updates by Uzay G. '26

Hey!

Here are some personal/academic updates. I haven’t been blogging enough as you can tell, but I want to be prolific this summer.

This semester I spent a larger than usual amount of time thinking about more personal topics and reflections in my life, which maybe I’ll unpack in other posts, but here are the more easily explained things.

Classes

Inference and information – 6.7800

This is MIT’s grad inference and information class. It is challenging. Each lecture has a bunch of content and it’s overwhelmingly rich – but it’s well organized and I learned a lot from it.

I think you learn a few things in this class, but mainly:

• what is the formal and statistical setup people have found to describe the problem of learning & inferring things given data, when evaluated with respect to the optimal solutions (and what are those solutions)

• what are the main challenges to these kinds of learning problems, and what are desirable properties or assumptions we can make to simplify our analysis

• given large amounts of data on a process we want to model, under certain assumptions, what can we say about the techniques that estimate that parameter asymptotically

• how can we quantify the notion of “information” and estimating randomness

You also improve on some meta-skills and get more of things like:

• the mathematical knowhow to take a bunch of scary symbols and boil it down to intuitive understanding and solve problems/prove properties of your algorithms

• statistical background to understand most not too niche ML papers

• a “physicist’s mindset” to inference problems: learning how to take problems and kind of guess their solution/play with the question to solve it

Secure Hardware Design – 6.5950

Veally cool. I did hacking stuff in highschool, notably capture the flag competitions where you need to break into different artifical systems to win. I like the hacker mindset because breaking things, especially secure things, makes you focus on really understanding something through and through.

You need to explore every angle and get a deep understanding of the way the components interact. Now, in many fields you need to learn something deeply, but often especially in school you don’t have real feedback loops on your learning, or at least feedback loops that aren’t artificial like exams. This makes your understanding more brittle, whereas when you try to break something you explore every facet, and breaking it adequately becomes the feedback loop for your learning.

This mindset of turning things into a game extends to many different things. Even in math, you often read about how great mathematicians have the skill of playing with mathematical objects and examples, breaking them down, seeing where things fail and where they work, to develop a good perspective of the field. I think it’s pretty key, but in cybersecurity it’s baked in.

This class was structured to match these ideas – fully graded on labs, where you can freely explore the content from culture and hack the systems they build for you. Very fun stuff and very competent course staff. Would recommend. You basically do low-level hacking and learn more about systems/system design along the way.

Topology/Quantum – 18.901/8.04

This duo was a bit weird. I was taking 8.04 for the first half of the semester, which is Quantum I, because I want to learn quantum physics at some point, even if I’m not a physics major. But I felt like since I don’t need this class for anything, I should only take it if the assessments actually feel useful for what I care about, which is mostly developing a deeper conceptual physics understanding. It felt like although the lectures were interesting what I was being tested on was more about being able to compute and do the math than the physical ideas, so I decided to drop it. And I added topology on add date! which is about halfway through the semester.

The content in topology is quite beautiful. Topology is about shapes, and in particular what properties of shapes are preserved under a few different operations, like continuous deformations and stretches, or bends, retractions, etc…

This seemingly simple question is really complicated! Once you’ve formalized these notions, how do you prove different shapes are truly different?

It’s hard in general to prove there is no way to map two things to each other unless you leverage something called invariants. Invariants are properties that don’t change under a given set of transformations. In this class you define the formalisms for what it means for different shapes to be “equivalent”, and then you define invariants that are preserved by topological equivalences, and learn to compute those invariants.

The content is quite pretty but I did not feel the way it was taught was very instructive for me, so I mostly ended up reading the book and doing problems, which was a bit sad.

Reading and Writing the Essay – 21W.735

Maybe this class is partly responsible for my less than productive blogging this semester. I very much enjoyed it. It’s a writing workshop where you can write about basically whatever you want in essay form, and you get a lot of detailed feedback from your peers and the prof.

I think the key part for me was reading a bunch of really good essays and getting precise, detailed feedback. This class has shown me how writing can really help me clarify my thoughts on something, and I’ve gotten better at conveying ideas in a pretty, argumentative way.

I think my writing interests sits in this zone between conveying aesthetic emotion/a sense of shared empathy and understanding and actually bringing out philosophical ideas and arguments.

What I wrote about:

• existentialism and committing to one’s values

• cultural elites and the values they reflect, contrasting French intellectualism and the American tech scene

• judgement as a form of love and deep empathy

• political polarization and the rise of discourse focused on making institutions feel safe rather than advancing particular values in a normative way

I want to unpack some of these more and maybe bring them to the blogs, but not yet.

Research

I started doing research in the Isola lab this semester, and that’s what I’m continuing in Boston this summer. I like my mentors a lot and think the problem I’m working on is quite interesting.

It’s pretty hard for me to context switch and do research properly during the semester, so I’m excited to focus for the next few months on this.

I’m studying in-context learning, or the phenomenon where language models like ChatGPT, given sequences of the form x, f(x), x’, f(x’), … ie input-output examples of a function f in their prompt, can generalize and figure out the value of new inputs, ie predict f(x”) from x”.

In particular in-context learning is interesting because it underlies a lot of the reasoning ability of the best language models, and it’s still relatively poorly understood. It’s important because we want to know to what extent models learn complex inner search procedures/inner optimization goals, and we want to be careful about what those goals are. Developing a better theory for in context learning is also useful to describe model internals and make better optimization procedures.

This project is in the field of theory of deep learning, which I’m considering specializing in, so it’s very exciting and I particularly enjoy working at the intersection of empirical ML and more theoretical conceptual and mathy thinking.

Climbing

I wasn’t super focused with climbing this semester, but I still got a bit better. I learned to channel a bit more calm and stoicism in the way I climb to not get frustrated, really consider where I’m getting stuck, and then try again in a more focused way. This helped a lot, and now I’m climbing at a V5 level. This summer I will be more consistent and try to make progress on my footwork, my main bottleneck.

Summer stuff

I’m very hyped to be in Boston this summer. Here are some things I want to do:

• do good research :)

  • learn meta skills

  • develop research taste

• climb

• discover new art (music/films/literature in particular)

• cook food

• travel and see friends

  • maybe bike to New York?

• Family!

• write

  • poetry

  • essays

• read some books/textbooks with friends

Reach out if you’re around :)

Parting song recs:

• Expert in a dying field

• Myopia

• Staying here