Kirill Eremenko: This is episode number 307 with AI engineer Marc Sarfati.
Kirill Eremenko: Welcome to the SuperDataScience podcast. My name is Kirill Eremenko, Data Science Coach and Lifestyle Entrepreneur. Each week we bring you inspiring people and ideas to help you build your successful career in data science. Thanks for being here today and now let’s make the complex simple.
Hadelin: This podcast is brought to you by Bluelife AI. Bluelife AI is a company that empowers businesses to make massive profits by leveraging artificial intelligence at no upfront cost.
Kirill Eremenko: That’s correct. You heard it right. We are so sure about artificial intelligence that we will create a customized AI solution for you and you won’t need to pay unless it actually adds massive value to your business.
Hadelin: So if you’re interested to try out artificial intelligence in your business, go to www.bluelife.ai, fill in the form and we’ll get back to you as quick as possible.
Kirill Eremenko: Once again, that’s www.bluelife.ai and Hadelin and I both look forward to working together with you.
Kirill Eremenko: Welcome back to the SuperDataScience podcast ladies and gentlemen. We’re super excited to have you back here on the show today. Today’s guest is super special. Today we’ve got Marc on the podcast. Marc is an AI engineer who’s working with Bluelife and helping us solve massive challenging projects. It just so happens that today or this week we are together in Switzerland and we decided to use this opportunity to record a podcast and in the process I got to know Marc a bit better and you will get to know him too. Here are a couple of things that we discussed. First of all, we talked about how the thoughts that you choose can affect the way you live. Very interesting, deep conversation there. Then we talked about university education versus online education. Marc completed one of the top schools on machine learning in the world and I think you’ll be interested to hear his opinion on how online education compares to in person university education.
Kirill Eremenko: We also talked about Marc’s systematic approach to solving problems; to solving machine learning challenges and you’ll find some valuable takeaways there. Then we talked about why Marc quit Spotify. Marc was actually building neural networks at Spotify, had an amazing job there and in this podcast, you’ll find out why he gave it up. We also talked about stepping out of your comfort zone, what it means and what kind of manifestations that can have. Those are just five examples of topics that are covered in this podcast. I’m sure you will find plenty more useful insights here. I’m super pumped for you to meet Marc. Without further ado, I bring to you AI engineer Marc Sarfati.
Kirill Eremenko: Welcome back to the SuperDataScience podcast ladies and gentlemen. Super excited to have you back here on the show. Today I’ve got Marc returning from the previous episodes. Marc, how are you doing?
Marc Sarfati: I’m doing great.
Kirill Eremenko: For those who don’t know, Mark is our AI engineer at BlueLife AI and it’s a company where we do artificial intelligence consulting to help businesses make massive profits with no upfront costs. Hadelin just left to the airport, right?
Marc Sarfati: Yes, he just did it.
Kirill Eremenko: So it’s just you and me now?
Marc Sarfati: Mhm.
Kirill Eremenko: What I wanted to talk about today is a critical thing you told me a few days ago.
Marc Sarfati: Yes.
Kirill Eremenko: It’s not about the thing, it’s about where it’s coming from.
Marc Sarfati: Yes, exactly.
Kirill Eremenko: What does that mean? Such an interesting quote.
Marc Sarfati: I’m very interested spiritually and I’ve been in a self-development journey for quite a while and to me it’s very critical to understand the notion of differentiating basically the thing from the place it’s coming from. You can do the main thing in two different paradigms which are completely different. One very simple example is when you are working on something that really excites you. You can spend hours upon hours upon hours working on it without feeling like its work. It’s just fun. Sometimes you have to do something that someone forces you to do and you really don’t want to do it and every single second spent working on it feels like a huge pain. Basically, you’re doing the same thing but the difference between these two examples is the place it’s coming from. It can either come from a place of inspiration; when you have the will and the joy and the happiness to do it or from a place of desperation; which is something you do to escape something.
Kirill Eremenko: Very interesting. Okay. So Tony Robbins would say you’re either coming from a place of fear or pain or a place of pleasure.
