00:32:42 >> Absolutely, before you mention about

00:32:44 what are your plan that you consider to be acquired,

00:32:46 Electra has received six offer for acquisitions

00:32:49 because for a couple of reasons.

00:32:51 One, we have a great team of talent.

00:32:52 The other one is we have over eight patents

00:32:55 around what I'm sharing and claiming here.

00:32:59 And then we have a trained machine learning model

00:33:02 for the last six, seven years and so on.

00:33:04 And the number of data points started to grow more and more.

00:33:07 We reached 1.2 billion data points in our algorithm

00:33:10 in the data lakes that we are talking about.

00:33:13 So the power of this machine is becoming extremely strong

00:33:18 and we are looking for partners like Stellantis

00:33:20 and others, OEMs, to deploy this to the mass market.

00:33:23 Because think about billions of electric vehicles

00:33:27 in a couple of years, but millions of electric vehicles

00:33:29 all connected to the same couple of batteries

00:33:32 between what could happen.

00:33:34 Your battery pack in your vehicle will improve performance.

00:33:38 Your vehicle can learn if it's going to catch fire or not

00:33:42 in a couple of weeks or a month or so.

00:33:45 This is extremely important.

00:33:46 So we want to deploy this to market.

00:33:48 >> You could do any, you could do, how do I put this?

00:33:51 You could do anything with, this is battery control software

00:33:56 so I was talking, I can't remember the company's name.

00:33:59 It's like something humanity, but basically,

00:34:02 there's a problem where, ah, windmills are great.

00:34:04 They create terawatts of electricity

00:34:06 and if there's nowhere for it to go, it's gone.

00:34:09 They generate it and it just goes away.

00:34:11 So Germany, they're starting to turn it into hydrogen,

00:34:13 but this other company was saying, hey,

00:34:16 if everybody plugs their electric cars in at night,

00:34:18 put the energy there, we'll pay you a little bit of money

00:34:24 to store our energy temporarily

00:34:26 and then in the middle of the night it'll go out

00:34:27 and then you'll draw in more energy.

00:34:30 But you can control for that kind of thing, right?

00:34:32 >> Absolutely, so we are exploring the V2X integration,

00:34:36 so when you connect your vehicle to your house, for example,

00:34:41 and you start to exchange energy.

00:34:42 The problem is that by doing so,

00:34:46 you're going to wear the battery,

00:34:47 but the question is by how much.

00:34:49 And what we do as a company,

00:34:51 we provide a very accurate state of health and degradation.

00:34:54 So we basically tell you in every instance,

00:34:57 how much value is left in your battery.

00:34:59 And that's extremely important.

00:35:00 We create a certificate of health

00:35:02 that you can click at any time and say,

00:35:04 oh, my battery is at 83% and this is the value, $26,000.

00:35:09 >> And this is going to also become more important,

00:35:10 not only for electric car adoption,

00:35:12 but people are going to get batteries for their homes

00:35:14 and you can analyze the battery in your house

00:35:17 and you'll know, go to sell a house,

00:35:20 this is what the battery's worth,

00:35:21 this is how much life is left in it.

00:35:23 Also--

00:35:24 >> To resell the batteries, or sell the vehicles.

00:35:26 Right now, people don't know exact value

00:35:28 because it's not related to mileage anymore.

00:35:31 Mileage doesn't matter anymore with electric vehicles.

00:35:33 It matters how much energy you have left

00:35:36 in your battery pack.

00:35:37 How much your battery pack was worth over time.

00:35:40 >> Could you talk about this?

00:35:42 What does average battery wear look like?

00:35:46 In other words, we know with cars,

00:35:49 they get to about 200,000 miles, internal combustion,

00:35:52 you're kind of done, you might get lucky

00:35:54 and they go a little bit further.

00:35:55 But what's a battery, after 100,000 miles,

00:35:59 what's a battery look like?

00:36:01 >> The issue with batteries, it extremely depends

00:36:04 by how do you drive the electric vehicles,

00:36:06 how do you charge electric vehicles,

00:36:08 and where do you store the electric vehicles.

00:36:10 So the answer is not really unique.

00:36:12 It depends by, are you driving in Canada,

00:36:14 or you're driving in South America or Miami,

00:36:18 are you extremely aggressive that every time

00:36:19 there is a green light, you push because you want

00:36:21 to enjoy the acceleration?

00:36:24 Of course, yeah, Ludicrous mode and so on.

00:36:27 So those are things that matters a lot on the batteries.

00:36:31 And then there are also differences,

00:36:33 whether it's an engine, the engine right now

00:36:35 is the ICE engine I'm referring to,

00:36:37 they are more or less the same, identical,

00:36:39 almost identical with each other.

00:36:41 Battery packs are not.

00:36:42 So even when they come out from the factory,

00:36:44 they're not identical, they start from

00:36:46 a slightly different health and so on.

00:36:50 >> And why is that?

00:36:52 >> Because batteries are a live organism.

00:36:54 It's very, it's not a simple mechanical aspect.

00:36:59 There is electrochemistry behind.

00:37:03 So we need to start to view batteries

00:37:05 as a very complex things, and of course,

00:37:09 the software can simplify and make sure

00:37:11 that what the software does, make sure that

00:37:13 all the batteries are similar to each other

00:37:16 and more or less on par.

00:37:17 >> To guarantee that this vehicle will have

00:37:20 the kind of performance that its twin does.

00:37:22 >> Correct.

00:37:23 So we do a step forward and we say,

00:37:26 even if they're slightly different,

00:37:29 the software layer will put them on a flat surface,

00:37:33 similar to each other, and then the way you drive it,

00:37:35 the way you charge, where do you drive,

00:37:38 we can still have an optimized control strategy

00:37:41 for your own battery pack.

00:37:42 And that's the paradigm they want to bring to market.

00:37:44 We call it EVI.

00:37:46 >> Okay, EVI.

00:37:47 So you heard it here first, a battery is a living organism.

00:37:50 So let's just give something for our audience,

00:37:52 something practical.

00:37:53 I think we know this, let me just gut check with you.

00:37:58 The proper care and feeding of your battery.

00:38:01 You mentioned how you drive, where you drive.

00:38:04 I assume if you live in a more temperate climate

00:38:09 that doesn't have a huge temperature variation,

00:38:12 your battery will likely last longer.

00:38:15 Charging it, we've always heard,

00:38:17 don't charge it all the way to 100%,

00:38:19 try to go, don't go to zero,

00:38:21 like 5, 10% to 80% on a regular basis.

00:38:24 >> Is that true-ish?

00:38:26 >> So it depends by the chemistry.

00:38:28 LFP can be 10 to 90, NMC can be 20 to 80%,

00:38:32 but it doesn't make sense that the user

00:38:35 optimize the charging, optimize the temperature,

00:38:38 optimize how much they press.

00:38:40 We need to have software layer that takes care of that.

00:38:43 So the user just plug it in, don't think about it,

00:38:46 and the software says, let me recharge it up to 76%,

00:38:50 because I know that the day after

00:38:51 you only need 20% charge.

00:38:54 So the software can be unlimited,

00:38:56 and that's why we are getting a lot of traction

00:38:59 with different OEMs, tier one suppliers,

00:39:01 and we are showcasing this solution,

00:39:05 and actually I can give you a preview,

00:39:07 since you asked about previews.

00:39:09 We're going to have an in-vehicle demonstration,

00:39:12 and we're going to show up at the MOVE,

00:39:14 follow-up conference, we are here at CES,

00:39:17 but the next one, MOVE, we're going to do

00:39:19 a very big in-vehicle demo, and driving from

00:39:21 Turin to London with our couple of solutions installed,

00:39:25 and showcase great technologies.

00:39:27 >> In what cars?

00:39:28 >> Yeah, so what vehicles?

00:39:29 >> It's early, it's early.

00:39:31 >> Stellantis?

00:39:32 >> It's one of the-- >> What's the vehicles?

00:39:34 >> It's one of the most sold vehicles in the world.

00:39:36 Let's say this.

00:39:37 >> It's a Corolla.

00:39:38 (laughing)

00:39:39 >> No, it's electric.

00:39:40 >> Could be a Tesla, could be a BYD.

00:39:43 >> Yeah, all right.

00:39:45 >> We'll circle back.

00:39:46 >> For sure.

00:39:47 >> Not a Maserati.

00:39:47 >> Who is your-- >> It's a great.

00:39:50 >> Actually, related, are you in a vehicle now?

00:39:55 Is your technology in a vehicle that Johnny and I

00:39:57 could go and drive today?

00:39:58 >> Not in production, but we have some pre-production

00:40:01 vehicles that are testing our technologies,

00:40:04 and are getting great results.

00:40:06 >> And this is with Stellantis only, or with other partners?

00:40:08 >> No, also other-- >> Others as well.

00:40:10 >> Yeah, absolutely.

