"Globes" names Mobileye Israel's tech co of the decade

Ziv Aviram and Amnon Shashua Photo: PR
Ziv Aviram and Amnon Shashua Photo: PR

Shashua: We went from being a supplier in 2014 to being a partner of BMW and Intel in July 2016.

Mobileye(NYSE: MBLY), the "Globes" tech company of the decade, which has already existed for 17 years, is a leader in technology for computer vision, machine learning, information analysis, location and mapping for advanced driving systems, and autonomous driving systems. Its systems are installed in 13 million vehicles around the world, and some readers have undoubtedly witnessed the performance of some of their simpler systems.

In order to get to know the company better, other than through its financial numbers, we conversed at length with Mobileye cofounders chairman Amnon Shashua (the company's technology brain) and CEO Ziv Aviram. They usually prefer to stay away from the media, and since the company held its Wall Street IPO, they have given very few interviews. We also talked with Colmobil auto agency (importer of Mercedes, Hyundai, and Mitsubishi cars) controlling shareholder Shmuel Harlap, the largest private shareholder in Mobileye.

These three men share the same theme: the auto industry is undergoing a once-in-a-century revolution - the biggest since the invention of the car - a revolution that will eventually make ownership of a car superfluous by integrating an autonomous car with a shared car. Mobileye is well-positioned in the forefront of the technological developments of an automated car, but in technology, of course, things are more dynamic than in other markets.

In any case, Shashua, Aviram, and Harlap are giving you an exciting glance at the world of the future automobile, and it is an especially recommended journey.

A cure for the plague of the modern world

"In two decades, it will be illegal to drive, because we drivers are responsible for 93% of the accidents, and road fatalities are a plague - 1.5 million fatalities a year in road accidents and 50 million more injured. I recently calculated the economic damage at $600 billion. It's a plague of the modern world, and we have somehow come to accept it with equanimity. There's a cure that will save 1.5 million lives and 50 million injured, so the autonomous car is a wave that has begun, and cannot be stopped. It won't be stopped," Aviram said.

Aviram adds that the automated car also has far-reaching significance beyond reducing road accidents, saying, "First of all, it will cut down on the number of cars on the road, because we currently spend only 4% of our time in a car. It's the most wasteful sector around. Think of it as buying a smartphone and using it only 4% of the time. It's illogical. The number of cars will drop, some say by half, some say less, because people will no longer own a car. You operate an app. Say I want to go from point A to point B. A car will stop next to me, and collect other people on the way. If you want to go for a weekend outing with the children, you order a minivan, and if you want to go to work some morning, you order a sports car. Discussion is become far cheaper, because if you leave the driver out of the equation, the cost of operating a car goes down by two thirds. But what is concealed in the huge motivation of the technology companies for entering this field is the pursuit of the leisure of the drivers in the car when they are going from place to place. This is the biggest economic motive that will not allow this trend to be stopped, because when you're a prisoner in a closed space, you can be shown commercials and given information. A lot of things can be done that are now done anywhere you're a prisoner. There's a lot of money involved here.

"If you ask me whether an autonomous vehicle will take off, my unequivocal answer is yes, there's no question about it. The technology is almost there, the world is almost there, there's an economic motive for getting there, and drivers will slowly start to get used to the idea that you can get rid of the boring aspect of driving."

"Globes": Amnon, how do you see the autonomous vehicle's progress?

Shashua: "I look 10 years ahead. I'm not a prophet or anything, but 10 years is something that can be described. Beyond that, it's all speculation. You can say for sure that two things will happen by 2021: there will be premium cars operated by an automatic driver on permitted high-speed roads, for example Highway 6 in Israel (Cross-Israel Highway). This car will travel safely without any hands. You can sleep, read a book, and the system will give you some time to reassume control of the car. The system can wake you up, and if you don't take control, it will make the car stop safely on the side of the road. That should happen between 2019 and 2021.

"The other thing that will happen in 2021 is that several players will launch technology for an autonomous car in a virtually delineated urban space designed to carry passengers. These players will include BMW, and Ford. Apple Computers has also announced it, and it is likely that Google will come up with such technology by then. It will happen technologically. It will take another few years before the company and the regulators allow such a vehicle to really travel without a driver. For a few years, there will be a driver behind the steering wheel solely for the purpose of collecting statistics in order to prove that these vehicles are really safe.

"Something else that will happen in about 2023-2024 is the continuation of automatic detailed maps. An autonomous car will have the technological capability to travel from point A to point B without a driver, and here, too, a driver will have to be behind the wheel at the beginning in order to prove that it's safe. Several years after that, if everything works, you'll be able to see an automated car traveling from point A to point B without a driver in both an urban space and an interurban space. It's pretty sure technologically that these things will happen. These plans have already been announced as manufacturing programs, and in the auto industry, once a certain activity is announced as a manufacturing program, the probability that it won't happen is next to zero. So 2021 is shaping up as a very interesting year in the development of autonomous vehicles. Anything beyond that is just speculation."

