Disseminate: The Computer Science Research Podcast - High Impact in Databases with... Anastasia Ailamaki

Episode Date: March 3, 2025

In this High Impact in Databases episode we talk to Anastasia Ailamaki.Anastasia is a Professor of Computer and Communication Sciences at the École Polytechnique Fédérale de Lausanne (EPFL). Tune in to hear Anastasia's story! The podcast is proudly sponsored by Pometry the developers behind Raphtory, the open source temporal graph analytics engine for Python and Rust.You can find Anastasia on:HomepageGoogle ScholarLinkedIn Hosted on Acast. See acast.com/privacy for more information.

Hello and welcome to Disseminate the Computer Science Research Podcast. I'm your host, Jack Wardby. Today is another episode in our ongoing High Impact in Databases series. And I'm really pleased to say that I'm joined by Anastasia Alamaki, who is a professor of computer and communication science at the University. I'm going to test my French pronunciation here as well. École Polytechnique Fédérale de la Zone, or EPFL as I always reach for that. Just use the acronym, right?

Yeah. Uh, Anastasia leads the data intensive applications and systems laboratory. She's also a visiting researcher at Google. And she also has co-founded, is a co-founder, should I say, of raw labs, uh, company that develops systems to analyze heterogeneous big data from multiple sources officially. I'm sure we'll be chatting about that at some point in the podcast today.

Yeah, so welcome to the show, Anastasia. Thank you very much. Cool. Well, let's jump straight in then. So I always like to start off by asking my guests about their journey and why they became a researcher and when was that sort of lightbulb moment where like, yeah, I want to research databases. So yeah, tell us about your story. So, I grew up in Greece.

I'm from Crete, which is an island where computers, at least to my knowledge, had not arrived when I was in the last grade of high school when one has to choose what they want to do in life What school they want to go to? I was there. I was filling out a form that I'm supposed to fill out to to declare my preferences for university departments and for university departments. And someone walks by and says, what are you going to choose? And I said, well, chemical engineering, because I really wanted to become a pilot,

but as a woman, that wasn't an option for me. So chemical engineering, great. Okay. But did you see that other school? I'm like, what's that other school? Computer engineering, which I completely missed because I didn't know what that meant as most of the words on that sheet. I was living and breathing mathematics and chemistry. So I'm like, okay. Yeah. I did. What is this? Oh, this is really brand new technology stuff and all that. Nobody

knows about it, but don't even bother because it's the only one in Greece and they only take a hundred people from the whole country. I'm like, oh really? And I put number one. Many years later, I finished. Five years later, that's the normal number of years in Greece, in engineering school. I finished as a computer engineer. I was working as a network engineer at the time back in Crete, not really thinking about graduate studies and everything.

Five years, beautiful five years in Crete. When another challenge comes around, I want to change what I'm doing. I'm not really learning as much as I wanted to. So looking around to find a job somewhere else. Again, same problem. Couldn't find anything suitable. The things that I liked, I wasn't the right fit for because of gender, because of underqualification, because of under-qualification, because of

over-qualification, whatnot. I was in the year 1995, four, five. And then opportunity knocked on the door in the form of a student who said, I'm going to go to the US to do a PhD. And I'm like, okay, what does it take to do that? That sounds good. Yeah, that sounds good. Right.

Yeah. That sounds like it's going to turn some heads. So I found out that there was, well, it was May, so it's not like I was going to be able to leave because I had to wait a year, but I didn't want to wait a year. So a friend of mine who was a student in Rochester said, oh, you know, we only have an incoming class of nine people. And one of them just said, no, why don't you go and take cheers and Teuffalls and all that stuff that us aliens have to take to enter an American university. So yeah, so basically the story is that I ended up in Rochester, upstate New York,

where instead of computer networks that I knew, I did computer architecture. And then after nine months there, I transferred to Wisconsin, where I started working on databases, a subject that I also had worked on while in Crete. And I ended up becoming an interdisciplinary student between databases and computer architecture. And then I got into the magical world of scientific applications, so started working on data management for scientific applications. Finished started working on data management for scientific applications. Finished my PhD in Wisconsin and then went for a short interview round and decided to join Carnegie Mellon University as an assistant professor.