00:40:12 We're not, I mean, definitely Stellantis

00:40:13 is a very strong partner, and we are working with them,

00:40:16 a great team, great innovators, for sure.

00:40:19 We're also working with others.

00:40:21 And the goal, the mission of the company

00:40:24 is deploy this tech to as many millions of people

00:40:26 as possible, because it's working extremely well.

00:40:29 It give you an extremely accurate remaining mileage

00:40:33 at every instance, so you know if you're going to reach

00:40:35 the grocery stop, or if you're going to grocery shop,

00:40:37 or if you're going to be able to pick up your kid

00:40:40 at school and so on, and it doesn't have,

00:40:43 we have 1% error of the state of charge

00:40:45 on the mileage remaining, extremely accurate.

00:40:48 We have an extremely accuracy of the remaining useful life.

00:40:51 How long my battery will last?

00:40:53 Two years, three years, eight years, 10 years.

00:40:55 Machine learning can actually predict that,

00:40:57 and we have an extremely accurate

00:40:58 remaining useful life algorithm.

00:41:00 And also we can predict any potential failure,

00:41:03 so the families are safer, and you know.

00:41:05 >> I'm predicting that you are going to be

00:41:08 in the Ram electric truck.

00:41:11 >> Okay, okay, that's a good guess.

00:41:13 Maybe we'll meet up later on in six months.

00:41:15 >> Okay.

00:41:16 >> I was going to say, so but this all sounds like,

00:41:19 you know, as someone who owns an electric vehicle,

00:41:22 this sounds like stuff I'd like to know now.

00:41:24 Is there any way, like I'm not going to buy

00:41:26 a new Stellantis tomorrow, but is there any way

00:41:29 I could like, I don't know, get an app on my phone

00:41:31 that my phone can talk to my battery?

00:41:35 Is that possible at all?

00:41:36 Or you're not in deep enough?

00:41:38 >> We don't want to, Electra is trying to go to production

00:41:40 with SOP, start of production 2025.

00:41:43 So we are talking with a couple of OEMs

00:41:45 to get to production, of course.

00:41:48 >> But it has to be in the vehicle, in other words.

00:41:50 >> It needs to be--

00:41:51 >> You can't retrofit it.

00:41:52 >> It can also be retrofitted on the analytics aspect.

00:41:55 I mentioned we do help with the design of the battery packs,

00:41:58 analytics, and controls.

00:42:00 The design and the controls is more pre-production.

00:42:04 The analytics can be even afterwards.

00:42:06 If you have the ability to connect your car,

00:42:10 your vehicle to our analytics platform.

00:42:12 >> Okay, so you technically could plug an existing EV

00:42:16 into something and the Electra software would say,

00:42:20 "Here's this, here's that."

00:42:21 >> About the app, yes, we could like potentially launch an app

00:42:25 in partnership with an OEM to show some of the highlights.

00:42:29 Your value of your battery, if it's going to value

00:42:32 10,000, 15,000, 20,000 dollars.

00:42:34 >> Well, the predictive failure, I think, is big, huge.

00:42:37 >> Well, this is, I was going to say--

00:42:39 >> That's got to be the killer app.

00:42:41 >> What's it called, Recurrent?

00:42:43 There's a process site, something out there

00:42:47 that can tell you roughly the battery health.

00:42:49 Do you buy into that?

00:42:52 >> That's an approach that is without machine learning AI.

00:42:55 Basically what you do, you hook up the battery

00:42:57 and you discharge a couple of times

00:42:59 and you see more or less where the battery's at.

00:43:02 The machine learning doesn't need that

00:43:05 because it's going to leverage the physics-based model

00:43:07 and it's going to start leveraging the machine learning

00:43:09 while you drive, so the accuracy is actually improving

00:43:12 over time and not decreasing over time.

00:43:14 So that's true, we can put and showcase a risk indicator

00:43:17 and say your risk is going up and up and up.

00:43:20 Maybe go to the mechanics or do something

00:43:24 in your battery pack, so we can give these kind of alerts.

00:43:26 >> I'm going to suggest a line of business,

00:43:28 which is if you can have a consumer-facing

00:43:33 or even a B2B model that will tell used car,

00:43:37 for used car sales of EVs, to tell the buyer

00:43:41 what the potential health of the battery is

00:43:43 and how much value there is left in it,

00:43:45 that's extremely valuable at this point.

00:43:47 Everyone's like, the biggest knock of buying used EVs

00:43:50 is like, oh man, when that battery fails,

00:43:52 it's going to cost you $25,000 to replace.

00:43:54 >> Well, and then also--

00:43:55 >> You have no idea what you're getting into.

00:43:56 >> Nowadays, a good year in the US,

00:43:59 15 million new cars are sold,

00:44:01 but 45 million used cars are sold every year.

00:44:03 >> And there's the Biden plan that is supporting

00:44:06 the used car sales, used EV sales as well.

00:44:09 >> But even figuring out the Biden plan,

00:44:11 like just you look at EV adoption,

00:44:12 it's just a line going up.

00:44:15 And so in like 10 years, a lot of the used cars

00:44:17 are going to be electric.

00:44:19 >> Let me know if you do that business,

00:44:20 I'd be happy to help you with that.

00:44:21 >> Yeah, yeah, yeah.

00:44:21 >> We can do a revenue split.

00:44:22 You got the idea.

00:44:23 >> Yes, yes, and I'd just like some cash

00:44:25 because I was sitting here.

00:44:26 >> What, who is your competition?

00:44:29 Is anybody out there that--

00:44:30 >> So we do have a couple of companies.

00:44:32 We are a scale-up company.

00:44:34 We are 55 people, we are growing.

00:44:36 We raised $26 million in three rounds,

00:44:40 and we are continuing to grow with our partners.

00:44:41 We already generate revenue on the double digit and growing.

00:44:46 We do have a couple of competitors in the US,

00:44:49 in Europe, and in Asia, and so on.

00:44:53 We have several competitive advantages,

00:44:55 like this 360 degrees approach.

00:44:58 It's very unique to us.

00:44:59 We offer a full spread of solution,

00:45:04 like design, analytics, and controls.

00:45:05 Other are only focusing on analytics or controls.

00:45:08 And also we are agnostic to hardware,

00:45:10 and agnostic to chemistries.

00:45:12 And that's something that was very hard

00:45:13 to build in the algorithm, but now that we have it,

00:45:16 every company is benefiting because basically

00:45:19 they don't need one year, two years of having us

00:45:21 adjusting the solution to a specific chemistry.

00:45:24 They can come to us and in two, four weeks

00:45:26 have seen already the benefits of that.

00:45:29 So it's massive.

00:45:29 And it's not only automotive.

00:45:31 So Electra is now focusing on automotive,

00:45:33 but think about we can apply this technology to robots.

00:45:37 We can apply this technology to warehouse robots,

00:45:39 drones, flying pods.

00:45:42 >> Yeah, anything with a battery.

00:45:43 >> Correct, anything with a battery.

00:45:45 >> And then, again, just because I love your shirt.

00:45:50 So AWS is your partnership in the cloud.

00:45:54 Blackberry, you work with Ivy, the product?

00:45:57 >> Correct.

00:45:58 >> Which is their cloud software system too, right?

00:46:00 Helps chat with AWS.

00:46:03 >> Correct, correct.

00:46:04 So we actually heard John Wall a couple of years ago

00:46:07 speaking about this idea of on top of the operating system,

00:46:10 creating a middleware.

00:46:12 That was basically the idea of to collect the sensor data

00:46:15 and create additional data sets, synthetic sensors.

00:46:19 And we love the idea because what we do at Electra,

00:46:21 we take inputs and we talk about the 100, 130 inputs,

00:46:26 and we come up with insights.

00:46:27 And as soon as we have seen the idea of Ivy,

00:46:30 even before calling it Ivy, we say, okay, wow,

00:46:32 that's exactly what we need.

00:46:34 And we started to work with the Blackberry team.

00:46:36 We actually were the first investment

00:46:38 of Blackberry Ventures.

00:46:39 And we started to build a partnership with them,

00:46:44 very strong partnership.

00:46:45 We have demonstrated several of our technologies

00:46:47 working with Ivy.

00:46:48 And now we are ready to production.

00:46:50 >> And for those going back,

00:46:51 John Wall is the head of Blackberry QNX,

00:46:54 which is basically the automotive platform,

00:46:57 software platform, real-time operating system

00:47:00 that Blackberry has created.

00:47:01 It's in a whole bunch of vehicles.

00:47:03 And then Qualcomm also is a partner.

00:47:06 >> Well, we partner with the chipset makers

00:47:09 and we have demonstrated our technology

00:47:11 to be installed into NXP, Texas System, and Qualcomm,

00:47:14 but it can be transferred to any chipset

00:47:17 that a company like Stellantis, Volkswagen, Tesla,

00:47:20 or even many others can utilize.

00:47:22 So it's hardware agnostic.

00:47:23 >> Very smart.

00:47:24 That's what it means, a hardware agnostic.