From accessories to a safety system

At the beginning, Mobileye was not involved in the autonomous vehicle problem. As Aviram tells it, the idea for the company was born following a presentation by Shashua to an auto manufacturer in 1998. "They asked Amnon if it was possible for two cameras to identify a vehicle, and he answered intuitively, 'Why two cameras? I think that scientifically, it can be done with one camera.' At the end of the presentation, we became engrossed with the question of how one camera could identify a vehicle. Later, we started wondering about why a company was interested in it. We were very good friends then, and he said there was probably potential here, so we should develop a company. Within five minutes, we decided that we'd develop a company, but we hadn't done it yet then. At the beginning, we took one engineer at our own expense, built a giant box, and put eight Pentium processors in it. We went back to that customer after almost half a year, and demonstrated an initial concept to him. We got a fairly large additional budget, and so, with this first budget plus an initial customer, six months after that meeting, we decided to develop the company. This story is interesting, because a startup that begins with a customer's needs has a much better chance of succeeding than a startup with technology searching for a market."

How did you go on from there?

"It took us many years to start explaining to the market the capabilities of computer vision systems, because the market at that time for road accident prevention systems was just beginning. The general concept was that radar was the right technology for doing it. A vision system was considered unreliable technology. It took us eight years before we signed the first agreements. We developed the company in 1999, and got our first manufacturing agreements in 2007 with General Motors, Volvo, and BMW."

According to Aviram, the company adopted two decisive special strategies. The first was the decision to work on one product that does everything. "An auto manufacturer won't use two systems, one of which does this, while the other does something else," he explains. "It drove us nearly mad. It was like running three startups simultaneously - recruiting a lot more employees and raising more money. Management attention had to be split among a lot of apps that we ran."

The second strategic decision was to go for a Tier-2 company, meaning that Mobileye provides the technology through Tier-1 auto parts manufacturers, not directly to the auto manufacturer. Aviram says that this decision proved much more critical than they initially thought. "All of our competitors are Tier-1, meaning that they supply a system directly to the auto manufacturer. We said, 'No, we'll make a chip and sell it to auto parts manufacturers like Delphi, Magna, Mentor, and TMW. They'll build a system around our chip and sell it to the auto manufacturer. When you work with Tier-1 companies, you can reach all the auto manufacturers through them. When you're a Tier-1 company yourself, you'll be limited to a small number of auto manufacturers, because such a company usually has a lot of people supporting each manufacturer. Tier-1 companies are really divided geographically: Bosch is very strong in Germany, Delphi is strong in the US, Magna is strong in South Korea, Denso in Japan, and so on.

"When you work through Tier-1 companies, you are actually using the industry to maximize the system's reliability. Every auto manufacturer has to decide whether to use our technology on millions of kilometers of trips in order to test it. This usually involves 10-30 vehicles traveling 24/7 and continuously recording. That's how we've generated 40 million miles of verification data to date. We've created real data, including road variations all over the world. The more scenarios you teach the system, the more reliable it will be.

"That's how we distanced ourselves from the auto manufacturers, and became a Tier-2 company. We managed to come up with a system that no competitor could even approximate. This became especially important since 2014, when the regulators in Europe told the auto manufacturers, 'If you want four or five safety points, you have to have preconditioned emergency braking.' For the first time, our system became an accessory to the system for a driver in preventing an accident, because it brakes by itself. The two strategies I told you about created a situation in which we won every tender we participated in over the past two years. Today, we dominate the advanced driver assistance systems (ADAS) segment. Actually, there are only three important auto manufacturers that don't work with us: Toyota, Mercedes, and Tesla, with which we ourselves decided to break off ties. This is unprecedented. Today, we're considered a leading supplier, although we have competition."

First autonomous vehicle in 2021

"Let's talk about where Mobileye was when it held its IPO two years ago, and where it is today," Shashua says, and tells us how the company fitted in with the autonomous vehicle trend. "The public was exposed to Mobileye only through our after-market, which is only a relatively small part of the company's positioning. When the IPO was held, Mobileye was busy decoding a direct signal from a single forward-looking camera aimed at preventing road accidents by spotting vehicles, pedestrians, road signs, and lanes with the use of computing information - artificial intelligence."

You are referring to a system with a beeper warning about pedestrians.