Spent eight wonderful years there. As associate professor, I took a sabbatical in 2007. I took a sabbatical to Switzerland, to EPFL. That was an interesting choice. It was mostly because, well, it was closer to Greece, so I could spend a year in Europe after so many years in the United States. But then I thought, you know, that's not going to be a place I'm ever going to move to. I mean, first of all, I'm Greek, loud and animated.

This is Switzerland. And second, at Carnegie Mellon, one thinks like if they leave Carnegie Mellon, they're going to fall off the face of the earth, right? So now it's like, what is this new school? leave Carnegie Mellon, they're going to fall off the face of the earth. Right? So now it's like, what is this new school? It's neither here nor there, whatever. So I came here and it was a fabulous surprise. The school is great.

The people are fantastic. Students are excellent. There's research money. It's fantastic. So I started thinking very, very seriously about moving back to Europe. And well, 17 years later, here I am. That's fantastic. Yeah.

It's really interesting as you were kind of telling us about your journey there, about how those sort of small interactions or small conversations can end up having such an outsized impact on your life, right? Like it's wild, like the sliding doors moments, you just think, well, if that person hadn't been next to me and told me, oh, have you seen this? Then you maybe could have gone a completely different trajectory. But yeah, it's always interesting to hear how someone's arrived at where they're at today. But yeah, I bet that is a bit of a coach clash also as well between sort of like the way life's lived in America, right? And the Swiss lifestyle is that there's a little bit of a, they're

not the exact same, right? But yeah. No, no, they're very different. To be, if I'm honest though, I really, I really think for the first thing you said, I really think that part of the blessing to have options comes from responding to challenge. For me, that is in retrospect, right? I didn't know that then. I was just doing that. I was just thinking, okay, I'll go there now. How's that? What's going to happen in five years? I don't know. But I'm going to go there now. That's all I know, because I'm going to regret it if I don't. So that's how it works in my mind. But responding to challenge, it turns out it opens doors for you. Right. So as to the lifestyle and, you know, how life is in Switzerland and in the US, in Switzerland, the lifestyle

is what it is. In the US, the lifestyle is very different depending on where you are, where you live. In Wisconsin, it's a beautiful liberal town in the Midwest filled with amazingly, genuinely nice people. In Pittsburgh, it's a steel city, fast, made of iron, exciting with depth history. Well, a Greek talking about history. It's very nice. It's motivated. And Carnegie Mellon was my haven. It was very nice. It's, it's motivated. And Carnegie Mellon was my haven.

It was really nice. And now EPFL is my haven. It's really nice. But, um, all I'm trying to say here is that the lifestyle is a, you know, the social part way of living is just one of the, on the parameters. There's the, there is a professional life. The people who surround you there are very important, have always been very important to me.

And then the family, the children, all of this together. So it's a mixture of all. I never left the place because I didn't like it there. I moved just because I needed the change because I wanted to discover something different. Yeah. Now, yeah, new challenge here for sure. I'm sure we'll, to a touch on the people, um, and later on in the podcast, I've got a few questions around then to start digging into that some more, but, um, before we do that, let's talk about

what's going on in your latest haven at the moment at EPFL. So, yeah, tell the listener what you're currently working on, what the various projects you've got going on at your lab at the moment. So, yeah, well, that's going to take a little bit. I always describe my research agenda as, schizophrenic version of a computer scientist. I think that I'm very lucky to be working in data management. It's never going to be short of excitement, of pure excitement.

Data is the source of knowledge and information. When I discovered it, it's the topic of science that I never left from. I branched out to different other disciplines, you know, and did interdisciplinary work, but always kept one hand close to data. And it seems to me that, so in the very beginning, I got excited as my, and this, what happened, what became my PhD topic, I got excited about data and their relationship with computer microarchitecture. Back in the day, we're talking about the late 90s.

