An interview with Mr. Patrick van der Smagt – A man who loves to see connections through research fields!
At the Second Workshop on Brain, Computation, and Learning (BCL) at Indian Institute of Science, we got a chance to interact with some of the leading minds in the field of Neuroscience and Artificial Intelligence. We interviewed few of them. One such humble person was Mr. Patric van der Smagt, Director of fundamental AI research at Volkswagen group. Currently residing in Munich Area, Germany, he is also a Chairman of the board of directors of Assistenzrobotik eV, a non-for-profit organisation for assistive robotics to help people. Being an enthusiastic professor in the past and head of several divisions, it was quite a fun to have an informal interview with him. Despite being a viewing generation (than reading), I hope you would love to have a read at following!
I am not involved in finding problems, I am only looking at solutions. I know it sounds funny but for me that is the difference between fundamental research and applied research. Like in applied research, somebody has a problem, he/she comes to see you and you say ‘Oh, I need bigger computer, tons of engineers and scientists and I have two years in hand to solve the problem’. In fundamental research, nobody comes to you and you say to yourself that, I have this silly idea and I am going to implement it and see what it does and half of the time nice things come out (and somebody can indeed use that). Advantage of being in a company like Volkswagen compared to a university, finding a problem to research on is always quite easy. Versus in University, you try to find solutions but nobody has a problem for you.
Do you think by looking at current progress of AI that at some point of time in future it will be capable enough to solve the major problems as Global Warming and Economic Bubbles?
Given an infinite amount of time, yes!
Ha Ha Ha. Cool! Lot of people lack the proper understanding of AI, they overestimate the stuff. So, from your experience, how can someone manage the expectations of what’s possible with AI and what’s realistic?
(Chuckles) That’s a very good question. So, most of the people highly overestimate AI. By the way I don’t like the word AI, because that’s the root of the problem. Because whenever you say AI, people think about human intelligence. We are talking about excellent machine learning methodologies which in themselves are very well defined and clearly enclosed in specific domain and can do some tasks even better than humans. For example, image and speech recognition. You could show that if you write a machine learning based program, it can surpass human level performance to solve that specific problem. But that’s the same as saying an aeroplane can fly faster than a bird and a car runs faster than the fastest animal. Because you created a piece of technology that is fit for that specific task and for that specific task only. But that doesn’t mean that you have solved general intelligence. It won’t be there this year and it won’t be there next year. And I am pretty damn sure it won’t be there in next several years. And it is silly to make these kind of predictions because technology always amaze us.
A.I. basically automates stuff. It reduces human efforts. So, which kind of jobs do you think will be most affected at first because such intelligent systems?
One good thing is our jobs will be the last ones!
I think in administration, there is a lot of repetitive kind of work is being done. And we should be happy if we could automate such tasks. And people doing those jobs will be free for some efficient work. From the era of industrial revolution, from last 200 years this has been happening. So what we have observed over all these years is that it doesn’t lead to unemployment but it leads to higher employment. Because the people can get off to more meaningful work than the boring work our brains are not made for. This applies for any repetitive work. For example, copying data from one excel table to the next table. I have seen people doing that. And people make mistakes because our brain can’t do that very well. There will be some societal challenges anyways, because the rate of development is increasing and it will be harder for society to take care of those people who will be losing their jobs. But there’s no alternative, right? You guys are holding your mobile phones right?
(Defending ourselves), Sir, they are just for recording the interview.
(Chuckles) I know. I know. But you know what this thing costs, right? Do you know why are they still affordable?
Because, they are made by people in countries having really poor life and really poor salaries. So, this kind of work which is going on right now, which is also kind of necessary to produce these things is not going to happen in 10 years from now. Because, people will start having life. Their economies are booming like anything. So, what is next? Going to some other cheap labor countries?
At some point of time it will saturate, right?
Exactly. So, none of us will be prepared to pay three times as much for these things. So you have to find different ways to keep up the same kind of efficiency as we have at the moment. Because, we have no alternative. Economy is not local to a single country or single continent but it’s a global thing.
Yeah. And what do you think about independent A.I. startups competing against tech-giants like google, apple, amazon, etc? Should they try in this domain despite having some big shots in the field?
(Pause) Yessss and Noo. When people take up on startups like autonomous driving, I think that is a bit courageous. Because Google, BMW, Volkswagen, Nvidia, Baidu, Lyft and many other well established companies who invest a lot of manpower, infrastructure and money in that because of potential of having a big use case there. I think that having similar approaches on small scale is silly. But on the other hand, these investment of resources makes them inflexible where small companies and startups are good at. Startups can change flexibly according to customer needs. At a big company, you can’t do that.
So, if an area is not too competitive there is no reason why you shouldn’t try to implement your small thing and try to make a mark. I have been involved with various startups in this area. There is always a decision you have to make at the beginning whether you want to build a product and go on full risk but have the benefits of scaling or want to go for consulting and have a busy but comfortable life. That is the decision you have to make.
Thanks for the advice. As startup enthusiasts, we will definitely keep that in mind. Next one is for beginners in AI. Does one need to know about complex maths, calculus, linear algebra if he/she wants to start on this path? And what resources do you suggest them?
The number one thing is Do Your Mathematics. Do linear algebra, calculus, etc. With that you can do anything afterwards. Also, computer programming is not a hard skill to learn. The thing is, these things are important to master. But most people do not even know how to multiply two matrices. And you have to know some basic things in order to work on all this stuff. So, learn these things first. And then learning some machine learning stuff is just a matter of few months, not even years!
Having said that, I am really sad that I don’t see a single person at university, at least the ones I am associated with who does not do machine learning in their Masters’ project. It gets so boring. Get a life to something else.
Laughter again 🙂
Yes, we need machine learning people. But we also need the other ones. So, get your basis and then you can switch. In fact I have been switching my field of research on a very frequent basis. I have worked in bio-mechanics, neuroscience, AI and stuff. The advantage being after some point of time you start seeing connections between all the fields. And that is an amazing feeling when you are working on something and suddenly a thought comes.. ‘Hey, that makes sense!‘
Ammm. I think that is sufficient. Many people know about what you do at work. So, we have some questions about your personnel traits. Do you like science fiction movies?
They don’t impress me actually. I get inspired by more human life based movies.
Which is your favorite book or novel?
That is always the one I just read. I have just finished Paul Auster’s 4321. It came out last year. And oh, I read some metamorphosis by Franz Kafka last week. That is more or less a favorite of mine. Its story is strange having a backdrop of first world war.
The last one, what do you want to build or design in the next 5 years?
There is this American group which looks into societally correct ways of using artificial intelligence. Keeping an eye on usage of these technologies in weapons and other industries would be necessary.
Pro Tip: On an end note, he added that one shouldn’t try to compare different regions or countries of the world. Everyone has their own set of problems. As far as A.I. is concerned, you need have a focus on solving your own societal problems like every country should have. Try to use machine learning and artificial intelligence in that aspect. Society problem doesn’t only mean getting the poverty off the streets, it is much broader than that. That’s like more efficient production, cleaner environment, etc where machine learning or data mining can play a huge role and have huge impact. This way every country can have a significant contribution to humankind!
On behalf of all the students who were interested in the BCL workshop as well as eager to listen what great minds had to say, I thank Patrick Sir for giving us the opportunity to interview him 🙂
Hasta la vista !