US-based GPU-designer NVIDIA is betting big on the automotive segment by providing high-end graphics for product design, infotainment and autonomous driving tech. Every automaker in the world uses NVIDIA’s Graphics Processing Unit (GPU) technology. Jaap Zuiderveld, vice-president, EMEAI at NVIDIA explains how Deep Learning will be the key to autonomous driving and how far is India from realising the autonomous mobility dream. An interview by Takshak Dawda.
What exactly is ‘Deep Learning’ and its connect with artificial intelligence?
Deep Learning is a form to work through an enormous amount of data and get some sensible information out of it to enable individuals or companies to make better decisions. That data maybe structured or unstructured but it is the sheer amount of data that is doubling in size every year. Hence, it has become almost impossible for an individual or a group of individuals to analyse the data.
Deep Learning is a form of analysing the data and obtaining statistics from it, enabling better, faster and easier decisions. Hence, it is within the sphere of Artificial Intelligence (AI) which is not only about robotics. If you simplify it, AI counts for every individual, in every industry.
Where does Deep Learning figure in autonomous driving?
You’ve got to approach it in two ways. One is sensing. You need to know what’s happening around you in order to make decisions while driving. That is what a human being does and the reason the industry is trying to implicate it is because of the deaths related to transportation. Road accidents are the third biggest cause of deaths worldwide after heart attacks and cancer. That’s how it started.
NVIDIA’s Drive PX platform is a supercomputer, the size of a lunchbox that delivers massive computing power required by the AI systems to navigate autonomous cars. It is used by all Tesla Motors' cars.
When I talk about analysing the data and knowing what to do, here’s the catch. The thing is data will never be the same because every road situation is different. Even if you wish to program every road situation, you will still never be ready. So if you think what model can help achieve this, it is Deep Learning. It teaches to make decisions in an environment based on certain parameters, which you input into a model. It then learns on its own what it needs to do and what decisions it needs to make.
Secondly, this is where the Graphics Procession Unit (GPU) comes into its power. You need to compute the enormous amount of data at low energy because you simply don’t have the amount of energy like a data centre. So the GPU is fit for that and makes it a pretty well-positioned product for autonomous driving.
How important is autonomous mobility for NVIDIA?
Today we are working with around 80 partners across the world. That includes Tier 1 OEMs like Bosch, Harman and Continental. We are also a part of the Open Automotive Alliance (OAA) which is basically for everyone who has transportation solutions. I wouldn’t call it just a car solution, it’s a transportation solution. It also depends on the region as every continent will have a different solution.
If you take India, for example, there’s no way you can implement full autonomous driving right now due to several reasons. However, you can take some first steps. Functions like autonomous stop and drive can be implemented. In India, you still need to control the driver because it’s a step too far from autonomous driving.
How much does automotive contribute to NVIDIA’s overall numbers?
A pretty sizeable piece of revenue will come from the automotive sector this year because of the existing businesses. If you look at autonomous driving, there’s only one carmaker, Tesla Motors, which has production cars in the USA. So it is still 3-4 years before this becomes a huge number. Currently, we are engaged with all the main OEMs but it’s at a very early stage. In the coming two years, we will see a lot of announcements and things happening.
In the third quarter ended October 30, 2016, the company reported a revenue of US$ 2 billion, up 54 percent from US$ 1.30 billion a year earlier, and up 40 percent from US$ 1.43 billion in the previous quarter.
What plans does NVIDIA have for the Indian automotive space?
We have a technology which everybody can choose. I think it’s for a car company to step up and say, ‘There’s an opportunity in India to do something’. Only then, we can do something and work on it. I don’t think if you put a Tesla on the Indian road, it will work. There are many reasons. For one, there are no charging points on the roads. Secondly, the car is extremely expensive and there will only be a handful of the population able to buy it. Thirdly, the infrastructure won’t allow us to drive autonomous. So we need to figure out where is India going and what are the challenges.
The challenge in a city like Mumbai is that it takes you upto three hours from one place to another. How do you solve that challenge? You can build an autonomous car, which is not cheap. But what do you make to improve the condition of a market? I think there has to be a very specific solution for this market. For this, we are willing to sit and share our knowledge and ideas with everybody. But at the end of the day, it’s for a party to step up and say, ‘Hey we believe and want to use the technology NVIDIA has. It will help us differentiate in a market like India and be successful.’
Has any Indian OEM stepped up yet? Are they interested in this?
Every OEM is interested in autonomous driving. If you look at autonomous mobility, there are five stages. But there’s also ADAS or Advanced Driver Assistance Systems. That was initiated to make cars safer with features like pedestrian detection and avoid collisions. For example, from 2018, every new car sold in Germany needs to have a pedestrian detection system. So governments are putting rules in place that compel the car industry to use such features. It’s the same what happened with airbags or for that matter with seatbelts. Some years ago, seatbelts were non-existent. Now, barring a few countries, you cannot imagine a car without seatbelts. Take into account ABS. I remember in 1992, having an anti-lock braking system in your car was extremely expensive. Now, almost 80 percent of the cars produced worldwide have ABS.
This is how you must look at autonomous mobility as well. A couple of years from now, there will be some new technology that will go mainstream.
