kia ghods

I'm a junior at Princeton, studying computer science and mathematics. These days, I spend my time thinking about machine learning, systems, robotics and control, and their intersections. Above all, I love learning, trying to make machines that learn and think, and thinking about our own abilities to learn.

github · scholar · linkedin

> language modeling, control, alignment at MinRegret Lab (advised by Prof. Hazan)

> $\nabla$ Gradient

> co-pres of Princeton ACM

> previously: CoCoSci (advised by Profs. Cohen, Griffiths) · Hillspire · Astralis

I always enjoy talking to new people, and am excited to hear new perspectives; feel free to reach me via email:
kia.ghods [at] princeton.edu

currently on my mind

• What is preventing us from having robotic models running on edge hardware instead of datacenters?
• Is there a way to probe a small model to consistently figure out if it will scale well?
• Why do all human languages converge on roughly the same information rate, despite wildly different syllable speeds?

previously on my mind

• Why does the same thing, entropy, show up everywhere? In thermodynamics, machine learning, telegraphy, ...
• Can you write a piece of DNA from scratch and have it actually do something?
• Why does a person who glances at a picture often understand it better than a machine that has studied it for hours?
• What does symmetry have to do with neural networks?
• Can an AI learn to lie, and can lies spread between AIs?
• Do language models have a voice, or just one voice?
• Why is attention slow, and where does the time actually go?
• How does a computer take a derivative/gradient, and how fast can that go?
• Why is the FFT so fast, and does its trick work for other kinds of matrices?
• How does a computer solve an equation it has never seen?
• When a machine sees two red balls and a blue one, does it know which is which?
• How does a pile of drones decide who looks where?
• When two strains of a virus share one body, do they fight each other -- and do real patients agree with what the math says?
• What is a virus actually doing, hour by hour, inside one person?
• Can cancer be turned off by gluing a small molecule to the right piece of RNA?

lately

• The Celestial Hierarchy (Dionysius the Areopagite)
• All of Statistics (Wasserman)
• Brave New World (Huxley)
• Online Control (Hazan)
• The Man Who Solved the Market (Zuckerman)
• AI Snake Oil (Narayanan & Kapoor)