Marc Sarfati: Yes, exactly.
Kirill Eremenko: You’re either chasing something that you want or you’re running away from something. It’s like a push-pull. You can either be pushed to do something or it can be like pulling.
Marc Sarfati: Exactly.
Kirill Eremenko: Which one do you think is better?
Marc Sarfati: It’s obviously doing things from inspiration. It feels like you have much more energy to spend. You have much more joy doing the things. It’s a no brainer.
Kirill Eremenko: That’s interesting. Actually, yesterday we had this conversation. We were at this client sites where they were undergoing the digital transformation. We’ve identified six use cases where we can add value and it was very interesting for me to have that conversation. Where you said, “Obviously I can’t work on everything. I want to work on the thing that I get inspired the most.” Right behind us on the wall there, right now, they were all written out there. How do you decide what inspires you?
Marc Sarfati: Usually you just know it. It’s really obvious. One of my main roles in life is to always follow my intuition because I really believe you have a GPS inside you that exactly knows where you want to go or where is best for you to go. I tend to really listen to basically how you feel. If you feel good about something, sometimes you use the words, ‘I have a good feeling doing this.’ To me, this is like a super powerful, guidance system that you can just follow and it will give you ultimately the best feeling and the best results.
Kirill Eremenko: Very strange hearing that from an AI engineer.
Marc Sarfati: Yes, I know.
Kirill Eremenko: How do you combine the two; the logic and the feeling?
Marc Sarfati: It’s hard to say. I come from a very scientific background. I started my spiritual journey, I would say, very recently so it’s hard to combine both. I don’t feel the need to combine both. I like doing data science. I like doing like the math, but I also enjoyed the spiritual aspect of life and I enjoy them separately.
Kirill Eremenko: Do you ever come into situations where the two contradict each other; your logic tells you to do one thing, but your feeling tells you to do it another thing.
Marc Sarfati: Yes, of course.
Kirill Eremenko: Do you follow your feeling?
Marc Sarfati: Yes. I try to always follow my feeling. The thinking mind is very strong so sometimes, no one will see why you want to make a decision and all the evidence show that you should do something, but the feeling tells you to do something else. Even though it’s hard, I try to always follow my good feeling.
Kirill Eremenko: Do you have an example like that? A recent one where all the evidence were just in one thing, but you decided to follow your feeling?
Marc Sarfati: One example is recently I was working at Spotify basically before doing consultancy and machine learning more in a freelance kind of way; and the situation there was great. I was doing a job I really enjoyed. I had a lot of flexibility and freedom working at Spotify. I’ve been passionate about music for a long time. Basically it was for, it was a good salary, so for everyone, this would be the dream job. There came a point where it was obvious that I needed to do something on my own.
Kirill Eremenko: Well, how was it obvious? It was like a feeling?
Marc Sarfati: Yes. It’s almost like a fire burning inside you.
Kirill Eremenko: Interesting. Okay. Has this feeling ever been wrong? Let’s say it this way, have you ever had this feeling but it comes from a place of fear? Sometimes we have feelings that are pushing us to do something or not to do something. Let’s say somebody might have all the evidence suggesting that he should quit his job or her job, but then they have a really bad feeling about it. It might be like an intuition that they need to follow or it might be coming from a place of fear.
Marc Sarfati: That’s a very interesting question. To me, I feel you need to develop a radical honesty with yourself. Ultimately deep down, you know the difference if it comes from a place of fear or inspiration. You just have to be honest enough with yourself to take the information without judging it, which is a very famous concept in spirituality, but basically observing what’s inside you without trying to label it. If you have this clarity, you view yourself within the lens of pure clarity. In my opinion, you will see the difference and you’ll see the answer.
Kirill Eremenko: How do you get that clarity?
Marc Sarfati: From watching yourself all the time; meditation definitely helps. Basically, trying to understand yourself without judgements.
Kirill Eremenko: Interesting. Okay. So you meditate a lot.
