The Cerebellum Beyond Motor Control
Searching beyond the streetlight finds a plethora of important cerebellar functions.
Interested in Machine Learning, Neuroscience, and Original Glazed Krispy Kreme Doughnuts.
Searching beyond the streetlight finds a plethora of important cerebellar functions.
Why Holographic Reduced Representations cannot be used to “Recast Self Attention”.
How Hopfield Networks are a special case of the biologically plausible Sparse Distributed Memory.
A simple modification to improve Elastic Weight Consolidation and Synaptic Intelligence continual learning baselines.
It is absurd that the second year of my PhD has come to an end. I am taking a minute to reflect on what I have learnt and want to improve upon.
Talent is a conversation starter on the under discussed topic of how to identify talent but not much more.
Having a non-linear uptick in utility avoids the Repugnant Conclusion.
Ted Chiang creates emotionally resonant and novel perspectives on deep questions about life and technology.
Working out how much memory it takes to train a Transformer GPT2 Model.
How to Build a Brain over promised and under delivered but I appreciate its ambitious goal to “build a brain” and the efforts made towards it.
Notes and themes from NAISys 2022 - Neuroscience to Artificially Intelligent Systems Conference.
Computer hardware and the need for fast iteration cycles.
What are rational, justifiable reasons to take the actions that define our lives. What is personal identity? Who am I? Who are you? How do we relate?
Computers deal with finite time and resources in many of the same ways that humans do. Algorithms to Live By engagingly reviews some of the most successful algorithms developed in computer science, and shows how they can be applied to our lives.
Science is good explanations. The significance of the Enlightenment. The multiverse theory of quantum physics. Optimism for the future and its enemies. Why First Past The Post is actually a great election system.
Notes from exploring the current state of Brain Atlases.
Research Questions I am interested in/confused by but am not actively researching.
Unconventional advice I wish I’d heard entering undergrad.
We show the heuristic Attention operation can be implemented with simple properties of high dimensional vectors, in a biologically plausible fashion.
I made a tool that tries to help you to optimally size your bets/investments, utilizing the Kelly criterion.
If your idea is so good, why hasn’t it already been done before?
Reflections on the relationship between copying and creativity across art, science, and evolutionary biology.
Resources in computational/theoretical neuroscience that I have found particularly influential.
After Phrenology by Michael L. Anderson makes three main arguments (I have ordered by compellingness):
I gave Journal Club to the Pehlevan lab on the paper An RNA-Based Theory of Natural Universal Computation by Hessameddin Akhlaghpour. I am sharing my slides here in case they are of interest to anyone!
Developing a new platform to identify viral genes responsible for immune suppression/invasion and applying it to COVID.
Designing a COVID Vaccine with the Gifford Lab at MIT CSAIL
Chesterton’s Fences are important, and very hard to identify/evaluate. With finite time, bountiful stupidity and inflated egos, it is too easy to not look deeply enough at existing ways of doing things and understand why they are the way they are before attempting to “fix” them.
My favourite questions from David MacKay’s Information Theory, Inference and Learning Algorithms.
Why humans are likely to be the most diseased organisms on the planet with entire classes of parasites that few other organisms face.
With the early announcement of the positive NIH double blinded RCT results, we know with high confidence that Remdesivir works. Looking at what little data we have and the other recently published China based Remdesivir RCT, I believe it is likely that our current Remdesivir results are a lower bound on its efficacy. This is because of when treatment began relative to symptoms.
A new way to sample from language models for text generation.
This essay was written for the Benedictus Young Scholar’s Essay Prize on the topic of education reform.