Learn Practical Data Science, Programming, and Machine Learning. 25% Off for a Limited Time.
Join our Data Science Bootcamp
Join our Data Science Bootcamp
Listen on your favourite podcast app
In our very first episode, we had the pleasure of chatting with Luis Serrano—one of the top voices in the AI space. Luis Serrano is a technology and science popularizer, researcher, and practitioner along with writing the best-selling book, Grokking Machine Learning. He has had a remarkable career, working at several tech companies including Cohere, Apple and Google. He’s also the brains behind popular ML courses on platforms like Coursera and Udacity, and the popular YouTube channel Serrano Academy, with over 135K subscribers.
In this episode, we unpack Luis’s fascinating journey, from his childhood and maths fears to a deep-seated passion for it and all things related, including AI and ML. We explore his career path in detail uncovering the pivotal moments and learnings, as he navigated through big tech players, changing gears from Maths to AI and Quantum AI, and how he ultimately found his true calling.
We further venture into the world of AI, exploring its profound impact on education and society—both the positive advancements and the challenges it presents, and how they are reshaping the world and future. And of course, we touch upon the human side of it all—exploring the themes of humanity and empathy and implications for the future. We wrap up the podcast ends with a fun and engaging rapid-fire round, again packed with bite-sized learning. So tune in, learn and get inspired!
Next Recommended Podcast: CEO of LlamaIndex, Jerry Liu on Generative AI, LLMs, LlamaIndex, RAG, Fine-tuning, Entrepreneurship
(AI-generated)
Raja:
Hi everyone. Welcome to the first episode of Future of Data and A I. I’m your host Raja Iqbal. So today it is a privilege for me to host Luis Serrano as our very first guest. Luis Is one of the best educators out there who has popularized identity media enlightening black models. I am personally a big fan of his tutorials on Transformers, Attention Mechanism and Embeddings and some of the more tricky mathematical concepts.
I like learning math by just visualizing it and looking at it as opposed to, you know, just sifting through the equations. And Luis actually makes it very easy for me personally. I’m an educator. I teach this topic, and I also learn from Luis. So I look up to him. So, Luis, it is a pleasure and a privilege to have you.
Luis:
Thank you so much, Raja, for inviting me. It’s mutual. I’m a huge fan of your work of Data Science Dojo, the way you all teach machine learning, data science and bring it to everybody in such a comprehensive way. I really admire. And so. So I’m very happy to be here. Thank you for having me. Yeah.
Raja:
So let’s get started. So I was looking at your LinkedIn profile and then trying to see where you’re from and all of that. So you’re based in Toronto now. You work for Cohere at the moment. But were you originally from there? Where were you born? What was what was it like growing up as a child? Yeah.
Luis:
Yeah. So I live in Toronto and I’ve lived in Canada, the U.S. or for a while bouncing back and forth. But yes, I am originally from Columbia. And yeah, I grew up on only child. My mom raised me by herself and she made a huge effort to give me all the education. That was her thing for me and just stretching every penny to give me the best education.
I’m always going to be very, very thankful for that. And that sort of gave me, that planted the seed for like me always being a learner. And that’s pretty much what I’ve been what I’ve been doing. Yeah.
Raja:
Okay. Yeah. I actually I owe a lot to my, both my parents, but most especially my mom. And someone was asking me in one of the panel discussions, so I, I told them my mom did a very good prompt engineering for me. Right. So, you know so she put that I guardrails and all of that. Right. So and she was an disciplinarian and I owe a lot to my mom, actually. Yeah. So I mean your mother, the same way, was she a strict, very academic? I mean, how was it like growing up?
Luis:
Yeah, she was definitely strict. I had two very strong women raising it because it was my grandma and my mom and they had very different personalities. My grandma was very sweet and soft. My mom was like a strong, you know, strict person. And so it’s interesting that I had sort of the to the two sides, but that the common denominator was strength.
Right. Like it was not an easy time for women never is but it was much harder back then and they had to develop this this resilience. And so, you know, growing up, I saw it. I see it much later. You know, you analyze it back and you and you see it much more. But yeah, definitely, definitely. She was strict.
Raja:
Yeah. So are you grateful to her for not letting your slack off? Right. So, you know, because when I was growing up, I mean, I had kids in the neighborhood and they would just go and, you know, maybe not as disciplined as my mom would make me. Right. So and then at that time, it was painful.
But I’d say, hey, why? Why do all the other get to play? And I don’t get to play. I mean, I used to play, but, you know, it’s almost like uncontrolled, you know, fun, right? So and that was not okay. So, I mean, how was it for you?
Luis:
Yeah. I mean, I was always kind of the shy kid growing up. Very shy. So I wasn’t that much into playing it, but I did. But I was kind of I was kind of nerdy all my life, like being inside and playing with, you know, Legos and all stuff. Like, I was always that I always had that side, so I was not that much trouble.
I said, like going outside and doing things. it was also like, you know, a city that was not so super safe and in a time that was not safe. So going outside was a challenge. They would always be like, now kind of stay that close where I can see. But yeah, I was, I was not that much told that in that sense. But I did have I was very I was a distracted kid.
Like, you know, my my mind was everywhere. So like, you know, paying attention, stuff like that, that was always like getting me to sit down and do homework was was very, very challenging. So I don’t know how she did it.
Raja:
Yeah, that’s that’s wonderful. So I was looking at, you know, I learned that you were part of the team that represented Columbia. So you represented Colombia as a high school student in and actually won a bronze medal for Columbia in the Math Olympiad. Yeah. How was it like? Just tell us, how’d it go? I mean, did you prepare a lot or how did it all start?
Luis:
The story, how it started. It’s funny. Yeah, I this was the that I mean, the highlights. If my life had a reel of highlights, that’s the first one. Yeah. Being able to go to this, this competition was just an amazing experience. I mean, people from all over the world and just tell me that there’s to solve problems and that it really was and, and being able to travel at that age with with your friends.
I mean that is to me that that the best experience I’ve had. And obviously yeah, you travel for like a week but you train skate like being an athlete, like you travel for a little bit, but you’re studying the whole time.
It started really funny because I used to fail mathematics in high school. I used to I used to hate math. And as a matter of fact, it was a great when I was the only one that failed in my class. We were in grade 8 and I was the only one who failed Mathematics.
I was failing that year, and so I hated mathematics like because it was just formulas and inequalities and it was like a following instructions, right? Like they tell you what to do and then they quiz you on how to do it. So it was like, but I didn’t know I hated it because of that. I thought I hated it because I hated mathematics.
Raja:
This is this is what I was going to ask you. Did you hate math or did you hate the way it was taught or you did not know what you hated?
Luis:
Now I know that I hated the way it was thought, but I thought I hated math. And so, so many people. And they said, I hate math. I’m so bad in math. I’m like, No, no, no, you hate technicality. You hate abstraction, but you don’t hate mathematics everywhere. Math is just the way we end.
Raja:
It is beautiful, right? It is just the most amazing. right?
Luis:
Yeah. It’s like music. Great. Like you listen to music and it’s beautiful. But if I just show you a bunch of articles and sheet music, then you just be like, wow, what does it mean? If that was what they showed us first, we would think it’s awful, but it’s beautiful. But in math, they show you the formulas first and not the thinking.
So I hated mathematics. And then we there was the national exam. I was like, Yes, in high school that it would always do the national math exam. They would take time off. And so I thought if I skipped hours of class, that’s wonderful, right? Because I was an awful student in every subject, really. I just can’t concentrate.
I don’t like I don’t have that, like, I have ADHD. I just I just cannot focus on anything. So I was an awful student, but at the time, you know, you don’t think, there’s a difficulty in learning. You just think a kid is stupid. That’s how we talk back in time, right? So anyway, I just thought of me as a person who is just not intelligent.
And then I took this exam, the national math exam, because I would skip hours of class and I would just be like, Well, I’ll do doodle for hours. And then I go into the exam and I think give the math exam, which is awful for me. But there’s no formulas.
There’s very little formulas. It was all puzzles, it was all games. And I love puzzles. You know, I was I was always nerdy, but not a good student. Right. Like I loved I don’t know. There were, like, magazines with the puzzles that I would used to do, and I would play games with puzzles and little like, I will always love that. But it didn’t occur to me that that could be math.
But in this exam it was just puzzles. And so I enjoyed it. I just like, well, you know, if I’m here for hours, I might as well. And so I really enjoyed the exam, but I didn’t think much of it. So I just left and, you know, and then one day I’m walking and then the math teacher comes in and said, hello, okay, congratulations, Luis.
And I’m like, And I thought to myself that she congratulated me because I finally passed an exam like it is being sarcastic. And I was very nice. It was like you did very well in the Olympics. And I was like, really? How did I do on? He said, Well, you were first. And I thought, my God, I’ve never been first in school at anything.
She said, No, you were first in the country. And I was like, in the country, How am I going to be first in the country and the stupidest one in the class, right? Like and so turns out I was the first in the country, and then I started liking maps.
But because I started seeing it differently, you know, and I did get lucky with some teachers, like this teacher, for example, was was actually one that displayed the very well when he started looking at what he did it, I started liking it.
But here the highlight that the price for winning the map link that you a top something like or something then the price is price is that in the summer in the vacation they send you to math camp so it doesn’t sound like a price.
Raja:
Yeah, that’s what I was to ask something. Did you see it? Do you see that as a punishment at that time? Right. So yeah, I mean, yeah, some you do a math camp now.
Luis:
So I went to math camp and it was, it was a lot of things because first of all, well, mathematically like first socially was amazing because in the school, like in the house, I was bullied, you know, I didn’t like it. And when I went to math camp, I was with people like me, you know, I just so socially was a huge thing.
And when you’re like that age, like social is everything, right? So to me that was already even if I hated the math, that was already the thing. And but here’s the main thing, right? Like, I walk into the first day of math camp and it was it was difficult like the challenging as hell and but they would never do what they do in school, which is they teach you the method and then make you apply it right.
They call it they don’t they don’t quiz you in that they start with a problem. So the first day we got there and the teacher just put the problem on the board and just know. And I’m like, Am I innocent?
That you didn’t tell me how to solve it? It’s like, Well, that’s the point you figured out. And then they would give us, you know, it’s like, well, I tried this and they would help and somebody would pass to the board and solve it and stuff.
And to me, that was that made a difference in in how I see education. Right? Because it’s not teaching us to be followers, but to figure things out. When I do that, when I teach, I never give quizzes after. I always quizzes before. I don’t go, How do you do this after I told you not before I tell you how would you do this?
And you get the most amazing ideas like you get, I get like ideas that sometimes better than what the standard thing was. So. So yeah. So that opened my mind. And then I was obsessed. And then I just continued. I skip a lot of school for going to the Olympics. I loved it. I would fail in school.
It was matter. I’m never there. Who cares? I barely pass school. But but yeah, the Olympics were where just my life and I would go there. There were two trainings a year in the summer vacation. In the winter vacation in Scotland. Yes, it was no, no seasons, but it’s the same. But it was in December and June.
And if you kind of keep going up and if you do well in the high levels, which are that was my last two years of school, then you get into the team of six people that go represent in the I’m going the math Olympia and everybody’s just obsessed to go to that They like everybody wants to go.
You worked so hard to get there and when you get there, then the training is superintend stuff like train. Like we had a camper who was like trained at midnight, you know, the whole day. It was you were just like an athlete, like your focus. And somehow we could focus on that. I could not focus on anything in school and classes, but but I would hyper focused when it was about solving problems and yarning.
We it well I mean some countries do amazing you know China and India like Russia they were just amazing and we were like, you know, so-so like we did okay. But but it was the most amazing especially shaped my way of thinking, have shaped the way I make decisions. It shaped absolutely everything. And I still in touch with Mary Falk, is the person who runs that all in good.