Marc Sarfati: I try to meditate. I have to say recently I was not as serious as I was before, but I tried to meditate every day for at least 15 minutes.
Kirill Eremenko: Is morning better or is evening better?
Marc Sarfati: I like in the evening before going to bed.
Kirill Eremenko: I would fall asleep.
Marc Sarfati: It’s a good transition from your active day to going to sleep.
Kirill Eremenko: Interesting. Actually this was supposed to be a FiveMinuteFriday episode; like a five minute one, but let’s just keep going. This is a fun conversation. We can make it into a big podcast. Since we’re on this really cool. So tell us a bit about yourself. Where you mentioned in the previous FiveMinuteFriday that you worked at Ecole Polytechnique, right? Is it a big powerful school on machine learning in Paris?
Marc Sarfati: It’s the best engineering school in France in general. It’s quite a general scientific school but then you have specific tracks inside the school. I focused in data science. The level in mathematics at Polytechnique is super high and data science is becoming more and more developed inside the school.
Kirill Eremenko: Why did you pick that field to study?
Marc Sarfati: My intuition. That’s a domain I really enjoy. I tried several courses in several fields. During the university, I tried economics, I tried biology, mechanics and I studied math and physics for a lot longer before. I really liked computer science in general. I did a lot of algorithmic too and graph theory. Computer science and applied mathematics was what I enjoyed the most.
Kirill Eremenko: Okay. It’s always interesting to talk to somebody who actually studied this at a university because a lot of our students and listeners on this podcast study this themselves online and through courses. What would you say are the main differences between studying data science and machine learning at a university; one of the top universities in France or in the world actually versus studying it online?
Marc Sarfati: I think the main difference is the level of theory compared to practice. I think in university, especially in the one I did, we were very focused about understanding the math behind the models and how they work. Sometimes in many courses, we wouldn’t even code on a computer. You really understand the mathematical principles upon which the machine learning is based. This gives a truly in depth understanding of basically what happens under the hood and how it works. I think it’s not necessarily useful for everyone to understand this because 90% of the cases when you want to do data science, you can use the tools almost as in a plug and play fashion. You need basically to understand how it works, what kind of inputs you can give, what it can understand and apply them as is.
Kirill Eremenko: Okay. What would you say about this theory, now knowing this theory; has it changed your mindset? Is there any benefit apart from the 2% cases where you actually need to change an algorithm or do research in that space or something like that where the theory would come and apply? Has it changed your mindset? Has it maybe made you look at problems in a different way?
Marc Sarfati: Yes. I think it makes you tackle problems in a more systematic approach. First I think if you have a deeper understanding of how the models work, you have a clearer idea and intuition on which models would work and would not. Also, it’s easier to understand when for instance, a deep learning model doesn’t train. Why doesn’t it train? It makes you question whether the architecture you chose is the best fit for the problem.
Kirill Eremenko: Okay.
Marc Sarfati: I think it gives you a bit more clarity why things work and why things don’t work so you can use this as a good signal to explore other possibilities for models.
Kirill Eremenko: Okay. What is your systematic approach to solving problems?
Marc Sarfati: I always usually start with the most basic version of the problem. For instance, I would try to predict whether it’s going to be sunny tomorrow or rainy or cloudy or the precipitation etc., I would start with the most simple use case like just taking my input data and fitting linear regression on the data to predict whether it’s going to be sunny or cloudy. Then I would try to add rainy, cloudy, like different outputs or maybe other inputs that are a bit more complicated to pre-process. Then I would try more complicated models like random forest, gradient boosting or simple deep learning models, etc. I would really start basically with the most simple, usually like a toy example and then I would build upon it with layers.
Kirill Eremenko: So not only would you use a very simple algorithm at the start, but you’d actually simplify the problem itself.
Marc Sarfati: Yes, always.
Kirill Eremenko: Are you always able to do that? The weather example is pretty straight forward. You would simplify weather to sunny and not sunny. In business use cases, are you always able to simplify?