Shashua: "Yes, but in addition to beeps, the vehicle also brakes automatically, and that really prevents an accident. The beep warns the driver in case he didn't notice - those are built-in systems, which is Mobileye's main business. Mobileye's very high value in the IPO was because this technology is essentially dictated by the regulators. In other words, it is not the driver or the end customer deciding that he wants technology for preventing accidents that determines the penetration rate. The regulator dictates these systems' penetration by setting rules stating that every new vehicle from a certain point onwards must be equipped with them. In August 2014, Mobileye had 90% of the market. We're talking about 90% of a market with 100 million cars a year. That's where the high value came from.

"But that was two years ago. Mobileye has spread its wings over the past two years, and the announcement with BMW in July, two months after the announcement with Delphi of cooperation in manufacturing of an autonomous vehicle, were indications of the success of that wing spreading. Mobileye realized that in order to be a significant player in the autonomous vehicle segment, it's not enough to decode frontal information with a forward facing camera. You have to deal with a range of sensors that provide 360-degree picture around the car. You have to develop artificial intelligence for more advanced sensing than what is needed just to prevent accidents.

"Another thing is that map drawing capabilities must be developed. Today, Mobileye is becoming a mapping company: not the familiar navigational maps, but maps called 'high definition map' in jargon - very detailed maps that place you on a 10-cm map. Once you're located within such a map, you know exactly where all the lanes are and the semantic significance of each lane. The problem with these maps is how to make them automatically, and how to revise them, so that they correctly reflect the situation all the time.

"We developed a way to do this using crowdsourcing. Like Waze, but the vehicle, not the volunteer driver, does the work and sends the information. Based on our artificial intelligence technology, the vehicle understands the road, finds all sorts of reference points, and sends information to the cloud. The weight of this information is about 10 kilobytes per kilometer, and our technology is in the cloud. It takes these pieces of information and builds a map around them. We started doing deals with auto manufacturers in order to attain this goal - it's really an achievable goal in the foreseeable future, even in the near future, for all the auto manufacturers to cooperate and contribute information to making this map, which is a critical asset for an autonomous vehicle."

Ziv adds, "Because so little information is involved, we can transmit it easily to the cloud using the ordinary cellphone systems, process it into an image, and return it to the driver. This is done in cooperation with the auto manufacturers. Actually, we're now signing agreements with all the auto manufacturers. Without it, it's impossible to put an autonomous vehicle on the road."

Teaching the car to drive

The most recent segment that Mobileye has taken up is the insight that the intelligence of driving, the decisions how to fit in with traffic, is a technological problem. "Without a very precise and detailed solution for learning human driving behavior, we won't have automated vehicles that can travel in normal urban areas. They have to travel everywhere - in Boston, Tel Aviv, and Jerusalem - and fit in with the traffic, just like a human driver," Shashua says.

"Assume that four autonomous vehicles simultaneously reach a junction without a traffic light," Aviram says. "If they behave according to the rules, they'll stay there forever. As human beings, we immediately solve this with intuition by looking each other in the eyes, seeing who's driving more and less aggressively, starting to move, seeing what the response is, and solving the problem somehow, without even thinking. Take the Place des Victoires-Nationaux in Paris, for example. The negotiating you have to do there with other vehicles is complicated - even unsolvable. We'll solve it by doing billions of simulations of situations in order to teach the computer how to behave in various situations. We're already achieved very good results, but it takes time to cover all these things. It will take an effort by all the auto manufacturers - the entire industry working together.

"As human beings, we essentially go through the same process. When you learned to drive, after a lesson or two, you already knew how to operate the vehicle. All the remaining lessons were spent on how to fit in on the road. What makes this problem more difficult is the fact that, in the next two decades at least, autonomous cars will travel on the road together with cars driven by human beings, and people's driving is unpredictable, and not always legal. Were all the vehicles to become autonomous at the same time, the problem would be reduced to a tenth of its scale, if not less."

Shashua: "The very advanced technology behind this is called reinforcement learning. Mobileye developed it over the past two years. From this standpoint, BMW's announcement in July was very significant, because what did it really announce? Not that they have a supplier named Mobileye, but that they have a partner named Mobileye. The CEOs of BMW and Intel, and Mobileye, stood on a single platform and told the world that these three companies were partners in building a mass produced completely autonomous vehicle, for which the target date is 2021. It's a very big step forward from being a supplier in August 2014 to being a partner in July 2016. What's behind it all is the technology."

Published by Globes [online], Israel business news - www.globes-online.com - on September 28, 2016

© Copyright of Globes Publisher Itonut (1983) Ltd. 2016

Ziv Aviram and Amnon Shashua Photo: PR
Ziv Aviram and Amnon Shashua Photo: PR
Twitter Facebook Linkedin RSS Newsletters גלובס Israel Business Conference 2018