You have the database system, which is a program that actually accesses the data, right? That its job is to bring the data close to the application that needs it in order to process it and make decisions and then provide some of the building blocks of the processing. Right. But then there was the operating system, which is traditionally the one bridge that one has to take to go from the application to the hardware.

And then there was the microarchitect architecture, which was this cloudy, very blurred motion of the machine and all the components that's there. But it's the operating system's job to know that. It's not the database system's job to know that. Well, what dawned on me while reading work from other people who worried about performance and the relationship with the different parts of the computational stack was that when you really want to optimize use of resources to make sure that your money, that you bought extra memory, you bought extra processors, you bought a fast program, they went to the

right place, you really need to look at how all the components of your computational stack interact with each other. And the thing that first dawned on me was that the database system's performance was very much related to how well we were using the processing and memory subsystem of our hardware, which was a very novel thought at the time. Now everybody worries about architecture conscious, cache conscious, memory conscious algorithms. But back then it wasn't the case. Only a few of us were worrying about that. So that was and is to this day, a major part of our work in the lab.

The database systems related to the underlying hardware and storage devices. So storage enters as a big player, of course, because the storage hierarchy is the lifeline, the blood that the database system needs in order to draw the data and make the processing happen. So that's the first part of my work. I'm kind of doing this high level so that we can move on later to the parts that are really important. So that's the one part. The other part, and then you will know why I'm talking about schizophrenia, has been the relationship between the database system and the applications.

So, and mostly the scientific applications. Initially, to be honest, because that's what we had access to from the university, right? It's not easy. Well, it's actually impossible most of the time to get any your hands on an proprietary software that's used or even more on customer workloads. But for scientific applications, that's not true. Plus, the scientific applications are the most desperate for performance because they don't have that much money. Nobody pays attention. So I started working with

soil science applications in Wisconsin. Then I worked with astronomers. I worked with earthquake simulation applications. And then when I came to EPFL, I worked a lot with neuroscience applications. Ever since the beginning, the problems were twofold. One is problems, one had to do, one class of problems was how to organize data, observational data. So think of a telescope scanning the sky and dumping the data on disks. How to organize that data so that it's faster to run queries on them, to analyze them. And then, or then simulation data, how to again organize it, but it's a different problem because that data is more like, it's more, it's derived from a different, from a program, from a simulation program.

It's not observed data, right? So we can tweak them. We can do different things with them, how to index them, how to make them available again to run scientific applications on top of them, visualizations, what have you. And when I came to BFL, one of the fantastic experiences I had was that there was this group of neuroscience and neuromorphic computing people and clinical neuroscientists who invited me to join them in writing a proposal which resulted in a 10-year effort called the Human Brain Project. I was one of the people who wrote the proposal, went to the EU, defended the proposal, and

then we got this amazing opportunity. So we built a number of different things with that team. We also built, among other things, we built a federated network of hospitals, a medical platform that could be used by Federated Network of Hospitals in Europe. That platform can be used to ask questions about aggregate behavior of adequate health information of patients in different hospitals.

I say aggregate because you don't get specific data. You get the numbers, you get averages and counts and things. And then you can use that information to create and you bind clinical information with biological information and you can create disease signatures which allow us to leap ahead really in personalized medicine, in precision medicine for different diseases. So that's the other side. That's not introspective computer science. That's more like computer science and applications.

And I went that far to show you how far that can go. And it went even further than that. What I'm thinking about these days, I'm thinking about automation in a very broad sense. For me, there's always been a question about when to do all of the work we need to do. So when you need to prepare a system for efficient execution, when do you do that? In databases, we traditionally load data, ingest data, put data in a data lake. You can call it anything you want. It's preparation. It's restructuring of the data so that it fits the specifications of the program that's going to use them.

And it's only fair to think that if you do that, if you restructure the data that fits the program, that fits the way that the data will be accessed by the program, then the program will execute the queries faster. Weather always bothered me because first of all, the queries are a lot and different and I may not know which ones they are. So the access patterns on the data may be very different. And also, I don't want to do work ahead of time for data that I might not use. And I'm a natural procrastinator. And I felt really loudly about myself until I read books about structured procrastination, which are there to make you feel good about yourself.

