ISSUE 448 · August 15, 2023
Talks & Conferences
Catching up on the weird world of LLMs
Great talk summarizing the last few years of the world of LLM development. Covers what LLMs are, what you can use them for, what you can build on them, how they’re trained and some of the many challenges involved in using them safely, effectively and ethically. The post includes an edited transcript and slides if you prefer to read.
North Bay Python | Simon Willison
Sponsored Link
Applied Data Science with Python
In this intermediate level course from Coursera, learn how to conduct statistical analysis, critique data visualization, apply machine learning, analyze social network connectivity, and much more. Enroll for free.
Posts & Tutorials
How to Match LLM Patterns to Problems
In this follow-up to his LLM Patterns post from a couple weeks ago, Eugene Yan explores potential problems when using LLMs and the patterns that help mitigate them. Covers issues related to both internal LLMs that you control and external LLMs that you don't control.
Eugene Yan
LearnDB
This new Python project is a good opportunity to learn how databases work under the hood. Essentially, it's a simplified sqlite clone that's been implemented from scratch with the intention to make it easy to learn and tinker with.
GitHub | Spandan Bemby
Wrapping C Code in an R Package
If you work in R and a collaborator who works in C gives you some code, what do you do? You could rewrite their code in R. You could say, "no thanks." Or, you could follow this step-by-step tutorial and package up their C code to make it available in R.
Jonathan Carroll
How to fill maps with density gradients with R
Plotting hundreds of data points on a small map of a city and have it make any sense is hard to do. This tutorial explores the issues and shows how to fix overplotted points on maps by creating bins or filled density gradients using R, {ggplot2}, and {sf}.
Andrew Heiss
Resources
Probabilistic Machine Learning: Advanced Topics
Kevin Murphy's Advanced Topics book is for researchers and graduate students working in machine learning and statistics who want to learn about deep learning, Bayesian inference, generative models, and decision making under uncertainty. The final version was just released and is available in print or follow the links to download a free PDF.
Kevin Patrick Murphy
Tools & Code
functime
functime is a Python library for time-series machine learning and embeddings at scale.
DescendantAI
causact
The {causact} R package is bringing numpyro speed to R. Learning computational Bayesian inference has never been easier.
GitHub | Adam Fleischhacker
Outlier
What a yarn!
Great story about tactile data visualization - with crochet! If you like the ideas here, be sure to follow the links.
Nieman Journalism Lab | Sarah Scire
 |