Marc Sarfati: One easy way to simplify is basically you can use one input feature instead of all the features you have. In the features you can have time series, you can have static features, etc. I would keep for instance all the static features that wouldn’t change over time. Just keep these ones and try to first make a very simple naive model of prediction and then add the features throughout exploration and the project.
Kirill Eremenko: Okay. Also increase the complexity of the algorithm.
Marc Sarfati: Yes.
Kirill Eremenko: Along the way.
Marc Sarfati: Yes. Basically you have, I would say, three main components, which are the input feature, the model, and then what you’re trying to predict. I would always start with the most basic I can in all of them and then I would try to improve, for instance, the features. Once I have identified a set of features that seem to be relevant and quite exhaustive, I would improve the model, etc.
Kirill Eremenko: When do you start the feature engineering?
Marc Sarfati: Very soon. It’s quite fast to have a very basic model. Usually the first thing I do is working on the features.
Kirill Eremenko: Okay. After a simple regression, what do you proceed towards next?
Marc Sarfati: It depends on the project but usually, I like using random forest regressor quite quickly especially in Python. You can almost use linear regression and random forest interchangeably with scikit-learn so it’s quite easy to make it run. They give you quite a good estimation of the importance of each feature in your input space. For instance, I would train a random forest and I will see basically, which is the importance, the weight of each feature in the forest, so that you have a clearer understanding of which feature affects the results the most. It gives you, usually, a clear understanding of the dynamics of your problem.
Kirill Eremenko: Okay, and then what’d you do? If the problem is not solved yet or if you need more accuracy or more sophisticated algorithms, which one do you choose from there?
Marc Sarfati: Usually at that point, it’s good to have some business insights because usually the experts in the field know… When you have something predict, they know which are the things that can make good predictions of this output. It’s good to have this discussion now to understand if something is not working at all; try to figure out whether it is because you haven’t used the right features or you haven’t treated them in the right way or if the input data are just too noisy. Basically straight to troubleshoot why it’s not working as you expect.
Kirill Eremenko: Okay.
Marc Sarfati: There’s no general rule.
Kirill Eremenko: Interesting. It’s segues into why we’re here. That’s the reason why we’re here. We were building this model for months and we got to a point where we realize we need more domain knowledge.
Marc Sarfati: Exactly.
Kirill Eremenko: We got on to planes, came here to Switzerland and this is our fifth day here; spending time with the client and getting domain knowledge. How do you feel? Do you feel you’ve gotten new insights from being here?
Marc Sarfati: Yes. A lot. Mostly it gave us a clear idea of how things work, basically the dynamic behind… Of course we cannot disclose on the podcast; the dynamics that basically rule the problem. This is of great help to understanding why the predictions we made were good but not excellent. It gives direction on which area you can work on and improve your model.
Kirill Eremenko: Okay. Very cool. From your experience, because we were walking around and talking to people, but I’m just curious for you what was the best way to get the domain knowledge? Was it like by listening, by asking questions, by maybe reading communications? Do you have any secret or any advice for somebody who’s going to be doing the same thing and looking for this domain knowledge? What’s the best approach to get it as effectively as possible?
Marc Sarfati: Interesting question. If you want to teach a computer to do something, you need to understand how you would do it yourself.
Kirill Eremenko: Yeah.
Marc Sarfati: To me, this is what we were lacking as basically the comprehension of the whole structure and the whole dynamics. To me, you need to understand the problem very well, almost as if you could do it by hand if you had enough computational power in the brain. You need to understand things very clearly so you can teach the computer how to do it. Even if it’s not, if then statements. Even if you just use a deep learning model that basically learns by itself, understanding the problem gives you a lot more keys to understand why your model is failing or why it is working and you definitely have much clearer analysis of the model.
Kirill Eremenko: Okay. What’s the secret? What’s the advice?
Marc Sarfati: Understand the whole scope of the project as much as possible. It’s easier to troubleshoot basically where it comes from. Then you need to explore, of course. If you don’t have a clear understanding of the project; there are many multiple factors that can influence the results and since you don’t really understand them you just put them aside. By understanding them very clearly, you can test every assumption one by one until you find which one is the bottleneck.