So, like, wait a minute, this is actually good. I am doing the right thing. Of course, I'm a procrastinator. At the very last minute, I have all the data. I have all the information. So why not make my decision the last minute? So I took that philosophy and put it squarely onto how I design systems. Enter in 2015, enter RoLabs.

So the company is based on the premise that there is no database and there is no database system. The world starts with asking a question. When you ask a question, I go out, I find where... I mean, you tell me which sources you want to be used for this question to be answered, but these sources are like your Dropbox or your Excel sheet or your other database system or whatnot. So I go discover the data. I go discover the ways to access that data, build an access path through that data, and then with these access paths, bring the data to processor and memory, and then create a tiny little database system, code-generated

really, which allows me to execute the query right then and there. And yes, that takes a much longer time than if you had prepared all the data in the world so that it can be used to answer that query. But I don't ask queries completely unrelated to that. So the second query is going to use some of the information that I gathered in the first one, both data but also code that I generated. And then the third query more and more. So I'm simplifying a concept that back then it was a bit out of this world. People thought I was a little crazy, but today it's something that a lot of people do.

Generation is very, you see it everywhere in systems. And yeah, it's expensive and we still haven't gotten it exactly right, but serverless computing is very important and it's very efficient and it's actually very easy to do if we can combine it with just-in-time code generation and data virtualization in this way. So what I'm thinking about now is that the three-dimensional world that I live in, which is data, hardware, and applications, workloads, has been an ever-changing one. Data is of a ton of different formats. Hardware is of a ton of different kinds and with a lot of different parts that are important.

And then workloads are very different, transactional, analytical and all that. And AI has come in and has smashed everything. So AI has come to the workloads and said, you thought it was only transactions and queries, here's machine learning. Bam! Nine percent is that. And then you've got data. You thought the data was like of 50 kinds?

No, it's 500 kinds. And you can't even enumerate them. So just quit trying. And then, you know, the hardware, you thought that hardware would just compute and memory? Well, yeah, you still can make either computer or memory with transistors, but I'm going to make accelerators and your chips are never going to look alike. Anything you know, and they're never going to look one like the other. You're going to have 10 computers. Each one of them will have 10 totally different layout microarchitectures.

And there you go. So that's what I'm thinking about now, how to make sense of all this mess here. And I'm really happy I'm a procrastinator by the way. Oh, that's awesome. I mean, yeah, I guess when you think of it in those terms and the sort of three different dimensions, the amount of heterogeneity we've, and I guess that's just been absolutely like, it's a smash and accelerated massively by the AI and the sort of, I guess the last two years in particular, probably.

When did ChatGPT become a thing? Was it like 2022? When it sort of came on the scene and just, just, yeah, took a different, yeah, made it a different ball game completely. Um, but yeah, I mean, I really like the sort of, uh, both ends of the spectrum. It has to be kind of, you've kind of been down to the hardware and sort of, how do you design those things in a, in a, in a, in a sympathetic way to kind of get more juice out of your system to get more.

And then obviously the other side as well, the application side as well. So it's, um, and that kind of explains why. So when I was actually doing some, like, do my homework for the, for this, for this child, I was obviously having to flick through your, your, your Google scholar, just to get a look at some of the, the top side. I think. And the number one was a paper from cell about the reconstruction simulation of the occult.

And I was like, is that Google scholar played a trick on me? But now it makes sense. Right. And so, yeah, yeah, that's my claim to fame. I have a cell paper with another 150 people. That's okay. My algorithm is in there though.

That's awesome. Cool. So yeah, let's, let's jump off and talk about a little bit, because obviously this, this podcast is about impact and whatnot. And kind of, I think the, your Wikipedia page says you've been, you've done over 200 peer review journals so when you look back over that sort of body of work what jumps out has been the thing you're most proud of? You mean of my work? Yeah of all the various projects and the work

you've done what you're most proud of and then I guess does that necessarily correlate with the thing that's had the most impact? Um, may I give you two answers? You can give me as many as you want. Yeah. Okay. So my, the top professional achievement I'm most proud of, and here I'm that serious is that, um, and, and please keep in mind that I am an educator. I have chosen.