00:47:25 Make as many friends as possible.

00:47:26 >> That's right.

00:47:27 >> It takes time.

00:47:28 It takes more time.

00:47:29 You need to write a code in a certain way.

00:47:30 It takes longer time, maybe take one or two years more.

00:47:33 But once you have it, it's transferable.

00:47:35 It's beautiful.

00:47:36 >> Are you, do you have a position on,

00:47:39 you said writing code takes so much time.

00:47:42 Wouldn't it be so much easier if everybody

00:47:43 was using the same code and it was all open source?

00:47:46 Or no?

00:47:48 Do you have any thoughts in that direction?

00:47:50 There's a Red Hat, Linux, and GM,

00:47:53 they're trying to go more open source.

00:47:56 Is that--

00:47:57 >> The issue with open source is that you need,

00:48:00 it's great for collaboration

00:48:01 and it helps innovation for sure,

00:48:03 but the level of safety started to decrease

00:48:07 because you basically need automotive certifications

00:48:11 that are very time consuming, very complicated,

00:48:13 and we want to make sure that your software

00:48:15 generated by a third party that you don't know

00:48:17 where it is and so on, is going to be certified.

00:48:20 >> ASLD, is that the one, right?

00:48:22 >> ASLD is definitely a big one to do,

00:48:24 but if it's an open source, who is going to pay for that?

00:48:28 Who is going to cover that certification?

00:48:29 That's why other operating systems are still not there yet.

00:48:34 Maybe in the future, but it will take some time.

00:48:38 >> We're going to start wrapping up a little bit here.

00:48:41 >> Yeah, I was going to say, I'm just,

00:48:42 and again, I'm just always curious,

00:48:44 and we're here looking at future tech

00:48:46 and what the future's going to look like.

00:48:48 Right now, a good electric vehicle,

00:48:52 or 300 miles of range and the battery weighs

00:48:56 1,500 pounds or something,

00:48:58 what's the next generation of batteries?

00:49:01 What are we looking at?

00:49:03 What do you see, trend-wise?

00:49:04 >> I think we're just at the beginning.

00:49:05 >> Well, I totally agree.

00:49:07 >> If you look at the battery,

00:49:08 every year we cut down 5, 10% of weight

00:49:11 by optimizing the chemistry,

00:49:12 optimizing the architecture and so on.

00:49:14 So if you project this to 50 years, 100 years ahead,

00:49:17 the battery pack will continue to get smaller and smaller

00:49:19 and maybe becomes as big as, you know, at this price.

00:49:22 >> There's got to be some sort of Moore's Law, right?

00:49:25 >> For sure, yes, absolutely.

00:49:27 But what I'm saying is that we're going to see

00:49:30 continuous increasing energy density,

00:49:32 decreasing weight, and better performance.

00:49:36 And of course, the software stack

00:49:37 will make a big difference there.

00:49:40 And we are going to also start seeing firmer updates,

00:49:45 similar to your iPhone or Samsung,

00:49:49 that they receive an iOS update or Android update.

00:49:53 We can do the same with your battery.

00:49:54 So we're going to start to see firmer updates

00:49:56 and that will improve over time.

00:49:57 >> This is already happening.

00:49:58 I mean, you know, with Rivian,

00:50:01 they did a vehicle update,

00:50:04 but you went from having 314 miles of range

00:50:07 to 328 via software update.

00:50:10 Hardware stayed the same.

00:50:12 So it's already starting to happen.

00:50:13 >> It's happening.

00:50:14 >> So let me ask you, you said optimize the batteries.

00:50:18 Do you have a viewpoint on what is optimum size and range?

00:50:24 I know there's a dependency on the charging infrastructure,

00:50:28 but, you know, like, I keep thinking about how

00:50:32 first, after Tesla, some other manufacturers start coming,

00:50:36 there's this whole range anxiety.

00:50:37 Oh, how much is enough range?

00:50:39 And I feel like the Germans,

00:50:41 Porsche had the right idea, right?

00:50:42 Taycan, it's only really 200 miles of range,

00:50:44 but it's got super fast charging.

00:50:46 It makes a lot of sense,

00:50:48 but the dependency is this really crappy

00:50:50 charging network in the US, right?

00:50:50 >> Yeah, and it totally doesn't work at all.

00:50:52 >> Totally doesn't work, but it would in a future

00:50:53 where we had more chargers, right?

00:50:56 >> It's still not enough range.

00:50:57 >> Do we need 600 mile batteries or is it 300?

00:51:03 That charge to 100% in five minutes,

00:51:07 like what do you think?

00:51:09 >> It really depends by the use case,

00:51:10 how you drive the vehicle.

00:51:12 >> America, America, America, impatient Americans.

00:51:15 You've been here 15 years, like come on.

00:51:18 >> We have a solution that actually understand

00:51:20 how you drive the vehicle and how much juice

00:51:22 you need for the day after.

00:51:23 We call it EVI Smart Charging,

00:51:26 and we have seen that most of the time

00:51:27 you just need 20, 25% the day after at maximum.

00:51:31 So probably your battery could be less than half,

00:51:34 so the cost could be lower.

00:51:35 But it really depends by use case.

00:51:37 So in theory, in a couple of years,

00:51:39 we might reply to a questioner and say,

00:51:41 where do you go, how long do you drive per day,

00:51:44 and so on, do you go over the weekend long trips and so on?

00:51:47 And after you answer, you have your ideal battery pack

00:51:50 for your specific usage.

00:51:52 So there is no optimal solution,

00:51:55 really depends how you drive the vehicle.

00:51:57 >> He's a real agnostic, he's not committing to anything.

00:52:00 >> No, let's talk about that because you're Italian,

00:52:05 now you've been here for 15 years,

00:52:06 but you're so deep and very passionate about batteries.

00:52:10 This is all counter to your national origin, come on.

00:52:14 What cars do you, do you love cars?

00:52:19 Do you have any gas cars?

00:52:19 >> I love cars.

00:52:20 I love motorcycles, I go to the MotoGP race

00:52:23 is anywhere around the world,

00:52:25 and I actually ride the motorcycle as well.

00:52:27 >> You Ducati guy, you like the Doctor?

00:52:30 >> Ducati for sure, of course Valentino Rossi is an idol.

00:52:33 >> Yeah.

00:52:34 >> For sure, yeah, no, he's amazing.

00:52:36 But I would say that I ride my bicycle in Boston,

00:52:41 so I don't have a car.

00:52:45 Maybe when we're going to have the vehicle in production--

00:52:47 >> What do you do in February in Boston,

00:52:49 you're not riding a bicycle?

00:52:50 >> No, I ride bicycle even with a snowstorm.

00:52:52 >> You're a maniac.

00:52:53 >> But I'm waiting for the EVI solution,

00:52:55 a production vehicle, and then I get maybe two of those.

00:52:58 >> Okay, okay, well, this was awesome.

00:53:01 >> This was fun.

00:53:02 Too brief, but that's the nature of CES, yeah.

00:53:07 >> Electric vehicles, where can we find out more?

00:53:09 This is great, he's got his hat.

00:53:10 >> Oh no, show them the QR code.

00:53:11 >> There's a QR code, I'll hold up to the camera.

00:53:13 You can like take a picture of this.

00:53:15 >> Oh, hold your phone up to it.

00:53:16 >> And then hold your phone up to it.

00:53:17 Electric vehicles.

00:53:20 >> Oh, yeah, yeah, yeah.

00:53:23 We have a LinkedIn, electricvehicles.com website,

00:53:27 also YouTube channel, electricvehicles.com,

00:53:30 so check it out, follow us.

00:53:32 >> So that's his vehicles, I think I'm going to predict

00:53:34 it'll be in the Ram Rev, and maybe that--

00:53:37 >> I think it's Alfa Romeo, the next,

00:53:38 the Julia replacement.

00:53:39 >> Or that Charger--

00:53:40 >> He's Italian, the Julia replacement.

00:53:43 >> Yeah, but what are they driving

00:53:44 in this long distance haul?

00:53:46 At least his popular car.

00:53:48 >> I would say write in the comments.

00:53:50 >> Oh, yeah, put your guess in the comments.

00:53:52 >> Right, right.

00:53:53 >> Anyways, Mr. Martini, thank you so much.

00:53:54 >> Absolutely, yeah, that was great.

00:53:56 Thank you very much.

00:53:57 Push for innovation.

00:53:59 >> Well, that was amazing.

00:54:00 This is, again, this was three weeks ago

00:54:02 we talked to this guy, and I've been thinking about,

00:54:04 like, you know, my truck, there's like

00:54:06 an $18,000 battery in there, and I would love to know

00:54:10 where it's at, like what's the state of it,

00:54:12 you know what I mean?

00:54:13 And like, I think this is, I mean, again,

00:54:15 you have a virtual simulation of something you own

00:54:18 running in the cloud that they can give you information

00:54:22 about, tell you how to optimize it,

00:54:23 or when you go to sell it, what it's worth.