Kirill Eremenko: Speaking of putting aside, we had this situation where one of the things that we were working on, we decided to put it on pause simply because by obtaining more domain knowledge, we realize that this is not the best place where we can add value. There’s other places we can add more value; we’ll come back to it later. That’s another form of insight that you can get.
Marc Sarfati: That’s super powerful. Any insights on how to use your time is always super helpful.
Kirill Eremenko: Just knowing what you don’t know is important. Even before you set out to get the domain knowledge, maybe write that out. Do I know what I don’t know or I don’t even know what I don’t know. Interesting, isn’t it? Okay.
Marc Sarfati: One thing I’d like to add that’s very powerful to me is also to be very agnostic in your approach. Start a project without any assumptions apriori. Sometimes you see people; data scientists that will do some prediction model and they will detect an outlier and then they will say, “Oh yes, it’s because of a bug in the measure or it’s a bug here and there.”
Kirill Eremenko: Yes.
Marc Sarfati: Everything happens for a reason. When you actually try to really understand what caused this that you didn’t expect, it gives you a clear understanding of the whole project. Don’t neglect the details because; I don’t know if it’s a saying in English, but it’s in France at least; the devil’s in the details.
Kirill Eremenko: Yes. In English as well. It’s a good point. Were there any instances like that recently for you?
Marc Sarfati: I would not necessarily be able to say it without disclosing more information on the project; which is confidential.
Kirill Eremenko: Okay. All right, now I want to ask you about how you maintain your level of adequate skills. How long have you been out of university now?
Marc Sarfati: Two years.
Kirill Eremenko: Two years. After leaving university for two years, what’s your go to method to make sure you are up to date with the cutting edge technology and you know these recent algorithms, because it sounds like a university is really intense? It really pushed you hard to be up there. It’s very easy to lose ends. This also applies to listeners who are learning through online education, right? You might put in a lot of effort to learn something and get good at it, but then your skills are going to get outdated unless you are maintaining them. What are your ways of keeping up?
Marc Sarfati: I would say practice. Practice consistently. I have a mentor of mine that said to me, “refine, refine, refine and all will be fine.” I really liked the sentence. It’s very powerful. I even noticed when I was on holidays for three weeks; when I came back I would open a Jupiter notebook and I could feel I was not as sharp as I was a month before. Of course it came back very quickly but practice makes things so much easier and automatic. During periods where I code a lot, I can almost start the beginning of the file with my eyes closed.
Kirill Eremenko: If you’re not working on a project, what do you practice with?
Marc Sarfati: I really like what I do so sometimes I just do random projects on my own. Sometimes even at 4:00 AM or recently I was in a plane; I was going to the US with friends of mine and one of them had a position starting in September where he had to learn how to code and I was like, “okay, let’s do some coding in the plane. I’m going to teach you a bit.” My other friend was solving Sudoku on a paper next to us and I was like, “okay, let’s code something that solves Sudoku.” We just spent an hour trying and making an algorithm that solves Sudoku automatically.
Kirill Eremenko: Did you make it?
Marc Sarfati: Yes.
Kirill Eremenko: In an hour?
Marc Sarfati: 45 minutes maybe.
Kirill Eremenko: That’s so fast. You are really fast. This is something that you are quite notorious for. How did you get so fast? Listeners, Marc once.. This is crazy. Marc once did a prototype for a project, not for this client, for another client, a web scraping thing. We were expecting it to be done in a week, it was done by morning. How do you get so fast?
Marc Sarfati: That’s hard to say. It’s hard to say. Of course a lot of practice. To me, coding almost reached an unconscious competence level so basically there is no loss between me having an idea and me implementing them. Basically, if I can think of the thing, I can code it. This is why I can code that fast. I think of the thing I want to do and I think for instance, ‘Okay, I’ll need to sort this array and then do this and that and this and this,’ and then I just do it. There’s no like, “How am I going to do it? Should I do this? Should do that? Let me pull up a tutorial. Of course, I don’t know everything I have to Google specific functions but when I see the problem, when I have the clear plan of what I want to do and I do it. It’s hard to explain.
