Technical AI Safety | Representation Learning | Data Scientist
greenwald-30c3-keynote. Glenn Greenwald's keynote at 30C3 (the 30th edition of the Chaos Communications Congress).
92dl-studies. Notebooks and notes on data-driven experiments derived from my studies with fast.ai's courses.
10transformer-visualization. Mechanistic Interpretability Tutorials, Results and research log as I learn from publicly available research, and experimentation. Evolving work, open ended, slow updates. Lots of incomplete work.
9chaitin-ait. Implementation of Chaitin's Lisp from 1995's The Limits of Mathematics -- Tutorial Version https://arxiv.org/abs/chao-dyn/9509010
7semisupervised-rice-africa. Using semisupervised learning to classify rice diseases
3anki. Anki for desktop computers
3containerized-rag-starter-kit. Scaffold for research purposes
2life. Life - a timeline of important events in my life
2circuit-tracer. Python
1consilience. The principle that evidence from independent, unrelated sources can "converge" to strong conclusions (https://en.wikipedia.org/wiki/Consilience)
1pgloader. Loading data into PostgreSQL
1clml. Common Lisp Machine Learning Library
1smlar. Clone of git://sigaev.ru/smlar.git
1Syft. Homomorphically Encrypted Deep Learning Library
1cmix. cmix is a lossless data compression program aimed at optimizing compression ratio at the cost of high CPU/memory usage.
1terminology-server. An UNMAINTAINED historical effort to create a SNOMED terminology server
1fastText. Library for fast text representation and classification.
1bertviz. BertViz: Visualize Attention in NLP Models (BERT, GPT2, BART, etc.)
1WorldModels.jl. WIP - Buggy and/or incomplete Implementation of recurrent world models in Julia - FluxML and POMDPs.jl
1elp_redux. Augmenting Episodic Linear Probes to Improve Data Safety
1