Sadie Bartholomew: Patterns in Computing and Art

SPEAKERS

Sarah Webb, Sadie Bartholomew

Sarah Webb  00:00

Later this month, the Supercomputing meeting, SC24, convenes in Atlanta. This year’s theme is HPC Creates and in this episode of Science in Parallel, we’ll be talking about the Art of HPC display, a special exhibit this year. I’m your host, Sarah Webb, and my guest is Sadie Bartholomew of the National Centre for Atmospheric Science in the United Kingdom and the University of Reading.

[Theme music plays.]

Sarah Webb  00:49

Along with her research software engineering work, Sadie creates digital art and submitted several works to the SC24 Art of HPC display. That special collection will be shown at the Georgia World Congress Center from November 19th through 21st. Sadie and I discussed the many patterns in her work within weather and climate, in coding and in digital art. She talks about the synergy and fulfillment she finds at the interface of computing and esthetic pursuits.

Sarah Webb  01:23

Sadie, it is great to have you on the podcast.

Sadie Bartholomew  01:26

Thank you so much, Sarah. I’m delighted to be here and being in conversation with you.

Sarah Webb  01:31

 I found you because you had these diverse interests at computing and software art and even music. And I wanted to find out what were you like as a kid? Did these interests start then?

Sadie Bartholomew  01:43

So I’ve always loved maths. I’ve always enjoyed art, so drawing and making things in general. When I was in first school, I remember that’s when, I guess, my love of maths began. I had a fantastic maths teacher called Mrs. McKay, who inspired me, and I also I was just, I guess, a creative kid drawing. I didn’t start painting until later in my life, but I enjoyed more traditional arts and crafts, I guess, when I was in middle school and in high school. This is the point where you have to start making choices about where you’re going in your, well, ultimately leading to your career, but where you’re going in your life. So there was a push and pull from the sciencey,  technical side and from the art side.

Sadie Bartholomew  02:24

Of course, when you get further along in life, you realize there’s a lot of creativity within the sciences and the technical sphere. But when you’re a bit younger, you might not realize that, and you may think it’s a more binary choice. But in the end, my family were very sciencey, so my dad was a medical physicist. My mother was a doctor until she had a career change. My grandfather on my mother’s side, an engineer. And I think partly because of that, I decided to take the scientific route. And I did A-levels, which are the higher part school qualifications in the British schooling system, in physics, maths, chemistry and further maths as well, which is an extra maths qualification. And I went on to study physics at university. But I always enjoyed in the background doing doing artistic things. So traditional drawings and paintings, landscapes, of portraits, it was quite realistic in terms of style back then. So later in life, I got into abstract art and even conceptual art, things like that. But at the time, it was more traditional based.

Sarah Webb  03:29

So let’s talk about your more technical work. First, talk about getting into math and computing. There are lots of directions you could have gone. How did you end up in computing overall?

Sadie Bartholomew  03:40

So I studied physics at Durham University. I did a four-year course. I’m not sure if you have the equivalent in the U.S., but this was what we call an integrated master’s. So it was a three-year standard undergrad, but with a master’s year tagged on to the end, so that it all comes under one qualification. And for my master’s year project, I specialized in particle physics phenomenology, which is kind of the area between the theory and the experimental aspect. So it’s about the interpretation and understanding of experiments, so say events in a particle collision, let’s say the Large Hadron Collider. But I found this really interesting, but it was very steep learning curve, I have to say. So I this is my first go at using C++ because previously, I’d only used a bit of Python. So I’d done some computing within my undergrad level of the integrated masters. But it wasn’t a big part of it, actually. So it wasn’t into until after university, really, I got into programming and computing in a more involved way, and I learned to really love it. So I was learning C++, but Terminal in Linux, that it was also something I learned in that year. And also with the phenomenology, there’s the experimental side that you have to know a bit about, and the theory side, which have to know a bit about. So doing all of that in a year was a really steep learning curve, and I did enjoy it, and I learned a lot, but it was quite intense.