00:54:25 >> Well, yeah, I wasn't kidding when I told him

00:54:27 there's an entirely other line of business, I think,

00:54:29 where he could just provide the service

00:54:31 of evaluating the life left in people's batteries.

00:54:35 >> I mean, check the Carfax, right?

00:54:37 >> Exactly, that's exactly what needs to happen.

00:54:39 So, super interesting conversation on a topic

00:54:42 that, yeah, that I didn't ever think we would,

00:54:44 you know, never even broached my imagination.

00:54:47 Similarly, the next one is going to be fascinating.

00:54:52 I always thought Mobileye, and I kind of introduced

00:54:54 this guy incorrectly, I always thought it was just like

00:54:56 LiDAR, radar, it's actually vision systems,

00:54:59 and it's basically--

00:55:00 >> It's so much more than that.

00:55:01 >> It's machine learning, and how do you see,

00:55:03 how do you make machines see the world,

00:55:06 see the road ahead?

00:55:06 >> And by the way, your car you own right now,

00:55:08 odds are high if you bought it recently,

00:55:10 it's helping Mobileye to see the world.

00:55:13 >> Right.

00:55:14 >> Like, currently, right now, it's crazy.

00:55:15 >> So you're going to talk to, literally, a professor,

00:55:18 Shai Shalev-Schwartz, from Mobileye,

00:55:21 subject matter expert on vision systems,

00:55:23 including cameras, LiDAR, radar, as you will hear.

00:55:27 >> Doctor, doctor.

00:55:28 >> Professor.

00:55:29 >> Professor, Shai Shalev-Schwartz, from Mobileye.

00:55:32 Thank you so much for joining us here at The Inevitable.

00:55:36 Very first question, for me, I've always thought

00:55:39 of Mobileye as, oh, they do LiDAR, tell me,

00:55:43 what is the best way to think about,

00:55:46 or describe Mobileye as a company, what do you guys do?

00:55:49 >> So, we are, let's look at it a bit from a bird eye view.

00:55:54 >> Okay.

00:55:56 >> If we look at cars, for the last 20 years,

00:56:00 many things has been changed.

00:56:01 But maybe the most important two things, in my mind,

00:56:06 are first, safety, cars are much safer today.

00:56:11 Second, driving experience.

00:56:13 So if you look at cars today, passive safety,

00:56:17 active safety, it's really different

00:56:20 than what happened 20 years ago.

00:56:22 Also, in terms of driving experience,

00:56:25 maybe not 20 years ago, but 30 years ago,

00:56:29 we were driving manual transmission.

00:56:32 >> Right.

00:56:33 >> Okay, then came automatic transmission,

00:56:35 ACC systems, automatic cruise control.

00:56:40 >> Right.

00:56:41 >> Lane keeping assist, and now more and more autonomy,

00:56:44 what we call level two plus.

00:56:45 So the driving experience is shifting from driving the car

00:56:50 to supervising or monitoring what the system,

00:56:55 what the automated system is doing.

00:56:57 >> Okay.

00:56:58 >> So Mobileye is doing things along these two lines.

00:57:03 In terms of active safety systems,

00:57:08 probably, we say high probability if you buy a car today,

00:57:12 there is a Mobileye system inside, okay?

00:57:15 Responsible for alerting you for dangerous situations

00:57:19 with the front-facing,

00:57:21 responsible for autonomous emergency braking sometimes.

00:57:25 Okay?

00:57:25 So the basis of technology is a front-facing camera

00:57:30 that looks at the scene,

00:57:34 detect pedestrians, cars, lanes,

00:57:39 and monitors the driver behavior, okay?

00:57:43 So that's on the safety side.

00:57:45 On the driving experience side,

00:57:47 systems like autonomous cruise control

00:57:50 and lane keeping assist also relies on Mobileye technology

00:57:54 in many of the cars.

00:57:55 Not all of the cars, but in many of the cars.

00:57:57 And now the next step for Mobileye

00:58:02 is how you're going both with safety

00:58:04 and with driving experience to the next level, okay?

00:58:11 And what we are doing is we are doing it in incremental way.

00:58:15 So the next phase for us is called supervision system.

00:58:18 Supervision system is surround sensing.

00:58:21 We have 11 cameras around the car

00:58:24 and they can understand the entire environment

00:58:30 around the car.

00:58:31 And basically, similarly to Tesla FSD,

00:58:35 these systems can drive everywhere.

00:58:38 - With cameras?

00:58:39 - With cameras alone.

00:58:40 - Yeah, just camera vision, not radar, not LiDAR?

00:58:43 - Not radar, not LiDAR.

00:58:45 So of course if there are radars

00:58:47 in some of the car models that relies on supervision,

00:58:51 there is also a front radar.

00:58:54 When it is there, we use it,

00:58:55 but it is not a requirement, okay?

00:58:57 The system can do everything just by cameras, okay?

00:59:01 What this system can do,

00:59:02 it can basically drive itself everywhere.

00:59:05 How we achieve it?

00:59:08 Combination of perception, computer vision.

00:59:13 - How do you see the world, yeah?

00:59:14 - Mapping.

00:59:15 So we believe in whatever you can do offline,

00:59:19 do offline, not online, okay?

00:59:22 We leverage the fact that many, many cars

00:59:25 have our system for safety.

00:59:28 The front camera.

00:59:32 This camera, in any cases in the car,

00:59:36 needs to detect lane marks, traffic signs,

00:59:39 traffic lights for safety features.

00:59:41 So if all this work is already done, why not use it?

00:59:46 So what we are doing is we are uploading

00:59:49 light information from the cars to cloud.

00:59:53 The information is here is a piece of lane mark.

00:59:55 I saw a traffic light.

00:59:58 I saw a traffic sign, okay?

01:00:00 And then we stitch all of the information together

01:00:03 in the cloud to build a very, very accurate

01:00:06 high definition map.

01:00:07 We call it REM, road experience memory, okay?

01:00:12 This is not just a HD map.

01:00:14 It's also a mapping of how people are driving.

01:00:20 So if you are in Rome, you want to drive like a Roman.

01:00:27 So we are collecting information on the typical speed

01:00:32 that people are driving, whether they are driving

01:00:35 in the speed limit or not.

01:00:36 - So one of your colleagues gave us an example,

01:00:39 like there's a curve, and so the car might be capable

01:00:43 of maintaining 75 miles an hour around the curve,

01:00:46 but to the occupants, it would freak them out.

01:00:48 And then the car knows via REM that, hey,

01:00:51 everybody actually slows down to 55 around the corner,

01:00:53 do the same thing, even though the car's never been there.

01:00:56 - Exactly.

01:00:56 So when you are arriving to a new place,

01:00:59 because you have REM, you are driving like a local.

01:01:02 - Right, right.

01:01:03 - Even funny things like, you know,

01:01:06 there are these speed traps, cameras, okay?

01:01:10 So we know to drive and then slow down like everybody else.

01:01:13 - Right, well that, in Italy or Austria,

01:01:16 'cause they're being very handy these days.

01:01:17 - It sounds a little bit like what Waze has been doing

01:01:20 in the sharing of the information in real time.

01:01:23 So does that mean that REM, there's a model in the cloud

01:01:27 that is like a real time map, high definition map?

01:01:31 So if--

01:01:32 - But with local knowledge.

01:01:34 - Well, if let's say three cars equipped

01:01:35 with Mobileye technology and are going up to the REM,

01:01:39 pass a point at this, one right after another,

01:01:43 they're literally updating.

01:01:45 Like if a, I don't know, a squirrel fell out,

01:01:47 it jumped into the road, the first car caught it,

01:01:50 the second car caught it, crossed the road,

01:01:51 and the third car hit it, that is all being logged.

01:01:55 - And we actually, exactly, and we actually think about it

01:01:58 as a next gen of safety models,

01:02:00 because if there is a hazard on the road,

01:02:02 also for autonomous vehicles,

01:02:04 but also for just manually driven cars,

01:02:08 you can alert the driver in a much more effective way

01:02:10 than just telling him that, you know,

01:02:12 like Waze, Waze are telling you

01:02:14 that there is a stop card on the side,

01:02:17 but they don't have the accuracy of where exactly

01:02:19 and how exactly it should affect you.

01:02:21 - Right, 'cause they have no sensors.

01:02:22 - Yeah. - Right.

01:02:23 - We have accuracy of 10 centimeters

01:02:26 of everything that's happening, okay?

01:02:28 So we can tell you exactly.

01:02:30 - Another thing, again, I was talking

01:02:32 with one of your colleagues earlier,

01:02:33 but like, you know, for a long time now,

01:02:35 since probably 2014, you know, Tesla first,

01:02:38 a lot of other people do this,

01:02:39 but you know, they have geo-fencing.

01:02:40 So in other words, there's a speed bump by your house,

01:02:43 and so instead of you having to manually raise

01:02:45 your suspension every time, the car says,

01:02:47 "I know where I am," and it does it.