And I’m still your visitor all the time. Such an amazing human that has made that so much for me that I’m so thankful to you and gets.
Raja:
And when you won the bronze medal. So I’m sure that Russia was there. Eastern Europe, I mean, some countries. Yeah. And they were there. Singapore, China. So South Asia, I mean, so these countries, they may have been there. So. So how was it like I mean winning the bronze medal. Was it for the first time at Columbia?
Luis:
Well, no, I got a I got to clarify something. There’s more medals, right? Like, if not first, second and third, like I was in third in the world. That third in the world was, you know, someone that’s not China’s only they they there’s a bunch of golds and a bunch of silvers and a bunch of bronzes. And so Colombia did get did get medals.
We got some bronzes sometimes we got some I think one did back in the day. China normally gets six golds or high gold They and that that the top countries do very well. But yeah, to get a medal you have to be ranked high and I was very, very happy. Yeah I think I think a lot of your math but a lot of it is your self-esteem.
And I think I, I think I had the math, you know, like it was I was very well trained. If I went back, I think I would go with more self esteem because when you come in and you know that you’re not going to country doesn’t do normally so well, it’s kind of like in the Olympics, right? Like if you’re competing in against somebody who’s just every time they win gold, you get intimidated.
And I think that I think that that helped. Like that week I went in intimidated. I think I yeah, yeah. I think we had great, great training. And in retrospect, I think I think, you know, I would have had more confidence, would have been better. But and I think most of the years I spent in the Olympics more than developing my math was developing like confidence, you know.
And I think that happened to a lot of I was like, we talk about it when I see with my friends with who went in. The same teams are like, Yeah, and I think I think we had a pretty good level and I think we should have been more confident, you know, But it did development, confidence a lot like it just kept, you know, did elaborate.
Yeah.
Raja:
Yeah, that’s great. And I’m assuming that I’m going from someone who hated math to someone who, as for the rest of his life and being passionate about math, do you think winning that within country competition or Olympian, was that the turning point for you.
Luis:
It was the turning point. Yeah, yeah, I think that that helped me throughout all my life. You know, I’ve always had difficulties, like I’m always going to have them, like in many things. It never was like easy after that, but it did give me a good start and then it made me take the bad decision of taking, which is to study mathematics.
You know, at the time in I think it was the time like the late nineties and I mean in, in Colombia, like a thing that the mentality was like, you have to do something that gives you money. And so everybody was like, Well, if you know math, you should. I mean, you know, you should study economics, you should study something like business that gives you money.
And I’m like, I just thinking, like, if I’m awful at school, I’m going to be awful at these things, right? I can only do math is the only thing I can do is the only like literally I have to figure out a career based on math, which at the time you will not is not obvious right now. You can think of data science.
So you can, you can study math and do well. But at the time we didn’t exist, so it was not. But I couldn’t do any other thing. And so it just it was just a decision I took. And then I, I never knew what I was going to do in my life. I just knew that I enjoyed it.
So, you know, I’m just going to one to grand In grad school, the decision was always easy because I just wanted to do a little more math or a little longer, and it turned out to be a great decision. But at the time building, I.
Raja:
And you mentioned your grad school like going to a masters and a Ph.D. in math, So did you always have this in mind that you’re going to be a mathematician or you thought that, Yeah, I will learn it and I’ll figure it out. Maybe, you know. So what was going on or were you just smelling the roses along.
Luis:
Really. I was smelling the roses. I always I only made the decisions on the next step when I went math. I just knew it was the thing I like. That’s it. And I thought, I’ll figure something out on the way.
When I was finishing math, I started understanding what the academic careers and so the academic career and you know, this is dear I’d like the academic career is is you know.
You will you do a master’s sometime semester sometimes some places you go straight to Ph.D., but the fact that you do a Ph.D. and then the postdoc, and then you become a professor, and then you do math for the rest of your life and that’s what all my friends were doing in math.
There are very few who would who would switch to other places, but the majority was just like being a professor. So then I thought, okay, I’m going to be a math professor. And so I took the next step. That was the Masters and the next step that was a piece.
And the next step there was a postdoc. I never questioned, should I do this? I was more concerned of where can I go? Like, how? What can I do to get into the best possible place or what can I do? Like that was the only concern where where do I go for the next step? But it was never should I do the next step or not because it was just, you know, the week and.
Raja:
Yeah, yeah. And then when you finish your Ph.D., your first job was at Google as a machine learning engineer.
Luis:
Yes,
Raja:
And you were working on personalization then with your recommendations.
Luis:
Yeah, right. Yeah.
Raja:
So it was great. And so it so looks like the teaching was not that top priority at the moment. What was going on in your mind? Because most Google everything is teaching most of it. It’s reaching, right? So but prior to Google, right to between your body and now. Right. So there was this brief period what was going on ahead?
Luis:
You know, I, I in retrospect, I would I would love to explain it as a story of how I planned things and how it was brave to make switch at the time. It was just a struggle. You know, mathematics. I was I, I liked mathematics, but I didn’t think of it in the way I wasn’t like everybody else.
My friends were very brilliant, like high level. They were thinking high, low, like they would understand things very quickly. They would do research. It was always in high level. They would speak in formula stuff. I always had difficulty with formulas and I was lagging behind a little bit by in classes. I would always take a little longer to understand.
I would always have to bring everything down. Mathematics is a very levels of abstraction. Like you learn the basics and then there’s lemons on top and then Theorem said at this level people talk and it sounds like philosophy, like it sounds just and I can never understand that. I always have to bring it down, right? And so that took me forever.
So for research, for example, I was slower, like I was a slower researcher than than my peers are. I always found that teaching was was easy. Like I would make talks, people would go like my classes were full. I was teaching was always came up easy, but the research didn’t and I didn’t. Some held a passion for my the research wasn’t there like it still did it and I had a great group and stuff, but I was always the one who was trying to understand it, to explain it.
So I was the explainer. I would keep the talk. But it was just the research, wasn’t it, to, you know, I wasn’t, I started wishing it really wasn’t my thing. And if you took like a year before because everything is academic, just your play a year ahead or something, the position like everything, not like tech. What did you play it two days of before?
So I started thinking that maybe I wanted to do something else and I didn’t know what. I just knew I wanted to survive. And the chances of of getting a job as a professor. But I wanted a place that I wanted and that I was going to do well in my life where not not that high. And so I started just looking at other things and just learn how to program.
I knew a little bit, but I wasn’t really that much. So I took some courses in programing and machine learning and I started liking it. And you know, at the same time as I was applying for professorships, I was applying for other things and, you know, exactly as I predicted that I didn’t, I didn’t get any professor position.
And so I started thinking, okay, I, now I really need to do something else. And I started interviewing in different industry. So I tried in finance or consulting things that my friends went. When they left academia, it was common to leave academia because it’s a bottleneck like this. All these students, there’s all these postdocs and there’s all these professors and not everybody gets a jobs.
And I was one of them, so I had to figure it out. And funny enough, you know, people think I went to Google because the world was my oyster and I just picked Google, knew it was actually the only place like I wasn’t, you know, I in the a lot of math but a lot of industries was not you know, I’d get interviews, but it was clear that I wasn’t for them and they weren’t for me.
Some help. Yeah, and a job as a programmer, I didn’t know how much had a program. So like, I, you know, I wasn’t that clear. But a friend told me who had left Google, who had left academia for Google, that he called me and said, Hey, these come to Google. And I’m like, Well, I don’t know how to program it well.
And he said, No, neither do I. The interviews, math. And indeed I went to the interview and it was a bunch of math questions, like a bunch of the stuff. I liked the puzzles I like to do, like the stuff I was doing from the Olympiads like that. Helpful. So I went in and did really well in the interview and somehow I wouldn’t have done well if if, if it was like another thesis company where they need someone to do the stuff immediately.
Google could take a bet on somebody because they had taken a lot of bets on academics that move and they don’t know programing that well, but they can you know, if you can push through a Ph.D, you can push through a lot of things.
And so they so I ended up at Google and that’s why I had those two years of of not teaching up or programing to I, I just wanted to have a job that I could eat. I didn’t really think about where my career was going and it seemed like I would open doors and and it did it, it opened doors.
Yeah. But I struggle at Google, you know, because I’m not I’m not a programmer. Yeah. So I can I can elaborate more.
Raja:
I can relate to so many things that you’re doing good. And sometimes I tell people to write to you and they say, Hey, did you? Because I’m also first I mean, some if someone asks me, I tell them I’m an educator first before anything else, like machine learning everything. Yeah, that’s right. So entrepreneurship, all of that, it comes later.
But at heart I’m an educator, right? So I enjoy explaining things. I can relate to So many things actually, that you mentioned right after finishing my Ph.D., I, you know, getting a good academic position. It is hard, right? So you can get a small, you know, college or you know, you can you can get it right. But getting into a good school as a professor, as a tenure track, it is very hard.
And when I tell them, I mean, Microsoft was the only job I had at that time being an international student, I don’t get it right, but I can relate to so many things, right? So when I came in at Microsoft, I was, as you say, because I’m the I, I would call them kids. Right? So who took me six years after my bachelor’s to get my rights.
And these the folks who had been programing for six, six years already, I was behind. Right. So I and I was behind not that I was not smart. I worked hard. Right. So but then you then I joined Bing and Bing. I was doing a lot of experimentation and then it became mathematical. I started teaching within Bing, you know, people machine learning was up and going.
So, so many things. I mean, one thing led to another. So I can I can totally I mean, at least I’m one of the people in this world who will totally relate to you how your journey actually unfolded. Right. So so that’s that’s amazing, right? So and I get that yeah.
Raja:
For sure And then you move to Udacity, right.
Luis:
Yes. So people are like, whoa, that’s so brave. But you left Google. Why did you leave Google? And I never tell the story.
Raja:
I get that too.
Luis:
Yeah, yeah, yeah. so I’m curious of how it white what how it changed for you. I’ll tell you how it was for me. Like I, I didn’t think about this was not a brave decision. I was struggling. I was buying from Google like I was. Actually, I’m not a programmer like I, I can program, but I’m very slow and I like to think everything just, like, could happen in math.
Like my struggles in math were because I am thinking high level and I have to bring everything to the basics. And with coding was the same thing. So I was like trying to understand the models and how they work. And I was like, this you recommend with like a neural network? How does a neural network work? And I’ll be obsessed on that.
And of course I had to write code asked and get things done. And so I, I was just slow, like I couldn’t do it. You know. And so I learned a lot. And after two years, you know, it was clear like that. And so it was, you know, my my, my boss was very kind. We were like had a good relationship and, you know, she said, no, you can like, you know, if you get these things done, then you’re okay.
But if you if you don’t, you know, you have to go. But, you know, I would. I would. I were you I would use this time to find something else. And so I thought I was very depressed because, first of all, I failed in you know, in research is what I wanted. And then he programing, which is, you know, three years I had two big failures a you know, after a life of many, you know, great things I had before.
So I was I was very depressed. But I had to go back and say, what do I want? And I thought, okay, I’m just thinking survival. I’m not thinking about writing. I’m thinking, how do I survive? I thought maybe if I go rest forever to another place, it’s not it’s not the level of the place that’s the problem here.
Like, I’m not going to do it. So I need to see what am I good at and whatever it is I need to go to that. And it turns out that, you know, when I was in academia, the research was so so but the teaching was amazing. And when I was at Google, it turns out that, you know, I was teaching on the weekends, like in webinars, and I was teaching like if people came to visit, like I would be the one giving that thought and stuff like that.
So I thought, I just have to teach like that’s what I have to do wherever it is. So I thought, maybe I’ll go teach high school, maybe I’ll go as a, as a teacher. Like I said, you as an instructor university or something, and one day I was taking courses online because I was taking a lot, of course, online.