Sadie Bartholomew  04:57

And I guess I was a bit disillusioned, but not just from the master’s year, but from the undergrad component by the academic environment. So I noted, and I had a lot of friends that were quite disillusioned as well by the kind of exam- and rote-based learning and the competitive atmosphere. Some of my friends went to do Ph.D.’s, but others we wanted to go into industry or find a role where we could use our technical and scientific skills outside of academics, basically. So I’d applied for a job as a scientific software engineer at the Met Office, which is the UK’s operational weather service. I stayed there for two years, and that’s really when I first got into computing. So, for example, that’s the first time I learned about testing and documenting code and using version control. And when I first did things like user support and user training, and I really, really enjoyed that, actually, and I learned to really enjoy the domain: so weather and climate, which wasn’t something I ever actively targeted, but I came to realize that it was really, really exciting.

Sadie Bartholomew  05:57

So specifically, at the Met Office, I worked on a workflow engine called Silk. It was designed especially for operational numerical weather forecasting, and it has a kind of a unique selling point, a USP, of being able to manage infinite cycling workflows, because there’s actually very complex periodic job execution requirements within weather and climate. So I stayed there for two years, learning a lot about software engineering and not so much, but still plenty, about weather and climate as a field.

Sadie Bartholomew  06:27

After two years, I felt like I’d like another challenge, so I accepted a job at the National Centre for Atmospheric Science and the University of Reading. I’ve now been there for coming up to five years. Day to day, I tend to develop and maintain tools which ultimately help the researchers do their research so in earth science in general, with a particular focus, I say, on climate science and atmospheric sciences, given my host institutions, but more generally as well, and within that, I also provide user support training, heavily involved with the CF Conventions Community, which is a metadata standard within earth sciences, particularly focusing on the net, CDF data format. But it’s also more generally applicable.

Sarah Webb  07:07

But Sadie’s technical work, particularly using the matplotlib visualization tool in Python, led her in artistic directions.

Sadie Bartholomew  07:16

In terms of the art side. this is something I tend to do nowadays, outside of work. But since I was at the Met Office, I’d been playing around with matplotlib, which is one of the main visualization tools for Python. And I’d realized that it’s really fun to play around and explore it as a tool for producing esthetic work. So not just obviously plots that it produces for research and for that you see in publications, but also just as a tool to explore different elements of initially mathematical functions. I did some contour plots on some gradient backgrounds of basic trigonometric functions multiplied together or deducing a factor, basically just putting random through the metric functions into the code and seeing what it would produce and zooming in on any interesting outcomes. But I went on to develop that with various different mini projects that focus on different, I guess, techniques or processes or inspired by some work by another artist, say, and it’s become quite mature.

Sadie Bartholomew  08:13

Now there’s about 10 mini projects in there which are all using pure matplotlib code. So effectively, Python code using matplotlib, but with different aspects shown and showcased. It could just be a one line to define a mathematical function and then several lines just to set up the matplotlib canvas and then plot it. It’s obviously interesting and important to show scientific plots, but it’s also quite interesting from, I guess a curious perspective is to see if I just change a little thing in the mathematical definition, which would ultimately completely change the XY plane of the contours that it shows, to see what behavior we could see. And there’s some really interesting and quite intricate details on that.

Sarah Webb  08:56

The SC24 session was what brought us together to have this conversation. How did you learn about the session, and how did you think about which pieces you might want to contribute?

Sadie Bartholomew  09:10

I was aware of it as a general conference in the background of the work I do so my team, which is the computational modeling services group within NCAS, which is the shorthand we use for the National Centre for Atmospheric Science, where I work. In our team, we do various work within computational science and high-performance computing. So some people work on aspects of high-performance computing management. So they work with Fortran codes and doing kind of parallel work, such as using MPI, open MP, for example. So there’s quite diverse work there, but within CMS and within the whole of NCAS, actually, we do some quite diverse computational science work and research software engineering, which is a term I like to use because the quite proud that the UK, where I’m from, kind of pioneered the research software engineering movement.