01:02:49 But another Tesla never does it, right?

01:02:52 But so this is like cars equipped with mobilized stuff,

01:02:55 the follow ram, they will know,

01:02:56 hey, there's a speed bump there, put the suspension up

01:02:59 if that's what you need to do.

01:03:00 - Exactly. - Right.

01:03:01 - So, and also, when we are driving with our system,

01:03:05 our automated system, which is called Supervision,

01:03:08 then it knows in advance that there is a speed bump,

01:03:11 and we know also in advance how the crowd

01:03:15 is driving at this speed bump.

01:03:17 - And then you can say, okay. - People take it 45, yeah.

01:03:20 - And then you can say, okay, I want to be like

01:03:23 the 20% low of the crowd, or 50%, or 70%.

01:03:28 So you can define as a driver how you want to,

01:03:32 how you want to compare yourself to other drivers.

01:03:36 Okay, so I like to arrive to speed bumps slowly.

01:03:40 Okay, so I can do it relatively to what others are doing.

01:03:44 Okay, so it opens a lot of options.

01:03:47 So one of the-- - Real quick, real quick.

01:03:49 Did you, 'cause I know you're Israeli, and Waze is Israeli.

01:03:53 Was there like mutual crossover between the two companies?

01:03:57 - Not really, not really.

01:03:59 But you know, great mind thinks alike.

01:04:01 - Well, I'm glad you guys are,

01:04:03 I was going to ask a Waze question.

01:04:05 I'm a huge fan of Waze.

01:04:07 We adopted them at Motor Trend very early on,

01:04:09 and the technology I still use it to this day.

01:04:12 One of the most powerful things they did

01:04:13 in terms of demonstrating the mapping,

01:04:16 'cause they had the most accurate and up-to-date maps

01:04:18 because of people who were using it, right?

01:04:19 And they would show this,

01:04:20 I think they showed the 24-hour map of Paris,

01:04:25 and you could see the peak of the day,

01:04:26 all these people, and redrawing the map.

01:04:29 Does something like that exist at Mobileye,

01:04:31 where you can, as the day starts,

01:04:34 you can start to see the usage,

01:04:35 and this cloud, this data, whatever,

01:04:39 set, grow and populate?

01:04:42 And it'd be a very interesting visual to see, right?

01:04:45 - So, yeah, it's of course possible,

01:04:50 but we are building it with layers,

01:04:54 and the most important layer for the start,

01:04:58 which is complicated enough,

01:04:59 is just understanding really to 10-centimeter accuracy

01:05:03 everything that is needed for an automated system

01:05:05 in the car.

01:05:06 But of course, these ideas are up and running,

01:05:10 and it depends on collaboration that we are doing

01:05:13 with OEMs, traffic information,

01:05:15 what else can we extract from this asset,

01:05:18 because it's a great asset.

01:05:20 - Now, with REM, you're using camera data,

01:05:24 but I know you guys also use radar and LIDAR

01:05:27 and other systems you make.

01:05:28 Is radar and LIDAR information fed into REM,

01:05:31 or is it just camera?

01:05:32 - So, no cameras at all.

01:05:35 What we upload to the cloud

01:05:37 is processed information from the camera.

01:05:40 - I'm sorry, processed information from radar and LIDAR,

01:05:42 does that go up as well?

01:05:44 - Yeah. - Why not?

01:05:45 - Okay. - Why not?

01:05:45 I mean, if the car has other sensors,

01:05:49 what we upload to the car is our understanding,

01:05:52 this car understanding of the surrounding.

01:05:54 And the idea is that understanding of the surrounding

01:05:57 is very, very accurate in the close range.

01:05:59 The real problem is what happens in the far range.

01:06:03 - Do you take any-- - But because--

01:06:04 - Just real quick, do you take any data

01:06:05 from the car itself?

01:06:06 In other words, probably take vehicle speed,

01:06:09 but like ambient temperature,

01:06:11 or I don't know, what, tire pressure,

01:06:13 or anything like that,

01:06:14 that might be relevant?

01:06:16 - Yeah, it is relevant, actually,

01:06:17 because the tire pressure, for example, is relevant

01:06:20 because if you look at how cars

01:06:23 are actually calculating speed,

01:06:25 it depends on the car pressure.

01:06:26 - Yes, yes, okay.

01:06:28 - And speed is very important

01:06:29 because in order to understand to 10 centimeters accuracy,

01:06:34 what we reach from the car,

01:06:35 we want to do what is called ego motion,

01:06:37 so understand how the car is moving

01:06:39 relatively to the ego.

01:06:40 And for that, you need very, very accurate speed.

01:06:42 So everything is connected,

01:06:44 and the nice thing is that it's not something

01:06:47 that we are doing for the mapping.

01:06:49 We're doing it anyway for the safety system

01:06:52 that is on the car.

01:06:54 So if we are anyway doing it,

01:06:57 why not piggybacking on this information

01:06:59 and just upload the card,

01:07:00 and then everybody can enjoy it?

01:07:02 Another thing which is very interesting here

01:07:05 is what we call semantic information.

01:07:08 So for example, who has a priority in some intersection?

01:07:13 - Right.

01:07:15 We talked a little bit about, yes, yes, this is interesting.

01:07:17 - So if you just look at the traditional way

01:07:19 to construct high-definition maps,

01:07:22 then someone needs to figure out the priority.

01:07:25 It's not something,

01:07:27 even if you have a lighter on the roof,

01:07:30 it doesn't tell you who has priority.

01:07:32 So either you need to read traffic signs

01:07:35 and have some logic on how you set priority,

01:07:40 or manually label that priority,

01:07:43 but then you have human errors.

01:07:45 In REM, we can look not only on the legal priority,

01:07:51 but also how the crowd interprets this priority.

01:07:56 - So in Rome, everybody hits each other.

01:07:58 (laughing)

01:07:59 - Scream, yeah.

01:08:00 I mean, look at some examples.

01:08:04 So when you are merging into the highway, okay?

01:08:07 In some countries, so there is a priority there.

01:08:10 The guy on the highway has a priority

01:08:12 over the guy that is merging, okay?

01:08:13 - Except for in Miami.

01:08:15 - Yeah, so in some places,

01:08:18 in some places, so you just ignore the priority and fit in.

01:08:23 In some places, people,

01:08:26 when they are arriving to emerge from the highway,

01:08:30 they are performing a lane change, okay?

01:08:35 In order to be polite for the people that are fitting in.

01:08:39 So really, you want to understand the culture, okay?

01:08:44 And this approach helps you understand

01:08:46 the culture of driving.

01:08:47 - Okay, I'd love to see your Moscow map.

01:08:50 I'd love it.

01:08:51 (laughing)

01:08:52 - Or the train, those intersections in Vietnam.

01:08:55 - Oh, like Shibuya.

01:08:56 - In Vietnam, with all the scooters and everybody.

01:08:58 - Right, right.

01:08:59 - Okay, so, 'cause I'm conscious we're running out of time.

01:09:02 In your booth, you have two vehicles on display.

01:09:06 You have a Polestar, and then you have a--

01:09:08 - Polestar 4.

01:09:09 - Polestar 4, and you have a VW ID Buzz,

01:09:11 which you don't get yet in America.

01:09:13 And they're--

01:09:13 - We don't get the 4 either.

01:09:15 - The Buzz has a bunch of really cool bulges and plates

01:09:20 and things on the roof, more than just cameras.

01:09:26 So, what's that all about?

01:09:28 What else do you guys do?

01:09:29 - So, it's all about how you reach autonomy.

01:09:33 And there are two paths, okay?

01:09:35 And it's important to understand

01:09:36 that they are really different.

01:09:39 One path is start from bottom up, okay?

01:09:43 So, start from front-facing cameras

01:09:45 and have ACC, LKA, stuff like that.

01:09:47 - Adaptive Cruise Control, Lane Keep Assist.

01:09:49 - Yes, yeah.

01:09:50 And then go to supervision system.

01:09:52 It's our next gen, which has cameras all around,

01:09:56 and it can drive itself everywhere.

01:09:58 But it's not, we call it a hands-off, eyes-on system.

01:10:03 Because the human driver still needs

01:10:06 to supervise the system.

01:10:08 It's not good enough in order to drive itself

01:10:11 and you can--

01:10:11 - So, camera base is not good enough

01:10:12 for full self-driving, shall we say.

01:10:14 - Exactly.

01:10:15 - Level five. - Exactly.

01:10:16 - It's not good enough for level five.

01:10:17 - It's not enough even for level three, okay?

01:10:19 - It's not enough, I like this.

01:10:21 - Okay, I will--

01:10:22 - Thank you, I like this a lot.

01:10:23 - So, how you go up--

01:10:24 - Real quick, real quick.

01:10:25 Just level three in the US at least means

01:10:28 that Polestar assumes the legal liability

01:10:32 for what the car is doing, not the person behind the wheel.