And then Udacity pops up and it was just like lectures by people who worked in the industry and they were really, you know, they were good teachers, you know, And I thought, I could do this, like I could teach online stuff. I’m always doing that. Like, I’m always explaining stuff all the time. And so I just applied and actually I was in Silicon Valley, so I was across the street.
So I crossed the street to that interview and the interview was Explain anything you want. And so since I always find understand these algorithms, I thought, okay, well, maybe let me explain how a neural network works.
And I used to think about them, would like, I don’t know, rabbits that, you know, I had big years and if you’re how big years your how you’re here and this spectrum here and small and then there was the alliance that would decide if like I made a really they explanation that I knew that I that I used to understand that would slow me down because I had to explain myself everything with cartoonish scenarios.
And I gave them that and they were like, my God, is exactly what we need. And then he started Udacity, and that was the first time that I absolutely clicked with a job, you know, because what they what I had was exactly what they needed was exactly what I had. And so it was the most productive period of my life we were building.
And it was when like I was starting to get super popular, you know, and discriminatory, I and I right, ran neural networks and all this stuff. And so I was learning this stuff and teaching it immediately with a team of educators that were amazing. And I, we created so much stuff, like so many courses. It was just the first time that I really was like in the perfect harmony with my environment.
And after that I’m a woman teaching, you know, I’ve been moving around, but it’s always gravitated towards course teaching why my jocks are always like in a company. The teacher and I do my own thing to what they did to channel and things like that.
Raja:
And that is, that is one.
Luis:
I’m curious of your transition now. How did you go? When did you say I’m a teacher? When did you decide like, this is what I’m going to do?
Raja:
Actually, I was a teacher all along. It’s because after undergrad I wanted to teach. Then I got into my Ph.D. program. I got my undergrad in Pakistan, right. So right after I was applying for grad school as well, got into a U.S. program right after my undergrad, fully funded. I said, okay, I will leave. Right? So but even right after my undergrad at one at the beach, right.
So I told you how I ended up in Microsoft. And within Microsoft, there were opportunities. I and they were there. We used to have a boot camp for all the new Bing employees. So and Bing I joined at the time, but there was there was a joke, right? So they are hiring people so they can have more users for being right.
So in those early days, right. So, you know, it’s early days. I joined Bing and then I so that the teaching, the bug was already there It’s always there so I would teach. And then actually when I left Microsoft, I did not leave Microsoft for teaching data science. Dojo did not happen at least until maybe another year after I started in Data Science Dojo.
So I started another something else, and then I was doing Data Science Dojo. It was a small meetup group in Seattle. I was teaching for free the same thing. I was just doing it for fun on the side, really, with no compensation in mind. The other startup was not going anywhere. And then I had this idea, Hey, why don’t I start charging people for, you know, more of this?
And then the rest was history. Right after that, when we launched. I mean, I think the timing was right. You know, we started around , right? So and the machine learning was everyone was trying to figure it out. So we created this curriculum. I still think it is one of of what I would need, a science bootcamp. It is one of the best curriculums because, you know, having done all of that at Microsoft.
So we were the only boot camp at that time that was beating a bit testing, for instance. Right? So many boot camps were out there because, you know, from design of experiments, all of that, we would teach in because I knew that it is an important thing and in the toolkit of data scientist. But after that, yeah, I mean, one thing leads to another another, right?
So I just said goodbye to the other startup and then my, my other startup, right? So it didn’t go anywhere, but this one started and then, you know, now.
Luis:
Awesome. Well, I’m glad. I’m glad you did.
Raja:
But I enjoy it. Right. So when people going back to it, right. So sometimes and I’ve seen many people. Right. So a lot of a lot of professional and they define their work based on which company they work for on they or what title do they have or what seniority, what another level that are they L or L ?
We very often I believe we live someone else’s life. You know we are. And there is this and, and so sometimes, I mean when you pick something that you love and you can make a living out of it, that’s the best situation to me because you know the paycheck and just a bonus you’re you’re enjoying it, right? Yeah. And I can see this happening in your case.
Luis:
Yeah. For me, like I heard once that when you it there’s a few like sentences that are like describing that like the shares but they are nice like one is like gives you like what do you do you, you’re not going to work a day of your life. It doesn’t feel like world based what you’re doing. And the know is that you don’t when you like something, you don’t, you don’t look for reasons to do it. You look for excuses, right?
Like I just have to do it. Then it’s my thing. And yeah, I think that sort of standardization off of our way of thinking, it’s it starts with education. I mean, they make us well in standard stuff and, and be be followers and be trying to like they try to mold us into the perfect kid student that then morphs into the perfect professional that has all these checkmarks.
And I don’t know, maybe there’s someone that for whom that works perfectly, but I think the majority doesn’t, you know, and, and I’m glad that I never I was able to fit that mold because then you were forced to look for your own thing. Where I like to look for what’s what’s your passion and a line to things.
You know, I think you bought online. What do you like to do for fun? Like a Sudoku or a crossword puzzle or playing, running outside, whatever. And what you want for the world, right? Like, what is it that that, that the purpose that you have for for humanity. And I’m sure that it aligns for you and in the same way as it aligns for me, right?
Like education is a where you can bring opportunities to other and level the playing field and, and, and that’s sort of the purpose. And then and just like to teach whether it whether it’s useful or not, I like to do it right. So yeah I think aligned.
Raja:
As so I noticed that you you also work for a company in quantum computing.
Luis:
Yeah.
Raja:
So what was, what was the did you go back to industry once again in a more for product development type role or what was your role at that company?
Luis:
Yeah, that was very interesting. And I just, you know, when I had like I do big career changes in my life and they somehow paid off even though there were difficulties, I kind of get confident of like career changes are good.
And so I was at Apple educating and teaching at the local Apple University is really enjoying it because I was given bootcamps and like internal consulting, but a very good friend of mine called Alejandro Perdomo, the top quantum scientist, turns out that he, he, him and I were talked a lot.
Then I would pick his brain in quantum computing the entire time. I was always interested. I didn’t know what was. And one day I decided to send him a message and I said, I have a great idea. You want to write at Quantum computing book? Okay. I don’t know. I had about one in computing, but I know how to explain stuff.
So you explain to me and I write the book and he said, Well, I work, but I have a counteroffer. Come work with me. And and I thought my my first thought was, I don’t know how to I don’t know any quantum computing. But then, you know, I also didn’t know programing when I went to Google and I learned there.
So I thought, okay, so it turns out that there was the job of a researcher, like it was a researcher in quantum computing and quantum machine learning. And so I knew that machine learning and a bunch of people knew that quantum computing and we’re learning and machine learning. And I came in knowing the machine learning and learning quantum compute.
So it was amazing. Learning from this guy was just amazing because he knows so much. And to me he was like doing a master’s, you know? I didn’t think a lot of people, even inside the company said, Well, you know, you know that this may not work. Right. And as a guy, okay, like doing a masters and getting paid for it.
Right. And it indeed was, you know, and I did some work, you know, I was it was research. And yet again, my thing is not research, but I was I was enjoying it, learning a lot. And we did get some interesting results.
And it helped push me towards working in generative learning. Quantum is very much involved in actually, I believe and I got this from this, you know, from my like but you know want them is more of a generative thing than than classical.
And I think that will be the advances. So I got more into generative learning and I yeah, I really enjoyed working on that. I, I learned so much and it was, it was a good bet, you know.
Raja:
And can you, can you explain like the simple in the for a layperson what is quantum computing for many many of us might not actually know what it it.
Luis:
Yeah so actually the thing is computers have bits right is a one or a zero and it’s a one or it’s a zero. That’s it. Right. Well one, one thing to do is have you bits which is anything between zero and one. Actually I, the simplest analogy is like a switch that’s on or off and this is like a slider that you can put it on off or like in romantic, you know, and like in the middle and you can put it in anywhere.
So imagine a number and anything between zero and one and there’s more because it’s actually a whole imagine a sphere switch okay a light switch that’s spherical. And if I’m on the North Pole, then I’m turned fully on.
And if I’m in the South Pole, then the light is fully off. And if I’m anywhere in the equator, it’s half, half and as high as I am, the higher I am, the higher the more light, the more the more one is.
And the lower you are, the more zero is. And it’s a full sphere. However it’s in, this is the one. This is the understanding that on a computing path, nobody really knows why or if it is, but is the only thing that matches with with the observations, which is that if you don’t look at it then it’s it’s somewhere the little the special point is anywhere in the in the sphere.
But the moment you look at it, it either goes up or goes down to the North Pole or the South Pole. And if it’s more hot, if it’s higher, it’s more likely to go up. If it’s in the North Pole, it stays in the North Pole. If it’s in the South Pole, it stays in the South Pole. If it’s in the equator, it goes .
And if it’s lowered, more likely to go down, you know, just like, yeah, if you are going to put the pole magnets on. One of them posed a little special point. So but here’s, here’s what’s interesting. I think here’s where the thing clicks and I’m going to be kind of high level here, but imagine that you’re in a computer and you want to do an operation.
Do the number one and two, the number zero. I want to square it or add one or something. I don’t I need to do an operation to the one and then the operation to the zero. I need to do two things. But if here I have a cupid that’s in between and I do the operation to that thing, I’m doing the operation to both at the same time.
The zero state and the one state. So I need to do one operation here while I’m doing two here. Now imagine this doing two more. As I was saying on the classical end, you have to do two operations to the zero on to the one, whereas in the quantum world they do one operation to the cubert in between.
So that’s due to one. It’s not that much, but imagine having many qubits here on a year. I have qubits that are entangled, which is something else completely different, and I do less operations when I have to do more operations here and it’s exponential, right? So here I have to do the n here once the the energy is less.
Of course, when I observe it, a lot of stuff gets lost. So that’s the catch, right? But if I do it, if you do it in a in a in a smart way, then you can save a lot of operations by doing things here correctly and measuring them correctly. And so imagine a imagine and things that are exponentially hard here are linear here.
Obviously, there’s a lot of complications, but that’s the principle. So you can do things like, for example, factoring numbers, right? Like, you know, that cryptography is based on if you can factor any number. So for that, for example, like I can say three times five is thing that’s easy for a computer and that’s also easy. If I have you which numbers multiplying is easy.
But if I say is three times five, that’s acting, which is much harder because if a computer’s given a millions of of of digits, a number, which means it’s not able to factor and it takes millions of years and that’s a basis of working right like it you know it’s right like so it’s basic biography that if I were to factor numbers I could I could break most cryptographic and and and quantum computers have an algorithm source out there and that makes it really much faster to factor by using this principle, you know, and others.
Luis:
So anyway yeah it’s it’s really interesting. I loved learning about it. I had to relearn a lot of linear algebra that I that I thought I knew and I didn’t understand so deeply. So. Ah, yeah, yeah. It’s, it’s, it’s a lot of fun, you know.
Raja:
And do you think that quantum computing can have any implications in terms of generative AI?
Luis:
I well, I conceptually I do, because here’s the thing. I feel like the biological computers are supervised learning machines, right? Because you put the input an image and the output which is yes or no if it’s a dog or not, like that’s it’s completely deterministic. Right. One excuse I’m not going to that. But when it comes to generative learning, classical computers are great, as we’ve seen, but they always have some difficulty, which is generating random numbers.
Right? Like anything you do again or something, you have to generate a random number and beat it through the thing. Or sampling is difficult because at the end of the day, what’s the simplest generative learning problem? The simplest generative learning problem is flip a coin like generate a number between zero, which is zero or one, generate a bit random.