Sadie Bartholomew  10:02

But across NCAS, we have a workspace Slack, particularly because we are embedded in different universities across the UK. But in the Slack group, one of the channels we have, which is a social and a nice networking group called Makers Meet, which someone had set up about a year ago. And the idea of this is that we can all share creative things we do in our spare time: crafts and arts that we make. And it’s some really diverse and interesting things that people do across NCAS. So some people make cards, some people do crochet, some people make ceramics. This is a really interesting, lovely mix, and a colleague of mine called Poppy Townsend had made a post saying, so there’s this Supercomputing Conference, and they’re having this session called Art of HPC, and it’s basically just a broadcast to people to, I guess, advertise that that was on, in case anyone was interested. If that hadn’t been shared, I probably wouldn’t have known about the Art of HPC session. So I’m very grateful to Poppy and to those who have set up that little Slack group.

Sarah Webb  11:04

I then asked Sadie to talk about two pieces that she contributed to the SC24 Art of HPC session. Those images are in the show notes at scienceinparallel.org and we have links to GitHub, where you can see more of Sadie’s mini projects and her code. She’ll talk about the role that the UK’s JASMIN supercomputer played in her design process.

Sarah Webb  11:28

Let’s start with that Collatz Kaleidoscope for our listeners. It’s this beautiful swirl of greens and reds and oranges. Talk about where this started for you, how this evolved. How do you tweak things over time to get the visual effect you want?

Sadie Bartholomew  11:47

So this Collatz design is one of two that I’ve submitted to Supercomputing 24, and the name Collatz refers to the fact it’s based upon this Collatz conjecture, which is quite a simple rule, or a pattern for mathematics, whereby based on starting with an integer and applying some rules based on whether it’s even or odd. So, for example, it’s also referred to as the 3n + 1 conjecture, because that’s one of the operations that you might apply. But you start with a given number, and after a certain amount of iterations, you’d expect it to return to one, and there’s actually an open hypothesis as to whether it will always return to one or not. I’m not sure it’s been proved. It’s an interesting mathematical topic in itself, but what particularly interested me when I was reading up about this was that this you can do a plot of the number of iterations it takes for the given number to have these operations applied to it, and in all the cases they’ve ever found or converged to one, and you plot the amount of iterations it takes to get to one, and you get this scatter plot, which has got a really interesting pattern, but I’d say it’s quite beautiful.

Sadie Bartholomew  12:53

So in my creative matplotlib repository, one of the many projects I did was basically just to plot that, but to zoom in on different areas of the ultimate plot that you get to see what it looked like. And I started playing around with different markers, so simply by changing the marker you’re using to plot it. And I realized that plotting some of these in an oversized format so that they overlap with very low transparency would create this really nice sort of blend of colors. Because what I tend to do with these create a matplotlib projects is create some code that encodes some sort of process. So in this case, the code is very short script. The process is basically just plot this scatter plot of the Collatz conjecture iteration times, and then apply some parameters to change such aspects as the marker size, the marker shape, what area of the plot you’re focusing in on, on the colors. In terms of the variables that I change, the colors are something that really interests me, because, like, color can really bring something alive or it can really change the tone of something. So I really enjoy thinking of color schemes.

Sadie Bartholomew  13:57

But yeah, so in this case, and as with a similar design that I’ve submitted, I’m using very vibrant colors, and I think it creates this really nice blend. I think when I was originally looking for the final designs to choose, because I use this process whereby I use, well, technically, a super computer, so JASMIN, which is a supercomputer that we have for environmental data. Obviously, this isn’t really an environmental project, but I got kind permission from someone in my group to use that, ultimately, as an educational exercise on on using Slurm. I was using job arrays, which is something I’ve not done before, basically, to produce various outputs. So I’d get 30 different outputs, and some of them, I’d basically eyeball them and see what looked visually nice. And some of them were a mess. But average, it was maybe about five that looked quite promising. So I’d look at those five, and I’d take those on the parameter sets there, and then I’d start to kind of hone in on those and tweak the parameters a little bit. Again, do another iteration where the parameters are changing a little bit, but from the more promising ones, and then do a certain amount of iterations until I got to a point where I would think, I was thinking: This is a design that looks really interesting. I’d end by doing basically just manually tweaking these parameters. When I just thought, I think I’m nearly there. I’m really happy. And these were the results. These were my favorites.