01:10:35 So, just by having just vision, Tesla,

01:10:38 it's not good enough to assume legal liability.

01:10:42 - But let me not talk about liability instead.

01:10:45 Talk about what we as a society,

01:10:47 because liability is the aftermath, okay?

01:10:50 - Oh, I agree, yeah.

01:10:51 - I want to say, okay, we need to be responsible, okay?

01:10:55 What is a responsible thing to do?

01:10:57 - You're saying it's the ethical thing to do.

01:10:58 - It's the ethical thing to do.

01:10:59 You don't want to put a car, which is not,

01:11:03 we call it MTBF, mean time between failure, okay?

01:11:06 So, the mean time between failure of a system

01:11:08 in order to be automated and eyes off,

01:11:11 it should be superhuman.

01:11:13 If it's superhuman, okay, we can discuss

01:11:15 how much we want more than superhuman.

01:11:18 But if it's not superhuman, for sure,

01:11:20 it cannot be eyes off, period.

01:11:22 We don't want machines that are worse

01:11:25 than humans to drive by themselves.

01:11:27 - So, to be superhuman, you need a combination

01:11:29 of cameras, radar, and LIDAR?

01:11:32 - In our opinion, yes.

01:11:34 In our opinion, yes, and let me explain why.

01:11:37 But maybe before I dive into it,

01:11:40 just going back to the Polestar versus iDubbbz.

01:11:44 So, bottom up is front-facing,

01:11:47 then supervision, cameras only.

01:11:49 Then you add sensors for redundancy

01:11:51 and you can go eyes off.

01:11:53 And gradually, you expand what we call the ODD,

01:11:56 operational design domain.

01:11:57 So, initially, maybe only on highways,

01:11:59 you can be eyes off.

01:12:01 And then more roads and more roads.

01:12:03 But this is a consumer cars that gets

01:12:06 more and more automated systems.

01:12:09 The other directions, let's call it top-down,

01:12:13 is robo-taxi, in which you say,

01:12:15 okay, I want level four from day one.

01:12:19 No driver in the car.

01:12:20 And then you say, okay, but it's a very,

01:12:24 very difficult problem.

01:12:25 So, I need many sensors.

01:12:27 Again, we want to be superhuman.

01:12:31 And maybe I want to be geofenced.

01:12:33 Maybe I want to start only with one city,

01:12:37 and even in this city, only one geofenced area.

01:12:39 - Right, so we've mastered the east side of a town.

01:12:43 - Or parts of Austin, or parts of the area.

01:12:45 - Exactly, so these are two direction

01:12:47 how to reach automation.

01:12:49 Eventually, they will merge, okay?

01:12:52 But do you want from here or from there?

01:12:54 Tesla are maybe the louder company

01:12:59 that talks from bottom up.

01:13:01 Waymo is maybe the louder company

01:13:03 that talks about top-down, okay?

01:13:06 Mobileye, we honestly, when we started developing,

01:13:11 we didn't know how the future is going to look like.

01:13:14 There were good arguments that it's going to be top-down,

01:13:16 good arguments that it's going to be bottom-up.

01:13:18 - It would be cheaper if it was just camera.

01:13:20 - Yeah.

01:13:21 - From a hardware point of view.

01:13:22 - You said you support both.

01:13:23 - And then what we said, we said,

01:13:26 the essence of the technology,

01:13:27 we build it in a way that it supports both.

01:13:30 So let's do both, let's hedge, let's do both directions.

01:13:33 Now, why is it very, very different from,

01:13:36 for example, the approach that Waymo took?

01:13:38 Waymo took a LiDAR-centric approach.

01:13:41 So for doing the top-down, they started from,

01:13:43 we require a very, very strong LiDAR, okay?

01:13:46 And then you cannot do both,

01:13:49 because there will be no strong LiDAR for consumer cars.

01:13:52 It's too expensive, okay?

01:13:55 But our approach, even for the top-down,

01:13:57 was that, first of all, you need camera-centric approach.

01:14:01 Roads are built for humans, and humans have eyes,

01:14:05 no LiDARs, okay?

01:14:06 - Okay, yeah.

01:14:07 - So the major sensor should be cameras, okay?

01:14:12 But it's not enough.

01:14:14 So build the system for cameras, okay?

01:14:17 And then add layers of redundancy

01:14:19 in order to reach the high MTBF

01:14:21 that is required to be superhuman.

01:14:23 So since we are doing it, we can do both,

01:14:26 bottom-up and top-down with the same technology.

01:14:28 - Let's fill in the blanks for the audience who's like,

01:14:31 well, I don't,

01:14:32 why, what is, why can't it,

01:14:36 what does LiDAR and radar do that cameras cannot?

01:14:41 Why do you need that?

01:14:43 And I know some of it's weather,

01:14:44 like seeing in the dark, all this kind of stuff.

01:14:47 So what are the advantages of those systems?

01:14:49 - So I don't want to say that they can do things

01:14:52 that camera can't do,

01:14:54 because you can drive with only your eyes,

01:14:56 and you can drive in all conditions.

01:14:57 So, okay, very good.

01:14:59 This is not my point.

01:15:00 My point is that these are different sensors,

01:15:02 and in what way they are different.

01:15:05 Cameras, by and large, okay, is a passive sensor,

01:15:09 not an active sensor.

01:15:11 Passive sensor means that there is no source of light.

01:15:13 We use another source of light,

01:15:15 either the sun, if it's in the day,

01:15:18 or the lights, headlights, okay, if it's at night, okay?

01:15:23 Unlike radar and LiDAR, they are active sensor,

01:15:26 meaning that they are shooting some signal, okay,

01:15:30 getting back the returns from the signal,

01:15:33 and based on that, understand something.

01:15:35 So this is the first difference between these sensors.

01:15:39 The second difference is that camera

01:15:42 is a very high resolution sensor,

01:15:44 relatively to radar and LiDAR.

01:15:47 So camera, today technologies of camera,

01:15:50 is much, much higher resolution than LiDARs and radars,

01:15:55 meaning that you can understand semantic information,

01:15:57 which you can't understand from radar and LiDAR.

01:16:00 So, for example, you can't read a traffic sign by LiDAR.

01:16:05 >> And LiDAR, there's like 150 lines of resolution,

01:16:11 whereas with cameras now it's in the billions?

01:16:14 >> Thousand, yeah.

01:16:15 >> A thousand, okay, yeah.

01:16:16 >> Millions of pixels.

01:16:18 >> Yeah, okay, so it's much better.

01:16:20 >> Yeah, yeah.

01:16:21 >> So this is the second difference.

01:16:25 Okay, the third difference is that cameras are 2D sensors,

01:16:30 while radars and LiDARs are 3D sensors.

01:16:34 What does it mean?

01:16:35 The raw signal of radars and LiDARs

01:16:38 is a point in the 3D world, okay?

01:16:42 On the other hand, in camera, what we really see

01:16:45 is a projection of this point on a 2D space.

01:16:49 So we see part of the information, not all the information.

01:16:54 And this is some of the reason that we have

01:16:58 a hallucination in cameras, okay?

01:17:03 Is because I can show you a video of a car in a flood,

01:17:08 and then I will zoom out and you see that it's a toy car

01:17:11 in a small pond, okay?

01:17:13 >> Right, right.

01:17:14 And so, for instance, for high speed, LiDAR is good

01:17:17 because it can see 500 meters down the road,

01:17:20 and it knows the dimensionality,

01:17:24 it knows something's coming 500 meters away.

01:17:27 Where's the camera?

01:17:28 >> Well, cameras can see also 500 meters away,

01:17:30 it's a matter of resolution.

01:17:32 And most LiDARs can't see 500 meters away,

01:17:35 so it's not just a matter of range, okay?

01:17:38 It's a matter of the type of signal that you are getting.

01:17:41 In LiDAR, you are getting, okay,

01:17:43 I know that something is 200 meters away

01:17:46 at an accuracy of 10 centimeter.

01:17:48 In camera, what you see is a reflection

01:17:50 or a projection of it on the 2D.

01:17:52 >> That's more what I meant, yeah.

01:17:53 >> And then the difficult part in camera is,

01:17:57 since the control of the car is done in the 3D world,

01:18:01 you need to lift back the information from the 2D

01:18:04 to the 3D, which is very difficult.

01:18:06 >> It's a lot of processing.

01:18:07 >> Exactly, this is why it requires a lot of processing.

01:18:10 >> So if you have 3D sensors,

01:18:12 then it doesn't require as much processing

01:18:13 'cause it's already in the 3D language.

01:18:15 >> Exactly, exactly, so who is better?

01:18:17 Which sensor is better?

01:18:19 My point is, all of them has problems.

01:18:23 All of them has disadvantages.

01:18:25 All of them have advantages.

01:18:27 So the game shouldn't be, you know,

01:18:28 this is the ultimate sensor.

01:18:31 Likewise, what algorithm, which algorithm is the best?

