ISSUE 446 · August 1, 2023
Insight
The Many Ways that Digital Minds Can Know
Detractors of LLMs describe them as blurry jpegs of the web, or stochastic parrots. Promoters of LLMs describe them as having sparks of AGI, or learning the generating function of multivariable calculus. This is an insightful post that helps untangle the seeming contradictions.
Ryan Moulton
Podcasts
Shiny for Python
Shiny is a well-known open-source framework that makes it easy to build interactive web apps using R. But starting last year, Shiny supports Python too. In this episode of Talk Python, Joe Cheng, CTO of Posit, introduces the framework and discusses the things it brings to the Python community.
TalkPython['Podcast']
Sponsored Link
Complete customer profiles in your data warehouse
RudderStack Profiles takes the SQL grunt work out of building customer profiles. You specify the customer traits, then Profiles runs the joins and computations for you to create complete profiles, so you can build better models, faster. And that’s just one use case. Get all the details.
Posts & Tutorials
Ultimate practical guide to conjoint analysis with R
Here's a practical guide for using R, {brms}, and {marginaleffects} to analyze conjoint data and find causal and descriptive quantities of interest, both frequentistly and Bayesianly. This is a well-written deep dive with lots of examples.
Andrew Heiss
Design Patterns for LLM Systems & Products
In his latest post, Eugene Yan takes a looks at practical patterns for integrating large language models into systems and products. The post explores academic research, industry resources, and practitioner know-how with the aim of distilling them into key ideas and practices. If you're thinking about LLMs or are already building, this is super useful.
Eugene Yan
Reproducible data science with Nix
Nix is a package manager that makes it easy to install software in isolated environments. That means you can use it to keep all of the code and packages together for each project and use Nix to help ensure reproducibility. In this multi-part series, Bruno Rodrigues shows how.
Bruno Rodrigues
Treemaps are awesome!
Treemaps are an underutilized visualization that are capable of generically summarizing data of many shapes and sizes. They're often used for displaying the files that consume all of your disk space, but with a few tweaks, treemaps can be a flexible tool for exploring and navigating messy data blobs.
Phronemophobic's Blog
How to Check 2 SQL Tables are the Same
Using just standard SQL, can you write a query that checks if two SQL tables are the same? Sounds easy, right? If you like this kind of problem, there's also a great discussion on Hacker News >>
Yisu Remy Wang
Tools & Code
tinyvector
tinyvector is a nearest-neighbor embedding database built with SQLite and Pytorch. It's less than 500 lines of code, stores all indexes in memory for fast querying, and it scales to 100+ million vector dimensions. SQL querying and integrated models are coming too.
GitHub | Will DePue
Announcing Jupyter Notebook 7
Jupyter Notebook 7 is the most significant release of the Jupyter Notebook in years. This new release includes real-time collaboration, interactive debugging, dark mode, internationalization, improved accessibility, and more. Here are the highlights.
Project Jupyter
Resources
Telling Stories with Data
This looks like a great new book that aims to fill the gaps that other data science texts often miss. Ultimately, it shows how to tell stories with data via detailed case studies. It starts with gathering, cleaning and preparing data and then shows how to use statistical models to analyze and share the results in a reproducible way. Free to read online.
Rohan Alexander
 |