That’s the simplest generative model in existence is generated that you basic unit of information and computers cannot do it like a classical. We cannot generate a random number. It can pretend, right? It can do a a pseudo random number, but they’re not random. If I were to recreate the conditions of the universe and press the same button, I get the same number.
You know? Whereas one on computer, it’s the complete opposite. Cannot control it. It gives you random stuff even if you don’t want to. And so the simplest generative model is and which is generating a random number is the is the hello world of a quantum computer. The first thing you learn in quantum computing is generate one random number and they’re random like they are.
This is truly randomness. And so if you make that bigger, then that the generative models, the quantum generative models don’t have the difficulties that the classical generative models have, which is, you know, that the sampling, you don’t know if you generate the right number to input the thing, you don’t have that problem here.
And so if you train a circuit in the same way as you train our neural network, you train a quantum circuit to start to start throwing out the stuff that looks like your data.
We actually empirically saw results that it it was better. Like experimentally, we compared it to an RBM or to an auto encoder or to GAN even. And this one got the same size, obviously, because you have to do apples to apples. A circuit of the of a small size actually did better than its classical counterparts in the experiments we did.
We don’t have, you know, a theoretical proof that it’s better but we ran the same thing with the same number of cubits as the bits here and it just it just had a better luck ugly when it got, let’s say like, like a generalized bear for example. Like if I give you bits, strings, string, sibling for and I always coincidentally gives you one with two once.
Right I say, not with an even number like things with an even number once if I do that data, the model learns to do to generate that data. However, the quantum one kind of learn beat rule started learning that it was an even number of ones and it started generating stuff that was not in the dataset, far from the data set that satisfied the rule, whereas the classical always kind of stayed close to the you would generate things that was close to the data.
The other one would come up with something crazy that was not close to being in the dataset, but somehow satisfied the rule that I gave the dataset. So we saw that and it was encouraging.
You know, obviously we need bigger quantum computers. There’s no question about that, because the classical music you want us. So that’s a huge like it may or may not happen, like technologically it may or may not happening, but I thought that at least conceptually, was nice, that it was better, that it showed better science that generating things.
Yeah. On the other hand, it was not good to supervise. It was and people have done great in supervising quantum computing, but it was then you see difficulties because they’re not deterministic. I could say one plus one and it says, well, maybe it’s two, but maybe it’s million with a small probability, you know. So found that supervised learning was a classical computer thing and general learning was a one in computer thing In in principle, yeah.
Raja:
And that’s great. I mean thanks for the explanation here. And you you ended up in Cohere toward the end, right?
Luis:
Yeah.
Raja:
How did that happen? You were already in Canada back then, right? So you were in Toronto?
Luis:
I was in Canada, yeah, I was. I was in all sorts of path. I was kind of like I was a researcher, but I always gravitated towards teaching. So I was working in like, you know, giving courses and blog posts and talks and stuff like that and yeah, so J. Alum are called me, which is you must know him, He’s a great educator, friend of mine.
And he said, Hey, we’re building a course in Alliance and you know, I don’t know if you’re free, but you know, with like.
Raja:
And Jay was a friend before Cohere.
Luis:
Yeah, we worked at Udacity together. and I loved his teaching. Like, it’s amazing.
Raja:
He’s amazing. Yes. Without a doubt.
Luis:
I learned, he teaches me the machine learning like so. So he called me at, and so I took it seriously because, you know, and then I interviewed and it turns out that there are two people in the team that was building the course were people I was a fan of already.
There was Sandra. It was like new to her and I wrote a book on LLMs and there was Muhammad who also wrote a book and great book in machine learning that I had to be in touch with before.
So I was like, my God, I want to work in this. It just made sense. And yeah, it started. The boom of LLMs was kind of starting actually a little before Chatty Betty came a month after. So it was in retrospect, I mean, I could say it was a great visionary move, but it wasn’t. I just wanted to be there.
Months before that boom of LLMs, I join an alum company. And yeah, and my job is to build the course. So we build this course called limb University.
Raja:
And one of the best by the way, I so I, I picked up a lot of concepts that I was not clear on from Cohere. One of the best documentations out there are and integrate.
Luis:
I’m so happy to hear. Yeah, I enjoyed that.
Raja:
Yes. I mean, I can imagine right? I mean, because it’s almost like you’re still working in industry, still connected to product, but still you’re more in a teaching type role, right? So I think, yeah, the best position to be in.
Luis:
I’m always going to be teaching and I love that I can talk to experts and ask them stuff like I just have like I can go and ask a million questions to somebody and then I love it. So I might feel like I’m doing grad school and learning stuff and getting paid for it. So I’m like.
Raja:
That’s that’s a that’s a good situation to be in, actually.
Luis:
Yeah.
Raja:
So let’s let’s move on to perhaps more of the the current and the air and how it has evolved. So reflecting on the evolution of the AI, yeah, can you paint a picture of how it has transformed over the years to the present day and how do you see it or envision it changing and, and maybe shorter term or longer term? What do you think is going to be changing? And, you know, let’s talk more about technology first and then maybe we can talk about the society angle in a moment.
Luis:
Yeah, so I’ll answer that, the technology one and then we’ll get to the society, which also will and opens up a whole new topic. I’m well, technologically it’s been wonderful because I got into A.I. like maybe about ten years ago when the boom of like, discriminate strategy. I right. Like the one that answers questions, yes or no questions or numbers, you know, as they say, mid cap or not, is the sentence happy or not?
Is the email spam or not? That could just answer questions that we can answer, but faster and sometimes more accurately. So that was wonderful. And at the time, everybody was it was a common thing to say.
Well, I think the next step is generative stuff like there was very small generative things, right? Like when you say like, I would create a very simple looking image or they’d spit out some some sentence that maybe made little sense, but it looked like it was wording things and so you could see it coming.
Gans Were big, you know, our beams, obviously, when they a large language models came in and the stable diffusion that builds images, that was a huge step that actually surprised me when I started playing with language models and I was like, Whoa, this started to me that was, that was big.
So after yeah, it’s always hard to tell, but I think after after discriminatory machine learning, then it was generative machine learning, and I think I feel like a big step will be multimodal reality.
Like I feel like now language models take text and return text image models return images or read images. Same thing with video and stuff. But like humans, we don’t operate like that. Like if I tell you something I know I don’t remember if I learned through text or through an image or through a video or somebody was talking or I saw something like, It’s all the same to us, right?
And then when when it comes out, it’s the same thing. Like I could I could explain a concept across a drawn image and I could also write it down. And it’s no difference. But for for models, it is different. Like either you do it with a language model or you do it with an image model. So I think joining all those three or all those in not three of those many will be a big step.
I think when we start and I saw it a little bit in general that I think saw by Google, I’ve seen a bit of that. And I think that that’s sort of the the next step on another step would be like, I think when you move, when you start joining it with things that move like robots or something, I think that would be a big step.
But there’s also another sort of conceptual thing that I find interesting, and it’s kind of like I remember that before neural networks, it was a lot like things were pretty clever, like algorithms were pretty clever. They were like probabilistic or something. You know, where you would study that the problem and solve it in a clever mathematical the way.
And then, you know, networks came in and just kind of like solve everything easily and it it felt like cheating right? Like is like if you put everything in a super high dimensional world and learn some stuff with that in your network, just so many parameters it it solves it better than the super clever Bayesian probabilistic math that you had.
And that was kind of disappointing. And I find that so biased that that me that may continue happening right like it it may be like I don’t know the idea I might put into a large AI which model of the data set of like linear regression and just just from words it solved it, right. Like it was like, this, like that kind of things.
And I wonder if like, like my which models are going to start like maybe just out of knowing a lot of text be able to solve stuff that not necessarily requiring text but they’re just so I parameterized that that the solution is somewhere in that space. So it may be disappointing in terms of the cleverness of the solutions, but I think it’ll give us better, better solutions.
And so also I guess this this may go into society, but also like the fact that is now easier, right? Like now it’s it’s more prompting, you know, like the previous generation had to do assembly and and cards, punch cards and stuff. I don’t want to do any of that. I do Python. The next generation is going to know English, right like what language and be able to code like that.
And I find that exciting. It’s I think that can bring many more people into a Yeah, not just the coders, but anybody who knows anything about any problem to solve is going to be able to use technology, not just the program. So I find that part pretty exciting, but it’s going to be more, more universal to everybody. Yeah.
Raja:
And you see as, as the increase in of the models start getting bigger and more parameters and more, you know, bigger context windows and you know, all that more compute being available, more data being pushed and where do you see this added? I mean, what do you think is the most exciting part of it?
I mean, do you think a bigger model is necessarily a better model I know is it’s because there is a lot of effort happening. And yeah, so and I would not name any company here. It’s a company X will come in and say, Hey, we have a bigger model and it has more parameters and more Docker limits. All right. So, so where do you think this is going?
That’s the first part. I mean, do you think a bigger model is better necessarily and that like the next part is how do you think this whole I would call it the race to a bigger or better model, you think in that what you will see the future is going to be?
Luis:
Yeah, that’s a very good question. No, I definitely think a bigger model is is a better model. But I do think that sometimes you need to sort of go bigger and then cut the things you don’t need. So I think it’s necessary to obviously a a bigger model doing the same thing where it’s a little better than a smaller model.
But then eventually you, you know, you bring things down you because a bigger model can can definitely overfeed or do crazy things. So I think I think getting bigger is one step in in the process of getting smarter.
But I do yeah I, I it’s it’s hard to tell again but I think I think we’ll find I think embeddings are the clue to everything and I think we’re starting to get better and better embedding and for anybody who doesn’t know, embedding is basically you take an image and be taken to a bunch of numbers or a sentence and take individual numbers because computers do numbers right?
So a voice or a video and taken to a bunch of numbers. So the more you can have that translation, the better. And so if you find a translation that is sort of agnostic to the type of information that sends the image of a dog to this very close numbers to the word dog and to that sentence dog, right.
That the sound a something like that, I think would make everything better. So I think I think working embeddings you cannot save efforts and you know, they don’t have to be bigger. We don’t need to have million million numbers per word like I think if I find a really good embedding, they have to be at least our sense.
You cannot you cannot summarize data in less space than than the data lives. But I think we’ll be able to find really good ones and then things become easier, right?
For example, ten years ago, classification of language was difficult, right? Like I if you want to make a spam classifier, I would need to give you a million spam messages and a million non spam messages, or at least a thousand tens of thousands, right, for for somebody to build a good enough model now for messages, You know what I mean?
Because the embeddings are so good that if I take the message this a few messages and I send them to the right numbers, then the math problem becomes much simpler because before the numbers weren’t so good. So the tail between spam and spam, I needed a really complicated region that would say, Here, you’re spam, you you’re not spam.
But now the embeddings are so good that I may have something very simple where all these messages get sent here and all the non spam recipes and just draw this line, right? So I think yeah, I think better embeddings just, just make everything easier.
Raja:
And the thing that people converge towards more and more toward smaller models, domain specific, small as that we started to see this term appearing like small language models as opposed to like language models. Right. A clever, I think clever play on words because they are still pretty big. I we get that. Yeah. Right. So so do you think that that’s where we are converging.
I’ve seen one of our one of the companies assembly AI they actually one of the partners for our bootcamp, they actually work on this conversational A.I., you know, which is they models called Nebula. So, and they claim that that is purely for in a support and conversational scenario. And it’s a, it’s a smaller model but a very domain specific model.
And then similarly, you know, maybe health care, financial. So do you see that as a viable direction that that is going to be in the future?
Luis:
I mean, I have seen, for example, embeddings get a little smaller, optimizing for search. For example, you can go from a few thousand to like, you know, a few hundred and they do better. So I think.
Raja:
They because of possibly the cause of dimensionality, right. So you and you’re not too sparse and now the concepts are maybe closer to each other if.