01:18:33 So you can hear some players are saying,

01:18:36 okay, we want end-to-end algorithms,

01:18:38 the AI will figure everything.

01:18:41 And other players are saying,

01:18:43 no, we want a decomposable approach,

01:18:45 we want to do it in parts.

01:18:48 What we are saying is, again,

01:18:50 every method has advantages and disadvantages.

01:18:54 Don't be religious about methods, okay?

01:18:57 You want as many as you can, build redundancy, okay?

01:19:02 And this is how you can reach 99.9999%

01:19:06 the MTBF, which is required in order to let

01:19:09 the driver eyes off the road, okay?

01:19:11 So just set the bar and do whatever you can

01:19:16 in order to reach it.

01:19:17 - Okay, this is great, 'cause you're giving me

01:19:20 the jumping off point for the next question,

01:19:22 which is, okay, so now I'm pretty clear.

01:19:25 Mobileye can provide a range of solutions,

01:19:28 inclusive of camera systems, vision,

01:19:32 you know, the radar, lidar, you build the chips,

01:19:36 you build the camera systems,

01:19:38 you support all these manufacturers.

01:19:40 On the road towards potentially full autonomous driving.

01:19:45 There's been a lot of hiccups in autonomous, right?

01:19:47 We saw some very well publicized things happen

01:19:50 in the last 18 months or so, Argo AI went kaput.

01:19:53 - Cruise.

01:19:54 - Cruise had that huge incident.

01:19:57 I'm curious, well, I'll just get to the,

01:20:03 'cause we're running out of time,

01:20:04 we'll get to the core question,

01:20:05 but I want to back out later and ask you

01:20:06 just for your overall sense of where we're going,

01:20:08 how fast autonomous is going to get here,

01:20:10 what do you think about it?

01:20:11 But what's your position on human in the loop?

01:20:14 You know, you talked about eyes off,

01:20:16 I saw some great demos in your booth

01:20:18 of some cool things that the car's doing,

01:20:21 it moves out of the way from a semi truck.

01:20:23 You know, I follow a critic in this space,

01:20:26 Dr. Missy Cummings, she was formerly

01:20:28 of the Biden administration, NITSA, DOT,

01:20:33 and she's been very critical of the way

01:20:36 a lot of these autonomous leading companies

01:20:38 are in their approach.

01:20:39 And one of her main arguments is that,

01:20:42 when you have a human in the passenger seat

01:20:43 and they're supposed to basically take over

01:20:45 when the superhuman system is not so super,

01:20:49 like that's bad.

01:20:51 Like you're basically, these systems

01:20:53 are essentially training people to not engage

01:20:56 and the reaction time for them to come off

01:20:58 of whatever they were doing--

01:20:59 - Wake up, get off your phone, whatever, yeah.

01:21:01 - Is, creates a real problem,

01:21:02 and I'm probably butchering that,

01:21:04 you know, in paraphrasing it.

01:21:06 But where do you sit on this,

01:21:09 what's Mobilized position?

01:21:11 - So, it's a very good point,

01:21:14 and Mobilized position is, again,

01:21:16 we have the bottom up and the top down, okay?

01:21:18 In the top down we can talk, maybe later,

01:21:21 about safety and about exactly what does it mean.

01:21:26 Because I think it's very important

01:21:29 to also crystallize what does it mean by safety.

01:21:32 - Yes, exactly, safety for who?

01:21:34 - Safety for who, exactly.

01:21:36 In the bottom up, in the consumer cars, okay,

01:21:40 yes, there is an issue,

01:21:42 and our approach is that we must be in a position

01:21:47 where human plus machine is better than human alone

01:21:52 and is better than machine alone.

01:21:53 We don't want to be in a position

01:21:55 that we put a semi-automated system

01:21:58 and end up with a system which is worse than human alone.

01:22:05 Okay, we want to be better, okay?

01:22:08 And part of the approach to tackle this

01:22:11 is from day one in supervision system,

01:22:15 there is a driver monitoring system.

01:22:17 So if the system is not good enough

01:22:20 to allow driver to be eyes off the road,

01:22:24 we must verify that eyes are on the road.

01:22:27 And not rely on sensors that detect

01:22:29 if he puts hands on the wheel

01:22:31 because then he will put something on the wheel

01:22:33 and/or will cheat.

01:22:34 - My trick is I lower it down onto my knee.

01:22:37 Can't tell a hand from a knee.

01:22:38 - Shh.

01:22:39 - Oh, sorry, sorry.

01:22:40 - So we want a driver monitoring system

01:22:43 that will make sure that your eyes on the road,

01:22:46 on one hand, we don't want to annoy the driver

01:22:48 and tell them every 20 seconds,

01:22:50 it's just annoying and it's not helpful.

01:22:53 We want it to be helpful.

01:22:54 And in order to be helpful,

01:22:55 you want a driver monitoring system.

01:22:58 This is part of supervision.

01:23:00 So in any step that we are doing,

01:23:03 we have the mantra of,

01:23:04 we want that the combination of machine plus human

01:23:07 will be better than human.

01:23:09 And we think that there are many opportunities here

01:23:12 in the fact that you have a strong computer

01:23:15 that looks 360 around the car,

01:23:18 opens up many, many opportunities for higher safety.

01:23:21 So if I go back to what I started

01:23:24 by the two things that shaped car industry

01:23:28 in the last 20 years,

01:23:30 safety and driver experience,

01:23:32 we want that supervision will help both.

01:23:35 It will help safety because the fact that you have 360,

01:23:39 the fact that you have road experience memories,

01:23:42 a mapping, okay, that alerts you on potential hazards,

01:23:45 makes the system safer than those

01:23:48 that only has front-facing camera.

01:23:50 And on the other hand,

01:23:51 we want to improve the driver experience.

01:23:55 - So isn't, but isn't the argument that until the system

01:23:58 of machine plus human is better than human,

01:24:02 the fail-safe shouldn't be a human,

01:24:04 you should take it out until you have a system

01:24:07 that is 100% of the time not reliant on a human overseer.

01:24:12 - So again, so we need to look at statistics

01:24:15 and see if the human plus machine

01:24:18 is better than human or not.

01:24:19 If it's not better than human, then yeah, you're right.

01:24:23 We need to wait with this approach, okay?

01:24:26 But if it's better than human,

01:24:28 then we should go for it because it's better.

01:24:30 - And we're seeing that with automatic emergency braking.

01:24:33 - Exactly.

01:24:34 - Is especially in the age of, you know,

01:24:36 while you're driving, that it is safer.

01:24:39 - All right, we're almost out of time.

01:24:41 How do you see the autonomous world?

01:24:44 Given all these sort of recent setbacks,

01:24:46 you know, our crews got their permit pulled,

01:24:50 at the same time Mercedes got level three

01:24:52 driving certification in the US.

01:24:54 - No, no, no, California, Nevada.

01:24:55 - California, Nevada.

01:24:57 Where are we?

01:24:58 How are we, are we on?

01:25:00 What year does the steering wheel leave the car?

01:25:02 - Yeah, all the bigger picture questions.

01:25:05 And is this like a blip in the roadmap or just, hey?

01:25:09 - So the way I see it, first I think

01:25:11 that this is not the right question.

01:25:13 - Huh?

01:25:14 - All right, so the right question is not

01:25:16 when's the steering wheel out of the car.

01:25:18 The right question is, when do you see jump

01:25:23 in the driver experience and safety of your car?

01:25:26 When your car as a consumer will be totally different

01:25:29 than what it is today.

01:25:30 And I think it is happening now,

01:25:32 and it is happening fast, and you will see it

01:25:35 in the coming three to four years

01:25:39 in a massive amount of cars.

01:25:41 And what's the revolution that I see?

01:25:44 Two revolutions.

01:25:45 First, level three systems which are really useful,

01:25:49 and I don't want to say who is not useful,

01:25:52 but level three systems that can really take

01:25:56 the driving in the boring elements like highways

01:26:01 and traffic jams up to 130 kph,

01:26:04 not just very, very slow driving.

01:26:06 - 80 miles an hour.

01:26:08 - Yeah, 80 miles per hour.

01:26:09 - Sorry, we have Americans in the room.

01:26:11 - So this is coming.

01:26:14 Mobilite has design wins on these cars.

01:26:16 They are coming, okay?

01:26:18 When you buy such a car, and you can have eyes off

01:26:21 on the road, not everywhere, not in urban roads,

01:26:24 but in highways, roads, and in traffic jams,

01:26:27 you will see a huge jump in driver experience.

01:26:29 So this is one thing that will change your life, okay?

01:26:32 You can buy back your time, okay?

01:26:34 - Or waste it on social media, either way.

01:26:37 - However you like.

01:26:38 (laughing)

01:26:39 The other stream which I think is important

01:26:42 is what we call Vision Zero.

01:26:44 What is Vision Zero?