Sarah Webb  15:12

What I found so interesting looking at them on the screen is that when you zoom in and you zoom out, the effect of distance is striking.

Sadie Bartholomew  15:21

That’s great to hear that because sometimes when we create art and we being anyone, you can really appreciate yourself, but you never know what other people might think. And obviously everything’s subjective. I’m interested to see or hear, I guess, because unfortunately, I can’t go to Supercomputing in November myself. I’m guessing there might be some big screen which these designs get displayed on, or potentially they might be printed out, and what effect that might have on the overall look and feel.

Sarah Webb  15:48

I want to talk about this other piece of Connections in Rotation, because it’s definitely another one that has a different effect when it’s zoomed in versus further out. And can you talk a little bit about how you made this one?

Sadie Bartholomew  16:00

This is based on another mini project from my creative matplotlib repository. It’s inspired by an Argentinian artist called Julio Le Parc because he had done some fantastic works, and this was before the advent of computers. So this was all done manually, at least as far as I’m aware, his process was to take a certain basic design, in his case, like a wedge from a circle and place that in a grid. But in terms of the whole grid, there would be some rotational angles at which the start and end point would be rotated. In terms of the wedge and going across the rows and the columns, the rotation would be evenly spaced to get between the start and the end point, which produces, ultimately, a grid of regularly rotated wedges. And some of his designs, there were multiple ones, and so he’d use this sort of technique to create various different designs.

Sadie Bartholomew  16:55

They’re quite intriguing, really, because they’re completely static. It’s a 2D traditional piece of art. There’s nothing moving, there’s nothing interactive, but it just gives this feeling of motion in terms of the rotation it has. But when I did look at it on the screen, I kind of scrolled down, and it’s almost like the motion that might it almost tricked my eyes into the fact it was moving. It’s quite clever in that respect. So I took that basic concept and other things that can be changed in this case are the array of angles, so in particular, that the start and end point across each row and how the first and last column are also spaced, because that is also regular, and that also determines, in turn, the spacing across the rows. And also the particular patches, what I call them in my mind, because that’s the matplotlib language for the basic design that’s being rotated each time.

Sadie Bartholomew  17:46

So for this piece of Connections in Rotation, they actually use some quite thin wedges. So imagine a circle with another quite big circle taken out. So like it was a symmetry, but there’s also a break in symmetry across the different angles, or at least in the designs that I converged on as being my favorite. Because like with the Collatz works, I emerged with whole series of different images based on the parameters of the rotation angles and the different basic shapes that I could start with. But this kind of emerged as one of my favorites, really. But one thing I did with Le Parc’s, one of his original work, I actually animated it, because you can do that with matplotlib quite easily. So with the patches, I just made them rotate at a regular speed. But I found I was actually a bit disappointed in that, because when the actual movement is happening, it’s maybe a bit less special than when it’s just purely static. But you almost feel like there’s a movement that is almost frozen, or even if you’re moving your head around, there’s a movement that your eyes are tricked into seeing.