Luis:
Yeah, yeah. Like maybe you live in dimensions, but it really is dimensional, but it’s flat, right? Like if I have a if I live in that three dimensional space, but I am mostly in a plane that I’m really in two dimensions.
So a lot of data is really more of a flatter than we think. And so you have a lot of advanced techniques can bring stuff in to like bring stuffing two dimensions down and then you lose the ability to a little bit of data, but you gain a lot in its simplicity and power of computation and stuff like that.
So I do I do think and I think this is done correctly. There’s a, an there’s a I like an approach of making everything bigger, but also an approach of when it’s bigger. Let’s let’s make it smaller and try to not lose any performance or very little. And I think that that’s the right way. Like I think thing that will continue.
And I don’t think things will be that and I don’t think the gains are going to be that things are going to be bigger and bigger. I think they’re just going to get more clever. As I said, data is high dimensional, data is complicated and we cannot go down. And if we’re trying to have models that think like us, I think we’re I think our brain has a lot of parameters.
So it’s it’s hard to bring that down to the very, very few if you want to fit basically all of humans rational thinking into another layer is going to be big, right? But I hope it’s controlled B.
Raja:
And perhaps. Let’s actually do this. So what I’m going to ask you is going to be at the intersection of society and technology. Okay. Thinking about guardrails. Right. So. So I know you are an educator and I know an academy. Most notably, they adopted this in in the learning and edtech space. And as more and more companies, they start acquiring or adopting generative AI and like language models as an instructional tool.
Do you see any of this there for that? Really any any risk in adopting LMS, not just for education in general? Let’s start with the general industry.
So, you know, anyone who knows how these things work, there are reasons to be concerned. Right. So from any reason to be. Any the reasons could be from anywhere from enterprise data security all the way to, you know, you know, wrong information, getting in wrong and to kids getting exposed information that they should not be propaganda. How do you see all of this?
Luis:
Yeah, I mean, everything. Everything has risks. And we’ve seen it with a I read because I the only thing it can do is propagate what we do. Right. Like take arts and make more of that. Right. And as a civilization.
Raja:
Acknowledge that that’s true with any technology is starting from nuclear technology to internet right to if you’re looking. That’s true with any technology, any groundbreaking technology to fire.
Luis:
I mean, you can use it to coke. You can use it to, you know, destroying something. You write it like. And everything has unintended consequences. Right. Like, I mean, even go way back, like agriculture, like it, you know, made life more comfortable. We didn’t have to go around hunting all the time. But it also, like, made things, you know, started things like property and wires and things like that.
And it you know, we were eating things that we were not prepared for, like that kind of things like. Like any, any, any advancement in, in society like, has its risk whether they’re like obvious or completely unintended and the Internet and that, you know, now we have access to all the information. But you know somehow we didn’t get that much smarter.
Like it’s a lot of misinformation. There’s a lot of manipulation and everything, you know. So I think obviously, this has its risks. And and we’ve seen in data science that like, you know, it basically it propagates our personality and it propagates our mistakes as a society, which are, you know, it’s society’s right.
This society’s sexist societies leaders said So a lot of the a lot of awful things that get that are in the data, whether like not hidden intrinsically in a way that it’s really hard to to detach and that goes in the models so that there’s no escaping that.
And I think we always have to be very, very careful of that. There’s always going to be guardrail. There never going to be enough. I think the only thing that we always have to keep in mind is that these things are tools like fire is a tool. You know, we need to use it. But at the end of the day, we can let it control us, right?
Like we can’t let it take decisions. I would use language models as ah, as a way to help me with a lot of things. But the moment I start taking like decisions based on a language model, then I think that, that that’s a problem. Right. I can elaborate on that. But let’s, let’s look at education. You know, I think looking at it in the in the in the positive side, I think there’s a lot to gain because education has a lot of flaws.
You know, like we sit in a classroom when we’re kids and we’re competing for grades. Why do we have to compete for grades? Why can’t everybody do? Well, you know, we’re working individually. Why can’t we work as a group? You know, at the end of the day, we serve like millions of years working as a group. We’ve been individualistic for not very long, you know.
So those kind of things start out wrong with education. And I think that they can be fixed with I mean, I use I used to depend on last language models here and tropical and all these ones I used them to, to learn like I, now I, I, I go into a language model and I start prompting it like my videos.
I make them like that now. Like I start asking it things like how to explain this to me, but then explain a different way. Now with this little example, help me understand this, okay, let’s focus on this little part of the formula and you can do that. And for the other things like you know about society. I go, okay, tell me about this.
And I had to start analyzing, right? Like, but don’t you think that this led to that? And it helps, you know, And as long as they use it as a tool and I check my sources because many times it lies to me. Like I’ve had problems with some of my videos that I put something that the language model told me and it wasn’t true.
And so I think that’s a great tool for education. I think that that if if, if kids or anybody like adults, anybody can can go in it, you know, ask it things and have it like an expert that you can talk to, it’s wonderful.
You know, obviously you can, you know, talking to an expert is better. But like many times, you know you that resources limited whereas you can be talking to these language models and and learning to be critical because what I was saying about, you know, education teaches you to be obedient.
It teaches you to follow rules to to do things as you were told and not never to think critically, never questioned a teacher in grade school or high school. Nobody ever goes, Hey, I think you’re wrong. This you know, we’re meant to follow orders. And so I think I think we can get you know, I think I think lesson response can be used as a tutor in a very, very productive way.
And I’ve been using it myself. And and so I’m very I’m very optimistic about that. And then other things right. Like, I mean, education is something that is very restricted, like traditional education. You have to live in a place that happens to be in a good location for a particular school. Some people don’t have that opportunity. You have to be in a certain socioeconomic, the place where you can, where you can stay.
Many people don’t have that access chronologically. You know, like if I if I’m and I want to start learning something new, society doesn’t really play well with that. You know, all those barriers can be taken down with with with tech education, with AP. So I’m very, very positive about that. You know, cautiously positive, but I but I am very positive.
Raja:
But as a as a high school science teacher or a math teacher, what do you encourage using to ADP or Bard or one of these applications?
Luis:
You have to I mean, it’s kind of like you can stop kids from using a calculator to do math. Like if you if I give a kid a hundred some and additions and subtractions, I can expect them not to use a calculator at home. Right. So what I can do is I make them do harder problems that they even with the calculator, they will still need to think.
And I think it’s the same thing with language models, right? Like I can’t stop students from not using Category D or any of these models. Right. And at the end of the day, they’re going to use them at working whatever they do. So I may as well just make them use them well. So, for example, you know, you can ask for an essay, but you can also ask for a page book and they need to use chapter three for that.
Right. Like, so they they would use it like so I, I think you you kind of it has to be like philosophy if you can’t beat them, join them because at the end of the day these things are better And at the end of the day, there there’s there’s a fear that we’re going to lose a lot of skills.
Yes. That maybe maybe our grandkids will not know how to write essays. But you know what? You and I don’t know how to make fire by rubbing rocks or you can’t do buffalo, you know what I mean? So, like.
Raja:
We love this analogy, though. It’s but because I see this as a natural progression when Google and you know the search engine there are Yahoo! Actually I actually love Yahoo as a company and for some some discussion for some other time. Right. I think sorry that we lost Yahoo! Because they’re the ones who actually brought search mainstream.
Right. Google happened much and Google was actually one of the like the more popular search engine that I mean, when we saw the search engine picking up, you know, we had the same concern. Right. So and, you know, all the devs, they use StackOverflow, whether they want to admit or not, everyone uses it, tried to, you know.
So I think these are tools just like you give this example, we don’t make fire, you know, ourselves anymore, right? So we have other means to do it. You? Yeah. Do you think that educators, high school or elementary school teachers, they also have to be somewhat upskilled in in these technologies because they cannot. And when I look at that.
Right. So if I’m an elementary school teacher, when I give them this assignment to write an essay or a poem about this or, you know, some something along those lines, well, like, it will take them less than a minute.
Right. So do you think the fundamental the way we approach education or critical thinking, thinking that these tools are going to be put or those approaches have to be put in context of the development that is happening in these tools and the educators, They have to learn these tools to create assignments that create critical thinkers as opposed to, you know, and who use these tools efficiently, by the way.
Luis:
Yeah, And I’m glad that we’re having to do that right? Like that. We’re forced to to change some things that needed to be changed. Right. Like these definitely needed to be changed. And it’s not an a full upscaling like it’s not like, you know, the history teacher has to learn how to program. No, like, they still have to know their thing.
But I think these language models are are are easy to use. And I think it’s not about being technically savvy, but it’s about being creative. Like if you’re a history teacher and you use you use a language model to where you make the students use language model to, you know, understand something in history to do analyze things, to throw ideas at the computer and haven’t developed them back.
I think that’s very useful and anything any, any, any field I think can benefit from this in, you know, education. Yeah. And I and I encourage the migrant crisis teachers to use it for their own benefit. It helps them as well and, and for the benefit of the students, you know.
Raja:
And what is the most exciting application or something that is closer to your heart that you’ll see will result as a it will either start happening or it will it will be better than it has been before.
You know, there is this real alarm or regenerative air era and any application, any area of impact in society and really any area in our life, You know, it could be education, health care, you know, industry, traffic, just anything we have any anything that you can think of where you think, wow, I mean, life is going to be so much better as humans.
We are going to be so much better as a result of Bell Mean, do you have any have you thought about this?
Luis:
I mean, concretely, like all things I’m excited about, you know, I think I think Madison I think that any like car like transportation, a lot of things can be can be made better. Yeah. Obviously, you and I are educators and I always think in education. So I always I always think, you know, the fact that you can now learn by by building things, the fact that you can interact because interaction is the most important thing.
I want to just be sitting down listening to information. So that’s what excites me more. But what really, really excites me of last night, which and I think that I, I want to be like, I’m optimistic. And if I’m if I feel optimistic because I have my pessimistic side, which is we’ll end ourselves by but my optimistic side says we will be able to use language models to have our sort of like like a second cognitive revolution.
You know, I think I look at the early humans, right? The early humans are they were just mostly physical creatures, right? Like the one that would the best and the strongest one. The the smart is like the with the best arms and strongest arms and legs and and something and and they use their brain very, very little for like little things like, you know, should I hunt that animal or that one?
But then they started, you know, the cognitive revolution. They started using the brain more to they went from like thinking about things that existed like a tree or an animal or something, to think about things that didn’t exist, like money, for example, which is an agreement right? Or like a certain mythology, like kind of a story of why we exist or the, you know, gods, things like that, that, that, you know, creating, creating stories.
And that was a huge step, right? So because we stopped using like, like the brain started being more the useful thing, and now we’re at a point that it we still have, you know, physical laborers, many jobs. But, you know, the way to succeed more is using the brain more so like we became from physical beings to rational beings through the use of, you know, machines and and everything.
Every time there’s a machine that does more for us, we become more rational and our jobs are more thinking. Now, I’m thinking that that’s going to be gone, too, because now language models can do a lot of our thinking. So we went from physical to rational because the machines could do our physical stuff. Now machines can do irrational stuff.
Where do we go, right? Do we become obsolete? Are we done? Like, are we as good as our machine? And that’s it. And I believe that that’s not the case. And I think we have to look at what is it that we can do that machines can’t. And if we do it properly, then we’ll be able to transcend to a higher level where the first level is physical stuff.
The second level is rational thinking, and then there’s a third level, which to me it involves anything regarding empathy, anything regarding emotional intelligence, anything regarding spirituality, things like intuition, you know, And it sounds silly, right, because these are basically taking decisions that don’t require your A.I.. And I really if you think about it, A.I. takes very good decisions, but they’re all maximizing some variable, right?
