In our second installment of our Glitch Lab Student Spotlights, we hear from Egemin Sahin ’23 about his work experimenting with and researching various A.I.s for our project.
Can you tell us a little more about yourself and what you do in the lab?
I’m Egemin, I’m a ’23 double majoring in Math and C.S., modified with Music. In the lab, I come up with ideas for what kinds of A.I.s we want to use, I do a lot of literature reviews on how we can implement these ideas into the computer, basically. Also I experiment with the A.I. itself after I put it on there. I’m more of an A.I. applicator, I guess you could call it that.
Can you tell us more about your process of working with the A.I. and anything new or surprising you might have learned over the course of this project?
Usually when I want to come up with a new idea or a new project, I’ll just think about what’s going on currently in my life that interests me and I’ll try to weave A.I. into that and I’ll start researching online about how to achieve this thing and go deep into the architecture of how the A.I. works, why it works, what kind of data it was trained on, what are the outputs usually look like, what are its downsides, what are its upsides, and after I’ve done that I install everything necessary which is usually what takes up the most time for the A.I. process, as there’s usually a lot of mismatching and outdated stuff like drivers which is usually the main thing that I work on. After I’ve got that ready, I just open it up to experimentation and I tell the Tiltfactor lab members to play on it and do creative things on it. And what I find really interesting about the whole process is when I first create an A.I., I’ll put in some things that I think are how it’s supposed to be, like how the input is supposed to be, and when I tell someone else about it they’ll completely put in something that’s a completely different format, completely different way of thinking and the output from the A.I. is vastly different just due to the input of our thoughts in this project, and I think it’s really interesting to see that kind of bias, not in a bad way but more so in a creative way. As much as the A.I. itself is seen as the artist here, it also matters a lot the data and also the input you give the A.I., which is pretty interesting.
What’s one goal that you have for this project that you hope to accomplish by the end of the term (or as a team)?
I think currently with this image, create image from text A.I., I think we’re at a pretty good place and I think I’ve achieved my goal of getting this A.I. to be able to create images from text. In the future, one of my goals for Tiltfactor is to implement more of music into this, we talked about it last term but we haven’t really done anything concrete with music, and that is alongside with computer science one of my greatest passions and I’d love to incorporate that into the A.I. and see what comes of it.