01:26:45 So the Sweden Parliament had a decision in 1997

01:26:49 to reach a state where there are zero fatalities

01:26:53 from car accidents, okay?

01:26:55 And the idea was that you should design roads

01:27:00 in a robust manner that will be tolerant to human mistakes.

01:27:05 I think the same will happen now with supervision

01:27:08 and systems like this, that they can bring levels of safety,

01:27:13 which I will not say zero, but they will give a huge jump

01:27:20 in safety, in that we want to decrease significantly

01:27:25 the amount of people that are being hurt

01:27:27 or die from car accidents.

01:27:28 - Which is a massive number.

01:27:29 When you actually read the global,

01:27:31 I mean, the US, we're back up over 40,000 deaths per year,

01:27:35 but globally, it's way higher.

01:27:37 I mean, some of that's poverty-related,

01:27:40 you know, a lot of things like that, but.

01:27:42 - So I think there is an opportunity here,

01:27:44 and I think that two to five years from now,

01:27:49 so we already have these systems in a small number.

01:27:53 The big jump is coming in the coming two years.

01:27:57 I think that in five years, we will also have,

01:27:59 in big numbers, eyes-off systems.

01:28:02 And then if we look back, in five years from now,

01:28:04 on today's interview, we will see a jump in safety

01:28:09 and a jump in driving experience.

01:28:12 Cars will look very, very different, okay?

01:28:15 So I think it's coming.

01:28:17 We talked about bottom-up and top-down, okay?

01:28:19 I talked about the revolution

01:28:22 that is coming from bottom-up.

01:28:24 I still don't know the top-down part.

01:28:28 The main issue there is not technological,

01:28:30 it's more business side of the things.

01:28:33 - Legality.

01:28:34 - Exactly. - Because it's so expensive.

01:28:36 - It's expensive, and can it be a viable business?

01:28:39 - Interesting. - Okay, okay.

01:28:41 - All right.

01:28:42 This is great, and I love that you put us on the clock

01:28:46 for five years, so we'll definitely chat with you,

01:28:48 I hope sooner than five years.

01:28:50 - Yeah, yeah, yeah.

01:28:51 - My last questions, two questions are,

01:28:54 what do you drive?

01:28:55 Okay, two.

01:28:56 And then, and this is a feel-free to brag a little bit,

01:29:00 or what can our listeners, or people watching this,

01:29:04 what is the best experience that they can have

01:29:08 in an advanced ADAS, or semi-autonomous driving system,

01:29:12 today, or maybe this year, coming out?

01:29:14 Is there a vehicle?

01:29:16 Is there a car on sale right now that you're,

01:29:18 a lot of Mobileye technology that you're proud of?

01:29:20 - That you support, that is, which one is it?

01:29:23 What would be the best case if someone's like,

01:29:24 man, I want to experience what this guy's talking about?

01:29:28 So, I'm driving today, the car that I'm driving

01:29:31 is a Zeker car.

01:29:32 It's the first system that came with

01:29:36 Mobileye supervision system.

01:29:38 And I can tell you--

01:29:39 - Zeker's a Chinese company.

01:29:39 - Zeker's, it's a Geely, it's part of Geely Group,

01:29:42 and it's a great car, and I love it,

01:29:46 and especially, it takes me from point to point,

01:29:50 okay, so I enter the car, I put an address,

01:29:54 it drives there.

01:29:55 - Is this the van?

01:29:56 Is this like the--

01:29:57 - No, it's not a van, it's a, how do you call it,

01:30:00 in compact car?

01:30:01 - SUV?

01:30:04 - It's not an SUV, exactly.

01:30:06 - Hatchback.

01:30:07 - Yeah, something.

01:30:08 - Shooting brake. - Shooting brake.

01:30:09 - Station wire. - Station wire, okay.

01:30:11 - So, this is a car that I'm riding,

01:30:16 and I'm using it, we all the time put new versions on it

01:30:19 in order, I test it myself, so I believe in hands-on.

01:30:22 I code myself, and I test the car myself.

01:30:24 - Oh, wow, okay.

01:30:25 - Okay?

01:30:26 And for fun, I ride motorcycles.

01:30:29 - Oh, okay.

01:30:30 (laughing)

01:30:31 So, the opposite, no safety whatsoever.

01:30:34 - That's good.

01:30:35 - Yeah, yeah, okay.

01:30:36 - But now--

01:30:37 - What kind of motorcycle are you riding?

01:30:39 - I'm riding a Honda.

01:30:40 - Okay.

01:30:41 - But--

01:30:41 - Which one, which Honda?

01:30:42 - The 750 NC700.

01:30:44 - Oh, big one, okay.

01:30:45 - Yeah, NC700, okay.

01:30:46 - Yeah, yeah, all right, nice.

01:30:48 - So, regarding car experience,

01:30:52 and what will be different in the coming years,

01:30:57 so Polestar, I wish that you could experience the Ziccer,

01:31:01 but it's not coming to US soon.

01:31:04 It's coming to Europe, by the way.

01:31:06 Polestar will have, again, the supervision system.

01:31:10 - So, we're gonna get, you know,

01:31:11 we'll get Polestars, Volvos, and Lotus.

01:31:15 So, the big Lotus SUV.

01:31:17 - And these are based on IQ5.

01:31:19 The next gen, IQ6, will be a big jump.

01:31:23 And we announced on, you know,

01:31:25 design wins for the IQ6, it will be a big jump.

01:31:29 We measure jumps by MTBF, by mean time between failures.

01:31:33 And we are getting to point where, you know,

01:31:36 most of the time you don't need to interfere with the car.

01:31:39 Okay?

01:31:40 But the highest jump will be when

01:31:43 we will have eyes off systems,

01:31:46 and this is coming, start of production, end of 2026.

01:31:51 So, 2027, I think this will be the big jump.

01:31:55 If you can buy a car, and you can drive on the highway

01:31:58 and traffic jumps without looking at the road,

01:32:00 this will really be a revolution.

01:32:02 - What's the current state of the art for MTB?

01:32:04 What's a good number?

01:32:05 - So, MTBF, mean time between failure.

01:32:08 So, to give you some statistics,

01:32:12 currently, we are around 50 hours.

01:32:17 Okay?

01:32:18 So, meaning that you can drive 50 hours

01:32:20 without intervention.

01:32:22 Okay?

01:32:23 And this was done, not by us,

01:32:24 it was done by people that evaluated.

01:32:28 - So, that's a lot different than my experience,

01:32:30 which 50 seconds would be a very high number.

01:32:32 - Yeah, but this also depends on the ODD.

01:32:35 It depends where you are driving.

01:32:37 Okay?

01:32:38 - I'm saying with every car I've driven,

01:32:40 I've driven a lot of level two stuff,

01:32:42 a lot of, you know, I know like Tesla and Volvo,

01:32:45 they don't even like to use the levels,

01:32:47 but it's always failing.

01:32:49 - Yeah, yeah.

01:32:50 But still, it sounds good,

01:32:55 but it's horrible in terms of, you know,

01:32:56 you cannot let eyes off.

01:32:58 - Right, and you don't want any failure.

01:33:00 - Yeah, right.

01:33:01 - Yeah, yeah.

01:33:01 - The road map is that these systems will go to,

01:33:06 so for eyes off, we're talking about millions

01:33:09 or 10 millions of hours MTBF.

01:33:13 Otherwise--

01:33:14 - How do you even measure, so that's like collectively,

01:33:15 in other words, you have a fleet of vehicles,

01:33:17 and every million hours driven, there's one failure.

01:33:20 - Exactly.

01:33:21 - Okay, wow, okay.

01:33:22 - Which is better than human statistics.

01:33:24 - Yeah.

01:33:24 - And this is where we want to be.

01:33:26 We want to be better than human statistics.

01:33:27 - Sure, okay.

01:33:28 - And then overall, you reduce number of car accidents.

01:33:31 - Right, okay.

01:33:32 - Right.

01:33:33 - All right, well we're going to have to leave it at that.

01:33:33 - Wow, professor, that was very enlightening.

01:33:36 - Very good.

01:33:37 - Thank you so much for your time.

01:33:39 - Thank you for hosting me.

01:33:39 - Yes, and we would love to check in with you

01:33:42 on the road to the next five years.

01:33:44 Maybe we'll come and chat once we--

01:33:45 - I used to work with your man over there,

01:33:47 so we know each other pretty well.

01:33:49 - But yeah, we appreciate the time.

01:33:51 Best of luck.

01:33:52 - Yeah.

01:33:52 - And thank you so much for being on the Inevitable.

01:33:54 - Thank you very much.

01:33:55 (dramatic music)

01:33:58 (dramatic music)

01:34:01 (dramatic music)

01:34:03 (dramatic music)

01:34:07 (dramatic music)

01:34:09 (dramatic music)

01:34:12 (dramatic music)

01:34:15 (dramatic music)

01:34:18 (dramatic music)

01:34:20 [BLANK_AUDIO]

Category

Transcript

Recommended