Either profit or the number of patients that are okay or the number of females that I corp is. It’s always maximizing or making, I think some some variable. There are a lot of decisions that we take that don’t require maximizing or minimizing some variable, and they’re normally tied to to do it to empathy, to caring about others, to thinking as a group, to thinking, you know, that I can’t explain with a variable.
And I think those are what makes us human. And it sounds silly, right? If I were to tell you, meaning not in a hundred a thousand years or something, you know, we’re not going to be thinking that much rationally, but we’re going to be meditating and coming up with the the the ideas, the the decisions that we have to take by buying, by intuition Like that sounds silly, but it also sounded silly to the caveman.
If I tell them that in the thousand years for thousands of years, they’re not going to be using their body so much. They’re going to be using only their brain when they only use their brain for a few things. Right now we only use our our intuition and our emotional intelligence for very few things. But I think that that’s what’s going to be the next level for us.
If we do it properly, we’re going to transcend that. We’re not going to be rational thinking humans. We’re going to be much more spiritual. So that’s what I’m hoping that that that this helps speed up.
Raja:
And how much are you and do you think humans are by nature, empathetic? Humans do have empathy by design? Do you think they are like that?
Luis:
We are. I think we’re born with that. And it gets domesticated out of us in in in childhood. Yeah, we get would turn into individualistic beings. We turn selfish, would turn materialistic. But I think we’re born without all those things, you know. And I think we need to like yeah, I think regardless of tech or no tech, we need to transcend that and like decolonize our mind from all those things, you know?
But we’re not we didn’t survive being individualistic for millions of years. We survived fight by working as a group and by not thinking of you and me as different people.
We are the same. We just, you know, happen to look like look like we’re separated. But but we’re not. And I think that what would save society is that if we if we get rid of that, get rid of a big ego and ah, you know, you start, start thinking as a group and I think, you know, I’m, I’m hoping that this step would will, this will be a step in that direction because at least we won’t be occupied the entire time with with rational
things. And we, we have to pretty much whatever the language model can do is what we have to start doing. Right. So I have hopes.
Raja:
And you’re hopeful about Right. So this is going to be for the better for us, right? So whatever the bad things are headed in, despite seeing some of the negative like deepfakes, for instance. Right. So, you know, or maybe, you know, propaganda and all of that, you think that we will be better off as a society as a result of, you know, all the developments that are happening?
Luis:
Well, yes. And I think the developments help. But I think it’s up to us. And I look back and I think it’s never just been technology and AI and I have that sort of disagreement with what many my tech friends because people think we’re one technology development away from being a utopian society where everyone’s happy and we sing Kumbaya Like, I doubt it, because in the past it hasn’t happened.
You know, like when the steam engine came in that, I don’t know, s, nobody said, let’s free the slaves, nobody. Nobody said, yeah, we can do work. That’s that’s, that’s freedom. Lately, nobody thought that if anything it made it worse because it made more trade and more work was needed. Like if anything, it made it worse, as so many people think.
Well, when we have.
Raja:
An easier to bring in slaves, right?
Luis:
I mean. And good. John has bought more. Yeah it was it made it worse. And so now people think well now that that doesn’t work, then we can just relax. Well, that’s nobody’s going to say that. No, nobody said, language walls can do a work now. Finally we can free the wage workers. No, we’re going to figure out a way to work ourselves until we die, because we’ve done that with every single technology development.
Every time a new technology comes in, we never say we’ll make it better. We make say, but it will stick the other way. So it’s up to us to you know, I think it’s I think it’s up to us to use these technologies correctly. I think I think they help. But I think, you know, it’s it’s the big changes in society have happened not because technology, but because of a strong change of thinking, a revolution.
You know, you get rid of an idea that that we had before, you know, monarchy came from the we thought the kings had divine right. And when that ended, we stopped thinking about that, you know, with the French Revolution and things like that. So so I think we need to colonize our minds of many paradigms that that we have existing.
And that’s more up to us. I think. I think technology can help us find things. And and I think, you know, I I’m I’m I’m, I’m excited about technological breakthroughs, but I, I think we need to use them properly.
Raja:
So yeah. So we’re you know, I hear someone who’s cautiously optimistic. Right. So you are, you know, because you’re a technologist and you see what could go wrong. But at the same time, I see someone who wants to see things being better in society, Right? So, yeah. So do you have any when I look at the data there, there has to be.
Not everyone is going to be hey, I’m a I’m a nice human being. I am not going to do this right. So then the society is they have to evolve, right? So now if I don’t have a stop sign or a traffic light, likely I will just zoom through. Right? I’m not going to stop. Right. So yeah, but speaking of that, I mean, what kind of ethical and social considerations are possibly some legal frameworks, some some law making?
Do you think along those lines that this society has to evolve, in terms of the laws? Because laws, how do you handle many of these things? Right. So yeah, just as an example, I’m not taking one side or the other, right? So New York Times So is open the AI, right? So how is it different from a human actually going and learning and becoming an expert?
I read the newspaper and and then we had this unfortunate situation when Taylor saw a picture, I mean, they were doctored and they were all over the Internet. Right. How do you how do you protect people? Individuals? And this has happened in a high school somewhere in New York as well. Right. So how do you protect individuals from this kind of behavior?
What kind of law making what kind of ethical social considerations maybe parents, they need to talk to their children about this? I tell you know, Yeah. You know, just like they would talk to their children about drugs. Right. So and then, you know, is it going to be part of our upbringing then, you know, how do you see that? I’m I mean, it’s maybe the question is a bit ambiguous. I’m looking at the society. How does it shape us?
Luis:
Yeah, it’s very complicated. You know, I mean, it’s like any technology fire, right? Like, how do we stop everybody from setting everything on fire? Like we have regulations, But that doesn’t stop me, you know, like, so we have the regulations. We we have the ways to stop it. You know, that the fire fires, you know, everything, but, you know, at the end of the day, it’s not unstoppable.
Like we we can stop people from setting things on fire. You know, I never will. The best we can do is a combination, as I said, a combination of regulations, a combination of ways to act when when these regulations are broken, and mostly just deter ourselves to not set things on fire. You know, like, I think that’s the most one, you know.
So I think for is the same thing you know can’t you can’t stop it. Like regulations are always late. It’s very hard it’s very hard to put to stop. So I don’t know like you can go back in time like these deepfake like things generate images that look a lot but steal the work of the artist. You can go back with back like you and untrained the model, you know, and you can’t trace it.
So you can say, well you this looks a lot like this it plagiarism or not like this kind of stuff is very ambiguous and is very hard to detect in how to hard to stop I would say. So some things we need to do, which is we can’t trust the voice or video anymore, just like I can trust an email.
Like if you get an email from signed with by my name that says Please for this, please wire these money. You won’t do it because it’s an email. Anybody could type that. Well, now if you get a video of me with my voice saying, Hey, I do, please forward, this is money, you won’t do it because you can’t trust this kind of stuff.
So we’ll have to figure out ways around it. But I think humans have to not, Yeah, not. Not. Not trust us. I feel like, yeah, we were teaching a course of responsibly AI right now, and it’s is very challenging and we have, it’s sort of three levels of responsibility. And the best that each one can do is like, I mean, which one should do the best?
It’s kind of like an analogy of, like electricity, right? Like there’s the electric plant, there’s the person who builds the oven that uses electricity, and there’s the user who has a kitchen and builds. And two is a each one has to be as responsible as possible. So the electricity company has to make sure that it’s steady and and that, you know, it doesn’t have bite spikes and everything.
The one who builds the oven has to make sure that it doesn’t have a short circuit or something. And the one who uses has to make sure that they don’t use it in the wrong way, that create a fire. And one has a small responsibility. So we have the creators of the language models, you know, go here, open the everybody Google.
They have to be as responsible as possible and not like like have to make sure that these models are as unbiased as you can and as safe as you can, which is not easy. Then the the people who build the apps have to be careful as well to make sure they’re not, you know, they’re not propagating horrible things and then the users as well.
So it’s the responsibility of all of us. But it is but it is hard to it is hard to regulate and it will always be. But what I think is that it kind of I don’t know, it kind of reveals cracks in society that need to be fixed regardless. I’ll give you an example. This is my opinion that when when like an image model creates plagiarized the work on which is an awful and it’s hard to detect as I said you know and artists are thinking you’re ruining you’re ruining my career, you’re ruining things.
You know, but it can’t be stopped. So when when I think of that, I think, okay, we have a problem. Can’t be sold. Is it that was already a problem before. And so this happened to me like I one, they put something into a language model and it was a mathematical question and it answered in a very mathematical procedure.
And I knew that some part of it was work of someone I knew. And so I when I talked to that person one day, I said, Look, I took a screenshot and I said, This is exactly you are thinking like, this is exactly your style. And it didn’t credit you like it didn’t you know, it was like you were steering, but it didn’t say and, and, and, and he said, That’s okay.
Like I, I like that my knowledge gets propagated, right? So I’m thinking, why does this person not get concerned that that got plagiarized? But the art is dust. And then the the answer was because the mathematician has tenure and a stable life and doesn’t have to sell theorems for a living. Like it doesn’t be like, look, I discovered at the end.
Raja:
Some very interesting viewpoint.
Luis:
You get it, right? So this person has a a stable life and so they can get a concern of propagating knowledge and not having to worry about getting credit. Of course they get credit in like big and stuff like that, but they didn’t have to worry about getting plagiarized by a language model. On the other hand, the artist does get worried because that’s a day of eating that goes away if they don’t manage to sell that image.
So is language models screwing artists, or are we screwing artists? You know what I mean? Like, have we been awful to artists beforehand by forcing them to have to sell their images to be able to survive and could we do better there? Because if we could do better there and then the artist was okay being able to like, be able to live without having to to sell their images, they wouldn’t care.
Maybe I’m not maybe they wouldn’t care if if something looks a lot like their art, they would probably be saying, my God, this is immortalizing me. You know what I mean? I may be wrong. Maybe an artist comes in and says, No, there is. You’re wrong, mate. I don’t discount that, possibly. But I think that a lot of things that I is bringing is cracks in our society that already exist.
So we need to be better with that, you know. so you know, I, I think, I think a lot of problems cannot be solved with A.I., but they, they need to go deeper in society and fix the original problem.
Raja:
Yeah, I love that. I love that. I mean, it is basically exposing the cracks that already exist in society. So do you think that the legislation and regulatory bodies, they’re ready for all of this? Because in the end of the day, right. So when casting, we invented cars, right? So we had to come up with traffic lights and stop signs.
Right? So, you know, we could not say that. Yeah. People rode horses earlier and, you know, big self-governed, right. So you, you have to have some sort of regulations. I mean, what are your thoughts? I mean, which country, which regions do you think are actually doing something and which regions the rich countries are oblivious. And of course, I’m talking about where you see not every country in the world, but countries that are more, you know, a more and ahead in technology. Right. So so where there is more technology, adoption.
Luis:
And good question. I don’t know of different countries doing different things like I feel like a lot of them are making efforts like there’s definitely efforts from the tech side, there’s efforts from the government side. And I think, yeah, I mean, I an industry cannot regulate itself because it’s a conflict of interest, right?
Raja:
Yeah. I mean, and many, many companies, I mean, the shareholders actually or the board actually gets to decide Yeah. What direction to take and you know so it’s a different class on connectivity but it’s a different optimize. I mean they are optimized in a completely different way, I think aggression as opposed to, you know, not necessarily best interest of humanity.
Luis:
Yeah, exactly. So if you if you regulate yourself, it’s going to happen. What’s happened with every industry that has regulated itself, which is, which is chaos. So I think there definitely the external, external forces regulating this. But it’s it’s a difficult problem because things change all the time. I mean, largo’s much slower, the hostile, you know, to think about things more carefully, whereas tech is kind of like, does it work ship, you know, launch.
