ISSUE 407  ·   October 4, 2022

Insight

AI for the Next Era

In this interview, Reid Hoffman and Sam Altman discuss the latest developments in AI and where the opportunities are now. A lot of insights here from the top. Reid is a partner at Greylock and is well-known as the cofounder of LinkedIn. Sam is the CEO of OpenAI and was previously the president of YCombinator. 
Greylock | Reid Hoffman

Sponsored Link

Curate high quality ML datasets with Nucleus today

Nucleus helps ML teams curate better datasets. Bring together your data, ground truth and model predictions to effortlessly fix model failures and data quality issues. Ready to improve your models? Create a free account, upload your data and start mining your edge cases!

Tutorials, Projects & Opinions

K-Means Clustering - An Explorable Explainer

Great interactive explainer that shows how the k-means algorithm works. Covers the algorithm mechanics, loss functions, drawbacks, effect of outliers and more. This project was recently accepted for the 2022 Workshop on Visualization for AI Explainability.
Yi Zhe Ang

Forecasting something that never happened

Counterfactual forecasting is the process of predicting something in the form: what would X be like had there been no Y. In this detailed post, Luca Serra walks through a real-world example from retail that estimates the cost-effectiveness of past promotions. Covers the approach used, tips, the evaluation process, and limits to the approach.
Artefact | Luca Serra

The anatomy of an ML-powered stock picking engine

This is a detailed essay about using machine learning to beat the market. It covers the design of a system called Didact AI, which was consistently beating the S&P 500 until it was shut down in July 2022. There's a lot here, including a discussion about market dynamics, the modeling approach, tech stack, feature engineering, the ML pipeline, and more. 
Principia Mundi

Forecasting gas prices demo

Great tutorial that uses gas price forecasting as a vehicle to explore three new packages: 1) statsforecast’s AutoARIMA implementation, 2) LineaPy for saving data and generating a pipeline that can be scheduled to run regularly and 3) Quarto for rendering documents from Jupyter. This is a step-by-step tutorial with code and discussion along the way.
Sean J. Taylor

The Illustrated Stable Diffusion

This is a gentle introduction to how Stable Diffusion works with lots of visuals to make it easy to follow.
Jay Alammar

How to build a repo of data resources from Twitter

Here's how to build an interactive Shiny document that turns your Twitter Likes into a searchable repository. This is specifically set up for data science topics but it would be trivial to modify it for other topics.
Dr Craig Hamilton

Data Visualization

Considerations for using text in data visualizations

There's a lot of info about how the right chart type and colors can improve visualizations – but not much about how to use words well. Here's everything you need to know, including phrasing, text placement, fonts, styles, colors, number formats, and more. With lots of examples.
Datawrapper | Lisa Charlotte Muth

How did you like this issue of Data Elixir?

👎    1     2     3     4     5   👍

Was this email forwarded to you?  Sign up here >>