Sadie Bartholomew  18:45

I just think it’s particularly genius that the artists such as Le Parc had done this sort of thing before the advent of computers, where we can quite easily set up code which will do this regular spacing of angles. And he must have very painstakingly got a ruler and a compass and created all these wedges and very regular spacing. It must have taken a lot of work and time, as opposed to some code, which can quite simply recreate that. I don’t think you should judge an artwork by how easy it might be to recreate. I think that’s quite a dangerous sentiment. I think there’s a danger that when it’s so easy to set something up, that things can become a bit cold or mechanical, and people could create, like, 500 different designs in the space of 10 minutes, and maybe not even see them, but then call them their own artwork. And I think it’s important to have that human touch where you actually evaluate it in your own eyes and think, do I like this? Is it nice? It doesn’t have to make you feel something like a lot of traditional art does, but is it something that you’re happy to put out there as something of interest? I do like the idea of, like I’ve done with my works. I’ve actually, I generated a lot of designs, but I threw away a lot of them because they just didn’t feel interesting, or some of them were a bit of a mess. Some of them just looked ridiculous. And I was like, well, that’s just looks like a plot that’s gone wrong,

Sarah Webb  19:15

I wanted to ask you, in general, how you think about the role of the computer in your artwork? Clearly, it’s a tool, yes, but is it inspiration? Is it collaborator? Is it something else?

Sadie Bartholomew  20:15

No, I guess I’ve always thought of it as a medium, but I’ve never really challenged that idea in my mind because, because it’s a good point, you could consider it to be a collaborator, as opposed to just a tool. So I always thought of it as just a medium. So for example, you could do a line drawing just by hand, and that becomes the medium of a line drawing. Whereas with creative coding, I guess the medium is the computer, or particularly a certain programming language which you’ve used, and then it comes under the realm of digital art, but it certainly can be something that’s inspiring. So for example, in the whole process of generating code to, say, replicate a technique. So for example, with Le Parc, works to create those grids of patches and create the rotations in doing so, the whole act of implementing that provides ideas in itself. So from there, I realized, for example, instead of doing just one patch per grid point, and I could do as many as I wanted.

Sadie Bartholomew  21:10

Obviously, with coding, I just set up 10 and it to plot 10 instead of one. If I wanted to do 10 or I could change the grid. I don’t have to do a regularly spaced grid. I could set up something in a more interesting geometry. So my computational brain allows me to think of ideas that I might not have thought of if I was, say, just doing that via a very traditional painting process, quite painstakingly like Le Parc. So I guess, yeah, there’s an inspirational aspect to the computer. Yeah, I would say, collaborate. It could work. But I think with computers, obviously they don’t really talk back. You tell them what to do, and they will do what you tell them to, not always in the way you want, because obviously they’re very strict. Sometimes you tell a computer to do something, and obviously, because they’re so logically bound, you might realize that what you asked it to do is not what you wanted at all, which is quite fun.

Sadie Bartholomew  21:58

It’s a part of being a computational scientist or just anyone, I guess, in general, who works with computers. You have to learn that it takes instructions very directly. But I think it is an interesting role being somebody who’s trying to do art via a computer. I realize there’s a direction in terms of AI as a creator of art, and does that count as a computer having more of a firsthand role in the creation of the art. I find it quite an interesting subject. I don’t tend to use AI tools, although I know there’s lots of fantastic work being done in the climate and weather field with such tools. But also, I know you can create pieces with the art, and I do. I guess I worry, because on the topic of creativity, because computers, or specifically AI. If an AI is trained on data sets that are, say, quite samey or consistent in a certain way, then the outputs will be, I guess, quite similar, which I think, to me, highlights the importance of creativity. If something’s a bit different or pushes the boundaries or has taken something in a different direction, then that could mean that the outputs of, say, some sort of AI tool will be more broad and more diverse.

Sarah Webb  23:05

We’re talking about the role of the computer in your artwork. And I guess I’m wondering, having matplotlib and some of these ideas into a creative space, does that influence your technical work at all, in a day-to-day way, or how you think about problems?

Sadie Bartholomew  23:23

Oh, absolutely. One thing I’ve come to realize, and it’s been, I guess it’s been a bit of a surprise, but it’s been really important to me, in terms of my career fulfillment, and how happy I am in my career, is that doing kind of software engineering and computational science, it’s a really creative job to be in. So, from the act of solving a complex problem or even something like debugging, I think you really have to use a lot of creative skills for that. And maybe it’s not seen as a traditional creative medium, but actually it’s fantastically creative and in a collaborative way, because it’s such a collaborative field. So the domain of atmospheric and climate and weather. It’s really important that various different teams work together, and say, RSEs and researchers, high performance computing experts, but also software engineering in itself.