It’s a different speed of things. So it’s yeah, it’s, it’s, it’s not easy. I think, you know, I think regulations are absolutely needed. They can only go so far. And I think it’s like security, right. Like you can put a police at every corner. But why are cities safer than others? And the reason is because in some places there’s less social inequality.
You know, if you go to a place where nobody needs to stop, they won’t grow up. If you go to a place where people are starving, there will be there. There will be insecurity comes from there, and no amount of police will be able to do to stop that. They may make things worse, you know. And so I think, you know, if our society was not you’re not in a necessity to to do something about that.
The only way to stop things. So again, revealing cracks in our society, you know what I mean? But yeah it’s it’s hard on like I don’t know the solution.
Raja:
Yeah it is our problem. I completely agree with you. I mean, I’m asking you a difficult question. Questions, I guess I like to, but I’m trying to figure this out. It’s. I’m trying to, you know, understand all of this, you know, as you as you as you learn more about these technologies. I do learn more about what they are capable of.
I mean, these are the questions that naturally come to your mind as a citizen of this this world. I tell you to try to see, try to understand how this is going to impact you and the future generations. So do you foresee more of a more of a symbiotic relationship between humans and I, you know, so we have traditionally talked a lot about human computer interaction, right?
So now Newt sees some kind of, do you think machines will surpass women intelligence? And if they do or if they don’t, I mean, so which whichever it is. And how do you see this kind of, you know, coexistence? How do you see it evolving?
Luis:
Yeah, Yeah. No, absolutely. I think they will surpass human intelligence as we know it. Like rational intelligence. Yes. Like that. Same way that they’re faster than us, right? They’re stronger than us, and they’re more precise than us. They they already succeeded. The human body. They were just missing the brain. And now they are succeeding the brain. So anything, anything that’s rational about us, any any decision that we take rationally and with data, anything data driven or any anything like that, I think they’ll be able to do it better.
And that doesn’t mean that we’re done. And that’s where where it where it it comes again that we have a higher level that’s not rational. And I think if if we managed to do a symbiotic relationship, it will be like what we have with the current machines, Right. Which is that Yeah. I mean if our car runs faster than we can pretend that doesn’t, right.
But I use it to go faster, right. I don’t race against it. I’d be silly to race against the car. Right. But I drive it and it gets me somewhere farther and in the same way. Yeah, I think if I. If we manage to develop our mind that is not rational, you know, like at our or higher level emotional intelligence and intuition.
And if we develop that much more than what we have it right now, we’ll be able to drive these cars, these, these models, these machine learning models to get us to much farther places. And I will be able to do things in unimaginable things. You know, I’m being super optimistic in this in this stature, but we’ll be able to do great things.
Raja:
But we have no choice. Right. You know, we we have to remain optimistic. Right.
Luis:
So we don’t have a choice. And and so will will be taking the high level decisions that are not based on some numbers, but that are based on much higher things and use that machines to do. They always want us to do the mundane, rational things that our mind does right now in the same way that now I’m using a bunch of machines to do the mundane things that my body would have had to do back in the day.
Raja:
I’m yeah, so yeah. And so yeah. So you have a lot of fans out there, right? So even within my team, my young, younger data scientist some, they were absolutely excited when they heard Luis Luis was going to be teaching at the boot camp and some of them were actually attending the session and they loved it. Right? So so for any aspiring data scientists in general, what would be ah, when I say data scientist, I still consider this a broader term, right?
So you know, it includes LLN Right. So traditionally we have used this term for, you know, classic machine learning and analytics, but assume now L am as part L alums are also part of this toolset. So what are some of the, what, what would be your advice for them to, how can they excel at what they do?
How can they be really good at what they do?
Luis:
Yeah. Yeah, no question. And thank you and say hi to them. I’m very happy to has very happy to give a talk at the Science Dojo and and it was a great crowd so yeah thank you. I’m yet to the ah well I mean I think one thing I did which I was forced to but some people are not is is that I had to follow what I liked.
I couldn’t do anything else right. So people are able to do things they don’t like. But even then, you know, you should still they’ll try to follow what you like. It’s like the gradient should be your joy. So like there’s we always kind of know, like we never sometimes have other ideas and paradigms and stuff, but we always kind of know what we like.
Like if you ask anybody, what would you do if you, if, if there was no if everybody wasn’t made, made the same salary and and there was job security, whatever you do, what would you do? And for me, it was teaching, but I never knew it concretely. And everybody has that. So I just try to find it, try to all try to explore and always be true to yourself.
And don’t worry about job title and things like that. And obviously that may be a tone deaf thing to say because we have to survive. And sometimes some jobs just keep, you know, more money than others. But to the extent that is possible, to the extent that you can explore, never stop exploring. That’s one. And in order to learn stuff, I feel like I have a way to learn things.
That is, I imagine that a lot is like soccer, for example. It’s somebody wants to learn soccer and they don’t know anything. What do you do? You give them a book of soccer rules and techniques and they read it and then come back. You know, you give them a ball and say, go kick it, run with it, explore, experiment, get some friends and play and kick it between you and each other and then enjoy it, you know?
And then later I can tell you what the rules are. And then I okay you and then I can tell you more. And then, and then if you become an expert, like if you become really good at it, then I give you a book and I say, Look, these are techniques or these are the rules, you know, And but but the book is the last step, not the first one.
And I in that were it drives me crazy that when people ask us, how can they get into the science and the first thing they do is or you have to know how to program, you have to know linear algebra, you have to know calculus, you have to know probably a lot of give them a list of things not that’s not like to me, that’s not good work for people, but to me that’s backwards.
And so what I say, you want to learn that they’re just just go in a language molester prompting it and start asking it how it works and start asking it stuff. Soon you’ll realize what it can do and what it can’t and how to hack it and how to. And that’s the engineering that’s like that, which may be the next thing you know.
I don’t know, but it may be the next data scientist. And then when you say, okay, well now I want to do one more step, I want to I do a model to ask answer questions about soccer. Okay. So then you go and read something and learn how to plug in some data to the like you you’re basically learning in the practical way.
And then later, if you’re so into it, then I’ll be like, okay, here, let me teach you something. You’re obsolete, you know, or programing or something like play with it, you know, play with it, enjoy it the way you would learn a sport and I find that that’s that’s how I try to learn things and that’s how it how it works.
And, and it’s more fun.
Raja:
A Yeah, have fun, right? So and so at the end of the day, I always.
Luis:
Look for that and for, yeah, the things that matter to you, right? Like if for data scientists, I say take a data set of something you like, maybe something of a mission that you have in, in life or maybe some or you enjoy some of you joy like they could, they just it’ll bat them play with it, plot it, see what, what relations they are.
Are, are, are they two groups of data, one here, one here. Well what are they? You know, like ask questions about that. But with your funding involved, like with your hobbies or whatever you like and that’s the way that it, it doesn’t feel like learning. It feels like, like you’re just exploring stuff. And then the the data science becomes the second thing because you’re just enjoying about something you like already, you know?
So I think that’s I think that that’s to me that the the way to learn and I when I teach, I try to have that. I try to have everybody do their own thing and have as personalized as possible. So it’s really different, you know, and we all have something different to offer. There’s no reason to put ourselves in a in a mold and denied to the world all we can give it right?
Luis:
Yeah. Yeah.
Raja:
So let’s, let’s switch gears and shift gears and go back to the personal side of you. Right. So. Right. So I follow you on on LinkedIn and I see almost like an activist, you know, someone who cares deeply about society. Humanity doesn’t matter whether they are from Colombia or they are, you know, from Luis’s family. So they be any people anywhere.
And you’re pretty vocal about that. So where did you get this from? And do you think that, you know, as that technology is to should I be worried about what’s happening elsewhere or should I be only worried about here? I want to be the best at what I do and I can’t change things, right. So.
Luis:
Yeah, that’s yeah, that’s some very, very strong. Yeah, yeah, that’s a very strong side of me and I, it’s always been, I mean when you grow up, you know, as I know in an unequal society, like it’s sometimes it’s hard to see injustice, but when you see it you can’t unsee it right. And the more, the more you see the like, you know, when you see imperialism and you see the genocides that are happening right now, I think we’re technologists, but at the end, we’re human first, you know?
And I think not being enraged when we see a genocide like what’s happening, the many that are happening right now, it’s it’s impossible. You know, I and I find it impossible. And I tried for a while to not post stuff, especially on LinkedIn, which is a different flavor than, say, Twitter or Instagram or something. So I kept it for a while.
But this this collated so much and we’re seeing it in real time. You know, this is not something that’s happening behind in the news or something. This is something that we’re actually seeing it being recorded with with phones. And it’s it’s very difficult to see it. You know, it’s very difficult to see that it’s happening and not be able to do very much about it.
And I wonder a lot what what knowledge, what what we can do. at the very least, I think with education we can teach people to be more critical because I think that’s the problem that we have, right? Like we are sort of domesticated to be followers in inner school. We just think of the picture as big.
Your I, we never questioned. So then we don’t question the news, we don’t question the corporations, we don’t question governments. And so we have some some some paradigms that we just never or never question.
I think there’s a very strong, strong thing here that that people have, which is frankly, that that we are taught that some people that to see some people as less humans than other, you know, if this kind of genocide happens in in say you a developed country in Europe or something everybody gets concerned which they should, you know.
Raja:
All of us should. Right. So, all right.
Luis:
Ever something happens we should all get concerned. But we are selective, right? Some places it happens and we go, my God, humans dying and other places it happens and nobody blinks an eye. If it happens in Palestine, nobody believes and I it happens in an Africa. Nobody blinks. And I Middle East, so many places, just people have less sensitivity.
And I think it’s just that inherent idea, but that some people are considered less humans than other. And and that’s not true. I think anyone that gets massacred or gets tortured anything, I feel it like if it was, you know, my my brother or sister, my mom or, and that’s what we should feel. We are a human race.
You know, I think we should we should all feel it. And I think we all feel it internally. But we have so many layers upon layers of of of colonization in our mind, our domestication, that we that we refused to are that we just we just it doesn’t get to the surface, you know. And when I think why why am being emotional and some people have been wonderful you know that Paul bigger and people have done a lot more vocal than me and have and have actually risked more than than I have.
I, I just think that, you know, what are you when you look at this and in or years what’s it going to be right. Like obviously history always figures it out much later. You know, we know slavery was bad. We know the Holocaust was bad. We know and you know, apartheid was bad. We know segregation was bad.
We know genocides that happened in Africa and Asia. We we know they’re bad. But at the moment, they are controversial. Right at the moment when apartheid was happening, there was like there was controversy. Like people were saying it’s good. Some were. They were saying it’s bad. Now we all know it’s bad. You know, So so in the future, we’re all going to know.
And the question is, what do we do about it and where we at least vocal about it. And I think I’ll regret not having been about it. And I think I think it helps. I think people sharing it. This is this is something new the first time that that we all kind of have a voice. And yes, it sometimes gets the algorithms getting quiet.
It didn’t stop. But I think this is a time when we all have a voice and we should use it. And I’m conscious that I have a little bit of privilege there. Right. Because some people write to me and they’re more junior and they say, I want to I want to be vocal, but I can’t because my career is in danger.
And I understand. And if I was in their position, maybe I be or some people are more, more important to me. And they have actually like people whose careers depend on them and they can lose funding and they can’t lose words and they also can’t be vocal, you know? So I’m in this kind of sweet spot and I’m and I feel like I can I have I have to and this probably a lot more that that I can do in a lot more that we can do.
I’ve been involved in having violence and groups and things like that. But yeah, I mean, it’s it’s difficult to witness. It’s difficult to feel, you know.
Raja:
Do you think it is because of our lack of critical thinking ability or do you thinking humans are inherently good? It is just that they don’t have the information and they are unable to process the information that is at their disposal?