Kirill Eremenko: That’s very cool. Did you do any touch typing courses or something like that?
Marc Sarfati: No.
Kirill Eremenko: No?
Marc Sarfati: No.
Kirill Eremenko: Okay. Very cool. Very cool. All right. We spoke a bit about that. I see you’re reading The Magic of Thinking Big. Are you doing it?
Marc Sarfati: Yes. It’s a very super interesting.
Kirill Eremenko: I remember it as a book that mostly teaches you how to be a good person.
Marc Sarfati: Yes. I’ll say yes.
Kirill Eremenko: What did is the main take away? So far? Are you just about well over halfway?
Marc Sarfati: Yes. About halfway. I’ve been like learning this kinds of concepts for a while now so the concepts are not brand new for me, he shines light on the details that I haven’t heard before but basically the main thing is the way you think will totally either empower you or disempower you, depending on the thoughts you choose to maintain.
Kirill Eremenko: That’s true. Interesting. It stems back to what do we started with.
Marc Sarfati: Yes, exactly.
Kirill Eremenko: That it’s not about the thing, it’s about where it’s coming from.
Marc Sarfati: Exactly the same situation; you can view it in very different angles that definitely change how you react to it. Even the words you choose. When there is something unexpected that happens, you can say, “Oh, we have a big problem. It’s terrible. It’s the end of the world, etc.” Or you can say, “Oh, we have a situation. It’s interesting that that happens. We’ll figure it out and once we have figured it out, we’ll have a deeper understanding of how it works and we will be able not to make this mistake again in the future. It’s like basically the same situation but so different angles to tackle it. It really changes the way you act.
Kirill Eremenko: I totally agree. I’m reading a book called How Yoga Works, given to me by a very dear friend of mine. I knew this before, that yoga is not actually just about the poses. Yoga, the word, actually translates as union. It’s like union of spirit, mind and body or union of your left and right hemispheres, creative and analytical and all these things. Actually in the book of yoga, there’s less than 10% about poses. This book, How Yoga Works is more of like a novel about a lady that saw a girl that’s walking from Tibet and gets stuck in a police station and teaches the captain there how to do yoga. One of the quotes; and she explains these quotes to them; one of the quotes says, “Things that are not themselves often seem to us as if they are.” It’s full of these quotes.
Marc Sarfati: [crosstalk 00:33:03] Mind boggling quotes.
Kirill Eremenko: If you stop reading for a second, you’re like, “What is that?” I think its purpose is to make you think a bit. Then they explain. It was interesting how they explained it or the girl explained it to this captain at this police station. He had these pens that were out of bamboo. It’s a piece of bamboo which you dip into ink then you can write with it. So she was asking him, “Is this a pen?” He’s like, “Yes.” “Is it a pen on its own?” “Yes.” “By itself, is it a pen?” He said, “Yes, of course. What are you talking about?” Then she looks out the window and there’s a cow there and she gives that pen to the cow and the cow eats the pen. For the captain it was a pen but for a cow it had found something to eat. Similar to the concept you described, this one is that our minds extend the meaning of things.
Marc Sarfati: Yes.
Kirill Eremenko: It’s doesn’t exist out there in the world in the way that we think it exists. The item or even phrase or event might have a completely different purpose. Therefore, it’s so powerful what meaning we give to it. It’s exactly what you said.
Marc Sarfati: Exactly. It’s a philosophical debate. What is truth? If what we perceive is only our perception of the reality, what is reality?
Kirill Eremenko: Yes. Interesting. I heard a recent interesting thought that we tend to equate ourselves to our faces. Like this is me, Kirill, this is Marc. I recognize you. But in reality, we’re actually sitting behind our faces. It’s this wet where three and a half kilograms of bio chemical connections and whatever else that is sitting behind the face. You have these five or six senses. Six because maybe gyroscopic can be counted as a super sense. You have all these senses coming in and you creating this model of the world. So we go all the way back to the matrix and all these things.
