The Parable of the Seeker
At age twelve, Alex joined a monastery. At age twelve and two weeks, he quit, finding the monks insufficiently motivated, too preoccupied with the comforts o...
Hello, my name is Kelechi Nwankwoala
I’m a software engineer, writer, athlete, and hobby-maximalist.
In the summer of 2024, I completed a 3-month batch at the Recurse Center, a self-directed coding retreat. While I was there, I focused on developing research-level AI expertise. I spent my first 8 weeks learning AI Safety and research fundamentals through Alignment Research Engineer Accelerator (ARENA) I spent the last month putting those skills to use in an independent research project investigating the use of Sparse AutoEncoders (SAEs) to selectively remove information from large language models.
After Recurse, I was selected to be a participant in Neel Nanda’s MATS training phase, where I worked on interpreting mechanisms of refusal in LLMs. Outside of AI research, I’ve worked on a desktop app to visualize personal time data and fine-tuning an LLM to sound like me.
Recent posts
Jo
Jo had an interesting sense of identity. Being a fraternal twin, she had seen what she could be, and so she didn’t think much of herself. Castor liked that a...
Not a Lot, Just Forever
Like the rock bears the weather/not a lot, just forever. — Adrienne Lenker
My Favorite Made Up Game
If you are reading this sometime between March and May, know that somewhere in New England skinny prep school children aged 13-19 are running over concrete, ...
Execution Failures are Still Failures (Calibrate Surprise by Making Weird Excuses)
“Almost only counts in horseshoes and hand grenades” — Frank Robinson
Sparse Autoencoders for Improving Unlearning in Large Language Models
A.K.A: Smallish Large-Language Models: What Do They Know? Can They Un-Know Things?? Let’s Find Out.