The top achievement most proud of my students. Add i cannot begin to describe how much more important. This is in my mind than anything else. that I have won an award for or received much, much valued recognition for. My students are the people who made all of this happen, have shaped who I am in terms of research. Otherwise, I would be all over the place and I would still try to make sense of the next thing that I see on the street or anywhere. They're really the people who are 30 people, and they're all systems people.

No one is similar to their peers. Each one is different. Working with each one of them for their PhD has been a different journey, one that I value a lot. And I have come to believe that as educators, our job is to take raw talent, which is our students, and turn it into happiness for them. So teach them how to take what they have, their smarts, their hardworking personality, their everything that they have as people, as professionals, and make a life out of it.

And I'm really proud of them. and technologically, if you will, speaking. I have to say that the system that we built here at EPFL, it's very difficult to choose. I have to tell you, it's very difficult. Like picking your favorite children, right? Yes. Because now, say this and I'm thinking, oh, what about the staged systems that we built? I was at CMU, my first project. At CMU, my first, first, first project was to build a database system, which, well, surprise,

I built a database system, which, however, was not the classical, like, monolithic database system. It was a bunch of little, what you could call today microservers. That's what people call today. So, you know, but back then in 2000, what I wanted to do was build like an engine, a query engine that was composed by servers, like tiny little servers. And each server could do only one job. So the epitome of specialization, right?

So one server could only scan a file, another server could only join two files and two inputs, another server could only do sorting, another server could only run an aggregate, right, do average. And each one would receive either one queue or two queues of packets and would spit out another packet wherever those packets, you know, the packets they incoming would have the information where to put it. And he would forget about the rest. So that's like the nightmare of every query optimizer.

So I would, and we built this system and it was fantastic because all of a sudden it was great too. It was really easy to debug. You could synthesize pretty much anything performance wise. It was great because imagine the locality. Imagine how well it used the cache subsystem because it only had to do one job every n milliseconds, right? So it's really, really, it was really good. Unfortunately, it's really hard to teach programmers to think this way. So it's very, so it kind of, I mean, a lot of people think it's, you can think of it as pipeline programming. A lot of people thought it was a great idea and some actually did apply it to their system as well or adopted similar philosophies. But it was great. So that

system I actually really liked. And then, I wonder if I ask any of the AI programs today to write me such a system. I'm pretty sure those damn things they will. But then the system that I'm really proud of is the one that we built around 2012 and then it became Rollabs 2015, which was the system that essentially got rid of databases and got rid of pre-coded database systems and used monoid comprehensions, which is mathematics essentially, right, category theory, to describe the different kinds of data and then describe how a code-generated infrastructure could, you know, code-generate processing on that data. And that's how the system worked. And I think this going back to mathematics in my mind as the only source of truth,

because mathematics don't have performance problems. Mathematics don't have scalability problems. They could care less about your limited computers because where we are limited mathematics hits you with this, you know, eternity. And then it's done. It's all, it's done. It's infinite. Yeah.

Infinity. Infinity is your answer right here. I got infinity. What do you got? Nothing. Yeah. Right. So with mathematics, I think we were able to wrap our mind around solving this, oh my

God, what query will the user, you know, am I going to be able to optimize it? We go around it in the most natural and beautiful way. And this is the system that I'm really, really proud of because it worked not effortlessly, obviously. I mean, what was the hardest was the commercialization of this system and still is. Because most people were not really,

you know, it's really hard to sell a technical achievement to, and everyone who's done a startup knows this, it's really hard to sell a technical achievement to people who have a today problem. Yeah, they are. So, um, so, but that's a system that I'm really proud of. I think it's, it was very, it's very, it's still is very, very innovative, very potent and, uh, Jenny Yeai actually did me a big favor.

