RSSAmplifier

Blog

Jacob’s Blog

This is a personal website, produced in my own time and solely reflecting my personal opinions. Statements on this site do not represent the views or policies of my employer. I have not signed any contracts that I cannot mention the existence of.

g-w1.github.ioRSS feed ↗10 posts

Latest posts

A protest against the Trump administration

When the Nazis came for the communists, I kept quiet; I wasn’t a communist.

Creating Interpretable Latent Spaces with Gradient Routing

Over the past few months, I helped develop Gradient Routing, a non loss-based method to shape the internals of neural networks.

I found >800 orthogonal “write code” steering vectors

Produced as part of the MATS Summer 2024 program, under the mentorship of Alex Turner.

Building intuition with spaced repetition systems

Do you ever go to a lecture, follow it thinking it makes total sense, then look back at your notes later and realize it makes no sense? This used to happen to me, but I’ve learned how to use spaced repetition to fully avoid this if I want. I’m going to try to convey this method in this post.

What I learned from doing Quiz Bowl

I did Quiz Bowl throughout my time in high school, and looking back on it, it was a pretty positive thing to do! In this blog post, I want to make a list of some of the life lessons I have taken from Quiz Bowl.

Taking into account preferences of past selves

Problem

In set theory, everything is a set

In Unix, everything is a file; in set theory, everything is a set.

The Race to Nowhere: Why Students Take Classes They Don’t Like and What We Can Do About It

[Epistemic Status: Wrote this as an ‘argument paper’ for English class. If I was doing an actual report, I would look way more into the truth of the matter and spend more time researching. Don’t take too much away from this. Putting it up here because I feel parts are relevant and might contain interesting insights.]

An alternative intuition for attribution patching

When I saw the following equation in the attribution patching paper, I got really confused and spent a bunch of time thinking about it:

The Puritans would one-box: evidential decision theory in the 17th century

[Epistemic Status: Pretty confident about the facts; less confident about speculating about what people actually do when presented with decision-theoretic paradoxes.]