The NVIDIA Quadro visual computing platform is used to design cars by almost all Indian OEMs.
I believe autonomous driving may also change the car ownership structure in the future. I think autonomous mobility will have a huge impact on society. I think this will change a lot of things. You can do your work, pay bills, have meals and do everything you can while you are driving.
Why would you want to own a car and think about parking when you can simply call for one through your phone and travel to your destination? The ownership structure will change completely. Of course, there will always be people who like to have something for themselves.
Look at Uber, which is a very easy method of transportation. Of course, you need to pay for it. But in the future, the society will look completely different due to autonomous driving. You will have more time in your hands, fewer accidents, fluid roads, and faster transportation.
10 years down the line, which will be the biggest autonomous market in the world?
That’s a very interesting question. It depends on how quickly governments, car companies and insurance companies are able to come together and make it work. Of course, where infrastructure is better, it will be easier to implement. However, that won’t be the only factor.
Think about this – when mobile phones were introduced, the adoption in the developed and richer was really slow. Because you had fixed lines, which everybody was used to and the call rates were extremely expensive as the mobile companies wanted to make a lot of money. However, some countries actually skipped the fixed (landline) technology and straightaway adopted the mobile technology, like in a millisecond.
The same applies to this technology as well. You cannot underestimate any particular country. Somebody who is much behind today might pick up much quicker. They are not worried about government norms and things that insurance companies are afraid of. They have less of a worry adapting to changes. I believe there are some countries which will surprise everybody and move into new technologies including autonomous mobility.
It’s now a matter of when rather than how.
Everybody is talking about it. It will happen. That’s for sure. We can’t say when but it has happened quicker than we expected.
A lot of start-ups these days are into AI. Is NVIDIA looking to collaborate with them?
We have a global program to engage with new startups. We support and we help them. There are a huge number of startups doing a lot of great work.
Two years ago, we did not speak about autonomous driving. There was only ADAS. Even now, there’s no company which is super successful and having big revenues from autonomous driving. Moreover, there are so many people who are jumping to the opportunity, which is huge. People talk about autonomous, but at the end, it’s all about giving solutions to your consumers. You have to look at the needs of the driver or the end customer.
I think autonomous driving will make the automotive industry go completely upside down. Firstly, we are moving to electric cars, cars running on alternative fuel forms. That’s big. So, mobility will surely be disrupted.
What is NVIDIA’s focus on the automotive sector and how important is it?
Very important. Actually, we don’t weigh it that way. It’s more about how relevant we are for the industry and I believe we are relevant. I think deep learning or AI will be the next big thing in the industry. What you see today, you won’t see the same 10 years from now. There will be new solutions which never existed before. It will change economies, jobs, society and I believe it’s a very exciting time. It might be challenging, yet it’s very exciting.
All Indian car manufacturers use NVIDIA’s graphic cards. The NVIDIA Quadro visual computing platform is used to design cars. We have the dominant market share in the designing field in India. What the Indian OEMs grapple with is keeping abreast with new technologies. However, aggressive companies like Tata Motors and Mahindra& Mahindra are globalising. They are going abroad and networking, exporting and competing on the quality level. To do that, you have to be alongside or ahead of the current technological curve. Mahindra& Mahindra has long ago stopped talking about a company that makes trucks and tractors. It now focuses mainly on future mobility.
NVIDIA's ADAS tech can differentiate vehicles, parked cars and also identify cyclists to pedestrians. From 2018, every new car sold in Germany needs to have a pedestrian detection system.
Our technology has simply been too expensive to give it enough space for the Indian market. Our digital cluster, which goes in the Audi A8, A6 and A4, is really expensive. If you put that in an Indian car, I don’t think anybody would buy that car anymore.
Similarly, if you move to autonomous driving, it’s not going to be cheap. In India as well, it will initially be a premium product. It will start with self-parking functions not straightaway to level 5 autonomous. It will initially be only for premium brands before high-volume cars can have the technology.
Will India continue to remain an unexplored island in terms of autonomous mobility?
Consider 1.3 billion people living in a relatively small space. At times, people forget to tend to the needs of the larger section of the society or the masses. The recent demonetisation move by the government is huge. It’s a bold step but let me tell this, the amount of data available – which nobody is doing anything with – is enormous! You are talking about 1.3 billion people. It’s a very complex thing to analyse that data and to do something with it. I think AI is a perfect tool which can help this country take some huge steps to get the full empowerment of what it can do.
The mistake what a lot of people make, even in India, is that they always look at the USA. They copy what works in the US and think the same thing will work here. It might work for 20 million people. Fine! But they forgot the rest of the population. If you copy what’s happening in the US or in Europe in terms of autonomous mobility, forget it!
I would look at it differently. I would rather take a step back and look at ADAS. People must look at technologies which will help the current market challenges and then build on it. Moreover, I think it’s a combination of governments really willing to make it work.
Mobility has been a huge challenge for India. Technological solutions should be specific according to the company and not be copied from other regions.
- Baidu autonomous car to use NVIDIA’s AI computer
- A sneak peek into the Car of the Future
- How machine learning will be the key to an autonomous future