A Year in Review 2024

I haven't been as active as I would have liked this past year, and since there has been a bit of an exodus from X/Twitter, I thought this platform might be best. I wanted to share my past year to better synch up with my network and collaborators.

Winter

I moved to Norway at the beginning of the year to stary a PhD. I also did some wonderful collaboration with some of the radiology staff at Upstate University. I was looking at attacks in computer vision and seeing how they applied to radiology. Here is a link to some slides for a presentation: https://docs.google.com/presentation/d/1TAFXM-GQo5nhw1MV5_RvmF1V_H3wdu6-/edit#slide=id.p1

Spring

In May, I continued researching attacks in computer vision and gave a talk at a meet up in Oslo

https://drive.google.com/file/d/1TPrk5jn9OaNDE_ZL4Q1UloOHoMVVp3VG/view?usp=sharing

In early June, I attended a research school at Kristiansand. We had to present our research based on this course in the fall. https://docs.google.com/presentation/d/1gyDezlKXHyD3bj5d145xR728kgfODTQo-CgecJpwf-8/edit?usp=sharing

I thought this was particularly neat. This plot is highlighting what an AI thinks is important when trying to predict the age of bones. I thought it was interesting that it was picking up on the roundness of hands. But, it also thought that the marker left by the technician was also an important characteristic. Since the smaller and younger hands need a more of zoomed in perspective, the marker typically isn't included with younger hands.

Summer

I went back for summer vacation. This past summer I participated in this "Introductory Technical AI Safety" program with AI Safety Student Team at Harvard. I got a lot of out of participating and applied the new techniques to my research.

For example, I tried to find where hacking knowledge is stored in LLMs:

https://docs.google.com/presentation/d/1i4JhR2ADN3-UlCdxby2IV1Laez1KEv1I/edit?usp=sharing&ouid=113772468977296182174&rtpof=true&sd=true

This research is being presented at IEEE Big Data in December.

I went to a DefCon in August, it is a hacking convention. During the conference, I attended a really interesting workshop that used a table top game simulation. The framing of the situation was that the people were part of a group trying to hack into hospitals while other groups tried to defend. The point to take away was that it was important to have plans and policies in place, since it is difficult to make comprehensive strategies on the spot, and you don't know the intent of actors and how they are working together.

Fall

In October I presented at IEEE 4th Cyber Awareness and Research Symposium 2024. {https://drive.google.com/file/d/1Tl9JsqFm4xcJG8fxDPe_j-JMdi_VOLay/view?usp=sharing

I also visited Helsinki to attend "Nordic AI Meet." The conference included students from across the Nordics, and I learned a lot. There were some great speakers discussing food science to the implementations of employment related systems. Poster:

End of Year

I attended at MMIV 2024 in Bergen last week. I presented some of my on-going research in using reinforcement learning to segment images in adversarial settings when attackers are attempting to manipulate the input data.

I am presenting at IEEE Big Data in Washington, DC. I am presenting my research using traceability methods and large language models. https://docs.google.com/presentation/d/1i4JhR2ADN3-UlCdxby2IV1Laez1KEv1I/edit?usp=sharing&ouid=113772468977296182174&rtpof=true&sd=true

I am also presenting working on applied homomorphic encryption. People are excited about this variety of encryption, since you don't have decrypt your data to use it. I really liked this project: https://github.com/ibarrond/Pyfhel

https://drive.google.com/file/d/1kD4YIJnxVQaTVFW8SDa5LTQbGpXvsHCq/view?usp=sharing

With the upcoming year, I am planning on participating at a research school in northern Norway (Northern Lights Deep Learning). I am presenting research on the stickiness of sensitive information with LLMs by looking at how peoples' passwords are retrained in models across different transformations and processes. I hope both the research and air will be cool.

Overall, it has been a really nice and productive year. I have meet a lot of really interesting and inspiring folks and hope it will continue in 2025.