Sadie Bartholomew  24:12

So they’re very creative domain, and the thinking that I use when I’m developing these pieces with my more esthetic brain engaged, it’s really helped me in terms of, say, trying to think a bit more originally, or trying to stray off the beaten path a bit. So, say I’m debugging when something’s not behaving in the way that I expect it to, and I just can’t work out why, it’s thinking about techniques, which might lead me to understand quite what’s happening. And usually it’s a very specific detail that leads me as a clue to realizing quite what’s going on. But I find it’s really helped to be able to think in different ways and engage different parts of my brain.

Sadie Bartholomew  24:50

And then there’s also the maybe more direct means of actually producing these pieces in matplotlib has made me a lot better at using matplotlib, which obviously is always useful in the research field. But certainly, I think the artistic side of me and the scientific side they’re synergetic. And it’s also really enjoyable for me, because when I was younger, I mentioned I was, you know, quite artistic, but I loved maths in particular, and I went on to study science, but not realizing that it could lead to a career where I’d be all really fulfilled in a creative sense as well as an artistic sense. So it’s been really, really brilliant to come to that. Been a bit of a journey, but to come to the point where I’ve realized that I’m using my creative mind as well in my job.

Sarah Webb  25:31

We’ve been talking about creativity in various ways. And so I want to bring it all back to, what does creativity mean to you, Sadie?

Sadie Bartholomew  25:42

I guess I’ll split it into, say, two components. So there’s the one in terms of creativity within computational science and within creative coding. And then thinking more generally as well. I’ll start with the specific case of in computational science and creative coding. I think to me creativity means joyous progress, actually, because I think creative thinking is essential to solving problems, particularly complex problems, be them, real world research problems, software challenges, anything technical. If you’re being creative, you’re moving away from the obvious or the mechanical, and that will allow progress, ultimately, not necessarily quick progress, but it’s speeding things up and making things more innovative and solving some of these very complicated challenges that we have in, I guess, any domain, but climate and weather is a particularly interesting one.

Sadie Bartholomew  26:36

In terms of creative coding. I think creativity is important because I worry that with creative coding, because it’s a computer-based process, that some of the designs could be construed as cold or mechanical, and that people might not engage with them in that sort of sense. But if you’re truly creative in that field, you can create something that feels very, not per se, organic, but it feels interesting. And it feels like it has structure, or there’s patterns or repetitions or themes there that are of interest that make you engaged, you’d have some sort of emotional response.

Sadie Bartholomew  27:14

And then, in a more general sense, I think creativity, to me, is just the utmost importance, really, because it makes the world interesting, and it challenges us, and it can bring us together as well. Because I think sometimes people can struggle to communicate, say a feeling or a state of mind or specific emotional response in logical ways. So through creativity and various mediums where you can express creativity, they can show more insight into where they are at.

Sarah Webb  27:45

Sadie, thank you so much. It’s been such a pleasure learning more about you and your work.

Sadie Bartholomew  27:50

Thank you, Sarah. I’ve really enjoyed it.

Sarah Webb  27:52

To learn more about Sadie Bartholomew, the organizations and artwork mentioned in this episode and creative coding, check out our show notes at scienceinparallel.org. We also have a link to a past interview with Joe Insley, whose work is also part of the Art of HPC display at SC24.

Sarah Webb  28:12

Science in Parallel is produced by the Krell Institute and is a media project of the Department of Energy Computational Science Graduate Fellowship program. Any opinions expressed are those of the speaker and not those of their employers, the Krell Institute or the U.S. Department of Energy. Our music is by Steve O’Reilly. This episode was written, produced and edited by me, Sarah Webb.

Transcribed using otter.ai and lightly copyedited.