I was about to say when you were talking about that initially, I kind of, I was thinking about the, the maps very similar to the word generosity, but I kind of went, I was thinking about it that way anyway. And so I guess that was a kind of help for like a PR boost almost. It made those calls with people a little bit easier, I guess, because there is such a departure from where they were, what they're used to at the moment. It's, I guess, a little bit closer, right?

Yeah. And you know what's funny is right before chat GPT became available and changed the world, that previous summer, I was giving talks telling people, we need to stop coding. We need to start thinking mathematics because that's how, that's expressive, right? Coding is a shoehorned thing that we use to talk to the machine, but we should be talking to the machine. We're human beings. Okay, we spent 70 years of coding. Fine, we had fun. Now, we're human beings. Okay, we spent 70 years of coding. Fine. You know, we had fun. Now let's get serious. Let's go back to mathematics, which are very expressive

and let's express ourselves like that and let the machines figure it out because we have compilers that code generate. So why not just do that? Now we don't even have to do the mathematics part. You know, we can just talk to the machine in natural language. How wonderful is this? And I think using natural language and, you know, making sure that we can use it for good, but use it in a way that's constructive, not cheating, not to fool people when we came up with something when we didn't and all that. Using natural language to really expand, not limit ourselves, right, is the new set of

skills that we need to teach the young people. Yeah, I completely agree with that message. And I kind of, I want to use this, this, this, this, this last discussion is there to jump off into a question about, and I always like to ask about bridging the gap between academia and, and industry. And obviously you've kind of experienced both. Now you alluded to sort of the startup experience a second ago. And so I kind of wanted to get what your current take is on the interaction between

industry and academia and I mean, raw labs is an example of those two things, I guess, working together as they should do, right? Like an idea comes from research and it maybe jumps over into industry. So yeah, so what's your take on that interaction between the two camps at the moment? reaction between the two camps at the moment? Um, in, in database systems, um, industry and academia has historically been a good five to 10 years apart. Meaning that academia would produce research results and if those were relevant to industry, that would happen five to 10 years later, I mean, they,

they internalization with product. Computer architecture has it worse. So I guess we should be happy because it's much easier, I mean, much cheaper also to change software than to change hardware, right? The investment is a lot more modest. But today, the relationship between computer science, academia, and industry is tighter than ever, I think, because essentially we need to listen a lot more attentively. We in academia need to listen a lot more attentively. We in academia need to listen a lot more attentively than before, not only because it's important

to but also because of necessity. I mean, funding, research funding, only comes around when the relevance factor is maximized. And the exponential increase in people has brought a beautiful exponential increase in data, in data, not in data, in data, but in ideas, in ideas, in scientific ideas, in industrial ideas. So academia and industry, I think, have never been this close. Adoption of ideas is also more immediate, I think, just because of the osmosis that

happens. However, we might be missing a little bit the scientific part. That might be a little oppressed because the relevance factor needs to be there very, very strong, stronger than ever. And we might be missing the research for the research part. I've always been fortunate to work on ideas that I found exciting. Somehow because I'm in systems, those ideas did find a home in at least discussions that I had with industry. Some of this didn't even find a home actually squarely in industrial product. But I am a systems researcher. People who are more on the theory side might be finding this, might have different views there.

I'm stricken on time, how we're doing. I've got two more questions if that's okay. Go ahead. Cool. Yeah. The first one, I think you maybe alluded to it a little bit earlier on, about sort of the thing you're most proud of was the people. And I kind of want to kind of maybe invert that question. Let me see. Who or what has had the biggest impact in the sense of what motivates you more? Is it the students or is it like any particular work or any individuals that have really shaped you and inspired you? Everything inspires me. It does. I know. I have to accept this. I mean, I know it's not a good thing. I know it makes you spread yourself thin at times, but I draw inspiration from a lot of

different dimensions of my life, of my everyday life, even from the most improbable things. I think I'm a bona fide engineer, is like that. And I always, I like, you know, making things, fixing things. Sometimes even with my hands. I built my son's computer after 30 years that I hadn't done that. Right before COVID started, he asked me for a computer. I thought, there's no way I'm going to spend this much money on a computer. And then I thought, come on, you used to do these things years ago in Greece, in the basement. Get your tools, build the computer. Well, you know, good news. You don't have to, you know, solder anymore. Just click stuff. Put it together, right? Yeah.