Marc Sarfati: Exactly. Yes.
Kirill Eremenko: At the end of the day, it is what it is, right? Cool. Very cool. Why did you quit Spotify?
Marc Sarfati: I always had in mind the idea of working on my own. I really view life as a game and basically I learned a lot working at Spotify. It was a great experience, but then I was like, “Okay, I want to explore new stuff and just play the business game; try new stuff.” Some things will workout, somethings will fail. We’ll see. I like playing and exploring the world.
Kirill Eremenko: So no regrets?
Marc Sarfati: Yes, no regrets.
Kirill Eremenko: Nobody must have understood that. It’s hard to understand.
Marc Sarfati: Yes.
Kirill Eremenko: It’s not just a cushy job, but a great job. Right?
Marc Sarfati: Yes, it was a great job. If you had asked me three years ago, what would be your best job? I think I would say AI for music.
Kirill Eremenko: Because you love music.
Marc Sarfati: I love music. I love AI. It’s like the best of both worlds.
Kirill Eremenko: What you said today at lunch was really cool; that Spotify is full of people who love music.
Marc Sarfati: Yes.
Kirill Eremenko: To give up something like that, you’ve got to have a lot of courage in the face of uncertainty.
Marc Sarfati: Yes. I got that a few times actually when I left Spotify. I also was living in London and I came back to Paris. This happened very quickly, almost from one day to the other. I sent my resignation letter. I moved back to Paris, I had a one month notice and I moved back to France. Many people told me, “You’re very brave and courageous to have left everything so quickly and coming back to Paris.” To me it didn’t feel like something extraordinary at all. It was just a next logical step. I had something in mind. I was like, “Okay, now it’s time to do it.” I had really this sensation. Basically, when people told me, “Oh, it must be a strange to come back, etc. It must be difficult.” I was like, “No, my two hands are still here and my two legs are still here, my body; my mind is still here and it’s all fine. I’m still here.”
Kirill Eremenko: I love that. My two hands, my two legs, everything basically.
Marc Sarfati: Things around me changed, but I was here.
Kirill Eremenko: Makes sense. That’s, very cool. I read a quote recently that ‘life begins at the end of your comfort zone.’ Right? Even though maybe in your case, Spotify was when you’re a [inaudible 00:38:19] ahead was a, jump, a leap forward. Such an exciting thing you doing new projects and so on but within the two years, especially at the speed at which you learn and code, you probably got to a level where now it became part of your comfort zone and to stay there would make you stay within your comfort zone. Interesting. I had a similar experience when I was leaving Deloitte. I did two years at Deloitte and then I went to Sunsuper, which is a pension fund in Australia; like an industry type of job. I only did 11 months. I didn’t even wait for 12 because I felt that’s it. Comfort zone has expanded and as you say, I would rather experiment and fail and learn and do it again rather than just stay within my comfort zone. Different people have different levels of tolerance to uncertainty.
Marc Sarfati: Yes.
Kirill Eremenko: What would you say to those who want to make a leap but feel some sort of hesitation?
Marc Sarfati: I would say it’s very normal to have hesitation. I always do, but I always try to basically take the first action that towards the end goal. If you feel like you want to… and sometimes this can be very extreme; if you feel you want to move to another city or something, just send your landlord a letter that you’re going to quit the apartments in three months. That way, once you start to have this ball rolling and have the momentum, you’ll have to figure out what to do. If you send, basically, to your landlord saying, “Okay. You can stop my contract or my lease now.” You’ll have to figure out another solution, right?
Kirill Eremenko: Yes. As a radical commitment.
Marc Sarfati: There was a training I wanted to go that happened in the US right after my masters and I was hesitating a lot going there; whether I should go there, whether I should not go there. Of course I had a lot of doubts and I was like, “Okay, I’m just going to pay for the training then I will figure out all the rest later.” But I know I will do it eventually because I’ll have to figure it out and I paid.
Kirill Eremenko: And you went?
Marc Sarfati: Yes. I did.