Luis:
I think so. I think we’re born good. I like to think that. I think this society makes us individualistic because it’s easier to control a group of millions of people that are individual and playing for themselves and for their interests and having different stratification where these ones have more privilege on this one side. So they have to protect it and putting people against each other, I think it’s it’s not natural.
I think we are born without knowing we’re different. You know, if we’re like I like like, like the answer or the bees or something where they are just a group and they work as a group. But the collective consciousness, they don’t really think, I’m innocent and you’re an act, but I’m going to try to get the best.
You know, they they work collectively. I think we are that when we’re born, but that that left free for others and that that gets quickly domesticated out of us by making us compete for scars, artificially scarce resources like rates and things like that, and by making us like pitting us against each other and making success an individual thing as opposed to a group thing makes us like that.
And then we’re easier to manipulate because it just feels like, you know, somebody who’s getting oppressed, This is not me. So why, why should I care? You know.
Raja:
Perhaps demonization as well, Right. So you you sort of you present the other kind as No, not civilized enough or not human. Yeah. And that’s how they are.
Luis:
And that’s how they got away with that. With the genocide of the indigenous people in the American Sun in Oceania, in if any nation in Africa, like that’s how you get away with with genocide by putting the other group as savages when many times that way more. But the way more white, like the Eastern cultures are so much wiser, you know, and indigenous are so are so wise.
I’ve been having now that I come back to Columbia, have the opportunity to talk to, to meet the indigenous and talk and learn more about about their culture and their way ahead of us. And in terms of, you know, in terms of group thinking and in terms of, of of everything. So I think I am I definitely yeah, I think this, this sort of a way of making the others look like they’re near-sighted.
It’s an awful thing, but it’s what is what allows for atrocities to to happen and for people to be to be apathetic about it. Right. Because at the end of the day is, is, is apathy what, what allows all these, all these atrocities to happen and empathy.
Raja:
I mean the irony is that a lot of us, we we keep talking about empathy and all the HPR articles and then internally within the, you know, the corporate environment, we taught about this empathy, that and teamwork and somehow as humanity. Right. And as the I zoom out of this, we forget about this, that these principles actually also apply as humans, you know.
Luis:
Where we’re learn to have like, you know, in a small group, right. Like in in in a in a group of our team or a company like try to work as a group but never, never across all humanity. You know, I we come up with the I think even question the nuclear family for example like you know it’s obviously nice that nuclear family some nice concept but at the end of the day we’re tribal like we’re we’re meant to be in bigger groups and care all about each other.
Right? Like when you when you ideally we would if you want to control people, you put them as individuals what you obviously individuals and reproduce. So we have to go one let’s compare that’s the most you know one can do so we have like sort of tiny tiny units of tiny and Hannah that the smaller the unit, the easier it is.
It is to control. But at the end of the day, I, I think I think there are some some small groups where you can where you see empathy happen. But it it should it should be much you know we should be able to to have empathy for every for every human being and for every living being because you let’s not talk about how we treat the animals or the planet, etc., during that story.
Raja:
And you’re optimistic about the decade and, the centuries to come as human?
Luis:
I hope so. I mean, I think I I’m optimistic. As I said, you know, that we can go the direction either we annihilate ourselves or or we figure out how to how to live in peace and how to live in harmony. So we have the power now to annihilate ourselves completely. So it could go either way. I don’t I’m hoping it goes the way of fire, the way of all.
I’ll figuring this out and and advancing as a as a society. We’ll see. I don’t know if the.
Raja:
Yeah, I mean, we don’t have any other option. And so pretty much we have to remain optimistic.
Luis:
Yeah. Yeah.
Raja:
So Luis, I’m absolutely enjoying the conversation, but I think the conversation, I mean, all good things do have to come to an end, so maybe I will just start wrapping up. So I will throw some rapid fire question that you So given your background as a mathematician, right, very. You have been in some based on your background, I created these rapid fire questions and you have to answer very quickly and maybe try to rationalize.
But that’s about right. So so one word answer calculus. Calculus or algebra?
Luis:
Algebra. I’m more of a discrete person than like a tiny deltas, a continuous person. So yeah, algebra.
Raja:
Like PI or Euler’s number.
Luis:
Or I’m more of an Euler’s number. Kind of underrated. Yeah, I like that. But it gets all the attention. Yeah.
Raja:
Okay. Yeah. And why is that?
Luis:
You know, I think I like e is kind of the number that multiplies most by adding the least. You know what I mean? Like, for example, if I one numbers to add to and I want to have the biggest product I can have like times one less times one, which is, but I can have ten times ten, which is.
And the best I can do is three, three, three, three, three. Right, because it’s the closest to eat. So the closer you are to you, the more you multiply by adding the least. For example, E to the X is always bigger than X to the year. That means a lot. So I that’s one of the reasons I like.
Raja:
Okay, that’s, that’s wonderful.
Luis:
So don’t get me wrong. Yeah, but likewise.
Raja:
Yeah. Because I mean pi is also, I mean a mysterious number so I mean that’s I put you in a tough spot here, right? So it was a difficult.
Luis:
Task but yeah. Yeah. It’s like which do you like the most, you know. Yeah.
Raja:
Bayes Theorem or Euler’s identity.
Luis:
Who? I love Bayes Theorem. But you can’t beat Euler’s identity. It’s just. I mean, base DRM is cool because you use it. We use it without knowing to make all our decisions. Right? Like we we have our idea of what things are and then we get more information. So we change the probabilities in our head. So base down is pretty close second.
But you cannot beat or this identity has the most beautiful equation in math like E to the I plus one equals zero or minus I sorry, one of those. Okay. I think it might be plus one equals zero. Right. Yeah.
Raja:
And I go to fine tuning.
Luis:
I always brag, but you need both. You know, fine tuning is like working your model in college so it learns more stuff. But then Rag is like giving your model a book to answer questions or Google search. I find that rag is the ultimate thing. That and then like Calhoun’s edition, so I’m going to go with that. But fine tuning you absolutely need.
Raja:
Yeah you have lived and both in Toronto and San Francisco right?
Luis:
Yeah.
Raja:
Which one?
Luis:
Toronto San Francisco school. But but and well Toronto’s to find more diversity join Toronto.
Raja:
And you like the cold.
Luis:
Not the cold no
Luis:
Yeah San Francisco has beautiful weather and love San Francisco but I’m going to go for Toronto.
Raja:
Okay. And forgive me for the pronunciation for the next that I’m saying so and correct me if I’m wrong here. Bandeya Paise or Ajiaco
Luis:
Ajiaco. and in front of the town. So I hear across the plate the bow ties the soup with potato and chicken, which is good for when it’s cold. But but that’s kind of cold. Not really, but then a mountain. But they have bison, the Medellin plate. I love Medellin, but I’m from Bogota and I’m going to go for the I here.
Luis:
Yeah.
Raja:
And I was looking at bringing up Aisa. Is it like Bahia in Spain? Is it something similar? Because the recipe that I saw, it’s similar rice and, you know, sausages and all that.
Luis:
It’s like a platter with everything it’s in. And the two, like you have it once a day and then go work the feels the entire day. So it’s got, it’s got plenty of rice ground beef or like crunchy or B an egg avocado. I’m missing half of the ingredients and I Republica corn thing I mean it’s if you have one you are in a food coma for three days I like yeah, but I yeah.
Raja:
Yeah. Okay. Yeah, that’s that’s great. So what is the most beautiful mathematical equation?
Luis:
I know I well, I already said Euler’s identity, so. Euler’s identity, but actually, I’m going to go with the other another a less known Euler phi vertices minus edges plus basis equals. I think that one’s pretty cool. So if I have a square, for example, what do I have for edges, for vertices and one and what two faces because it’s the inner outer envelope minus four.
Plus two is two. Right. And actually if you’re, if you’re in the sphere it’s two like in the plane, it’s this huge sphere. But Taurus is different if you’re in a place with holes, it’s different. So it’s B minus E plus F equals something different with the genius, with the whole sun. The number of holes that you have in your surface may not, but that one is pretty cool.
Yeah, yeah, yeah.
Raja:
And what is the most underrated or underappreciated mathematical concept?
Luis:
Look, Yeah, that’s. That’s a good question. I think a bunch. I mean, I think one that I, but it’s less obvious in this like you can have for example I can make a polygon of as many sites that I want. I can have a triangle or a square of four sides or a Pentagon or at Gon and a gone.
Luis:
And it’s very easy. But I polyhedron I can only have a few of them, right? I can only have a fire hydrant with four faces, a cube with six faces, an octahedron, a dodecahedron or icosahedron. When you or three, four, six, eight, . And why can’t I have a polyhedron of any number of sites that is regular?
And believe it or not, the answer is not with areas or lengths of things or angles. It actually comes from the B minus he plus F equals to equation. It’s it’s a matter of you can’t count and do some he said some points in the correct way unless you are in one of these few cases. So that’s that’s under-rated I think I don’t know like yeah like maybe yeah that would be one fractals.
I also like, I mean I think I tell you, you know we have dimensions, right? Like an edge of dimension. One plane is dimension to cubes dimension with three fractals to have any dimension like logarithm of something like that, I think that’s pretty cool. I think those two.
Raja:
Yeah. Okay. Now, favorite mathematician.
Luis:
I have to galois. Yeah, just tremendously underrated. He was a genius. Died at in a duel with a huge revolutionary and died at and he would have done a lot more. You know, math would be different if he had lived longer, but he’s the one who answered, you know, you know those old mathematical questions of, like the Greeks had.
If you can trace the angle, like you can bisect an angle with square and compass, cut it into. But trace acting is hard. It’s impossible to do it with strength or it’s finding a square with the same area as a circle, or finding a cube that is twice as volumetric, as big volume as the idea. The Greeks pondered with those questions many years ago, and it was like Galois in that developed the math that you could use to solve those problems and had nothing to do.
It was Galois Fields and and groups and come up with the most beautiful proof that you cannot do those. So yeah, Gallo was amazing. And the other one actually it’s closer to my heart because it’s a party of Alexandria and I’m a huge fan because she was more of a popularizer like she was that she would. He does what I like to do, which is take the stuff that exists and understand it better.
And so like, I’m I am I am a huge fan because she is she broke many, many barriers and so it would be my year notes to I think are are my favorite mathematicians.
Raja:
Yeah and I promise this is the last one right. So favorite machine learning or a textbook?
Luis:
I’m. I’m biased. I like, I like memory rocking machine learning. That’s a very biased answer. But,
Raja:
And this and that, you vehicle work your peers from here, right?
Luis:
No, that’s. That’s mine. Yeah. So, I mean, I’m sorry where I hit. Does everybody like it? I have a hard time understanding machine learning books. There always go to the formula is immediate and hard, but I’m going to give you some some have a list because actually Jay Lamar’s book is coming up pretty soon on a little else that one’s really good and my other teammates have great books.
Still, Sandra Kubelik has a book on and Jeopardy! And and Mira has a book where Alma has a book release Planes machine learning in my like images and stuff. It’s a PDF. You take it, take a look. It’s yeah, it’s about like you can you can like it’s all written with it with images like explaining machine learning. But like it’s all, all visual really helps someone like me and needs to learn with, with Bishan.
So yeah, my and those those I really like.
Raja:
Okay. So Luis will have to close it now. I know you will. You took the time out on a beacon, so thank you so much. It was a privilege to have you at our first episode. I really enjoyed the conversation.
Luis:
Thank you. I really enjoyed the conversation, too. That’s wonderful. And thank you for the work you do. And I’m a big fan, so. So thanks for having me. It’s been a pleasure.
Raja:
Thank you so much.
Luis:
Luis Thank you.