But you know, the kid ended up with a 12 core gaming system, right? Just in time for COVID as well, right? So I guess it got used quite a lot during the pandemic. Great. So yeah, good timing. In 2020, in February. So, so inspirate, and I was inspired. I was inspired that time. My mind was racing left and right. I think, um, keeping an open mind is extremely important, right? Don't limit yourself. I have a student who's working on LLMs and

I have another student who's working on concurrency control. There's nothing more beat up or worse, triggering yawns more than concurrency control, right? Not even queuing theory. Queuing theory is expensive and difficult, right? Concurrency control. But she wants to do concurrency control. Who am I to tell her no? Because when I wanted to do what I wanted to do, people thought I was crazy. So, you know, maybe she's on to something. we'll see. So, trust and let young people flourish. Just be the power. Don't pull them places. Yeah. And yeah.

And that's really nice. Yeah, that reminds me of a quote I quite, quite place about. Everyone, I can't remember who it's from, everyone thinks you're a madman until your ideas triumph, basically, I think is the quote. I think it translates from Spanish, so I'm not sure I've got it exactly correctly, but that's the rough quote, anyway. I think I had that at the start of the opening page of my thesis, I think. But yeah, anyway, cool, right? So yeah, just the last question now.

We've kind of mentioned a few kind of future looking things over the course of the chat, but yeah, I kind of want to get your take on the future going forward, obviously exciting advancements that you see and how that might impact the field of databases. Are we just going to be consumed by LLMs and generative AI or is it, what's that going to look like? What will the future look like? I guess is what I'm asking. You get the crystal ball out and let's see. We're having this conversation in 10 years time. What would it be talking about? I guess. I think that computer science is changing at a rate that none of us could fathom.

And there are many different futures that might happen. Many different people say, you know, go toward one direction or the others. It seems to me that most of them will happen. It's just because we can. It's because we can. Diversity today is so multid-dimensional. It used to be men and women, right? I used to be one of nine kids in a classroom with close to 150, 160 students. I was nine girls, 145 boys. And now that's just one dimension of the diversity we're going for.

It's diversity in, in not just in, you know, who we are physically, biologically, you know, in terms of background, but how we think. And that's the biggest power. The Gen. AI, I'm gonna say it again, it's one of the best things that ever happened to us. All of us, researchers, engineers, scientists, and the world, and it's a huge power. And yes, we are gonna use it wrong.

We have used every single new discovery wrong before we use it right. But it's a learning curve. So how what the future will look like scientifically, which is what I'm strongest, I think I'm fittest to talk about. Scientifically, I think we will be faced with even more difficult problems to solve. But as computer scientists, we have an edge. The edge is that we tame complexity through abstraction.

This is our power. So the more the complexity, we apply more abstraction, we go along all the dimensions of the abstraction, we find the right subset of dimensions that can be most promising to solve a problem, and then we carefully see what level of abstraction we need to go to in those dimensions. And then we solve the problem and we have solved all the problems to date this way. So the more the complexity in computer science, we say, bring it on. Yeah. I like that.

That's brilliant. What a message to end on. That's fantastic Anastasia. Yeah. Thank you very much for the chat today. It's been absolutely lovely. And I'm sure the listener will have found it very insightful as well. I'm sure about anything we've mentioned, I'll drop links to in the show notes so the listener can go and you can go and find those. And I think also use this opportunity

to plug two of your students' episodes. They came on the other episode, Hamish and Eleni. It's episode 51 and 22 if I've got my notes right. So yeah, listen to go check those out as well. You can go and learn some more about what's happening at EPFL. So yeah, thank you very much, Ernest Deger. That was awesome. It was a pleasure. Thanks for having me, Jack. Music