Kirill Eremenko: Was it worth it?
Marc Sarfati: Definitely worth it.
Kirill Eremenko: Well, that’s a cool story.
Marc Sarfati: Also, the more you go out of your comfort zone, the easier it is. For instance, it can start with taking cold showers in the morning. I know you do this every day. You talked about this. I did this for a few months last year. Well, even just like when you go to work, just use a different way. You maybe cycle to work if you’re used to taking the bus or a walk in the different streets or do different things and this will give you more diversity in your thoughts. Basically, it extends your creativity and your thought patterns and it allows you to think more widely I would say.
Kirill Eremenko: Interesting. Stepping out of your comfort zone doesn’t necessarily always mean to be more ambitious, fast, strong, brave, doing unexpected things that people will be surprised at. Sometimes stepping outside of your comfort zone and I’m just going to use myself; actually means the opposite. It’s becoming more humble, more caring, more soft with people. Something you don’t do before. That’s one thing I definitely need to step out of my comfort zone and work on. It’s like developing closer, deeper connections and relationships with people because otherwise I find myself rushing around the world and doing lots of things. I forget that you can connect with people on a deeper level. They are people in my life that I connect with, but not everybody and not that you have to connect with everyone. There are certain relationships you can develop and I’m not used to that, but that’s also an example of stepping outside our comfort zone.
Marc Sarfati: Yes, very similar. Sometimes you’re afraid of saying to people you appreciate that you appreciate them. It can feel a bit scary; or saying, ‘Thank you.’ To me, it’s going to be out of your comfort zone, but super positive and not as you mentioned, super ambitious, adventurous; just allowing yourself to open up a bit more. It’s a great skill.
Kirill Eremenko: Great skill. There are a lot of examples. Whatever makes you uncomfortable basically is maybe something you could look into to try and step out of your comfort zone. Interesting how in life you can develop lots of different things; machine learning skills in one hand, self-development on the other. What else are you into? Are you into sports or anything like that?
Marc Sarfati: Not so much anymore. I used to play tennis and basketball for a while [inaudible 00:44:09].
Kirill Eremenko: But you play guitar, right?
Marc Sarfati: Yes, I do play the guitar. That’s very cool. Music is great. It’s a great passion too.
Kirill Eremenko: Cool. What’s next for you? You getting in a plane in a few hours?
Marc Sarfati: Yes. I’m flying back to Paris and have the dinner with my family tonight.
Kirill Eremenko: Nice. Very nice. Very good weekend. And then maybe looking into some different projects we picked up here. Gotcha. Cool. Before we wrap up, what would your one piece of advice be to those who are entering the field of machine learning, deep learning? What’s one big… If you could give yourself three years ago, one piece of advice, would it be?
Marc Sarfati: Enjoy the process. Enjoy, have fun. Learn new stuff, enjoy learning it and if you have nice ideas of things you want to implement, like toys you want to make with AI, try this. This is, for me, the best way to learn is having a project you have in mind which inspires you and then work on it, work on it until it works. It’s a great source of motivation and learning and…
Kirill Eremenko: Fantastic. Fantastic. Well thanks a lot Marc for coming.
Marc Sarfati: Thank you for the invitation.
Kirill Eremenko: Awesome. Have a safe flight today.
Marc Sarfati: Thank you.
Kirill Eremenko: Thank you ladies and gentlemen for being here today on the SuperDataScience podcast. Thank you for joining us today for our conversation with Marc. I hope you enjoyed the valuable insights that Marc was sharing and also our conversations on things like our thoughts. The thoughts you choose can affect the way you live. I found those very, very valuable. As usually you can get the show notes at www.SuperDataScience.com/307, that’s SuperDataScience.com/307. We will link to all the materials mentioned on the show notes and of course you can connect with Marc there as well.
Kirill Eremenko: If this episode sounded inspiring to you and your business, your enterprise wants to work with Bluelife, you can always find us at www.bluelife.ai. On that note, thank you so much for being here once again and I look forward to seeing you back here next time. Until then, happy analyzing.