stlite (again), Data Visualization and Open Source AI-a boon or a danger
Articles that have piqued my interest over the last few days and weeks
Just a few things that have caught my eye recently…
stlite: Serverless Streamlit— Run Your Apps in the Browser
(AlanJonesCoding blog)
I wrote about stlite in a recent post but here is a quick tutorial on how to use it. stlite is Streamlit ported to Web Assembly which, when combined with the Pyodide Python interpreter allows us to create Streamlit apps that run entirely in the browser.
No more pushing apps to GitHub and deploying them to the Streamlit Cloud, no more Docker or Heroku, just host your app as a static web page. The downside is that it initially takes a little while to load and there are some restrictions on the Streamlit features and Python libraries that you can use. So, not a 100% replacement for Streamlit but worth a look.
Open-Source AI Is Uniquely Dangerous or Open-Source AI Is Good for Us
Which is it? IEEE Spectrum has published two articles that have taken these opposing points of view.
The danger of open source AI models is that they can easily be used for nefarious purposes: deep fake videos can put false words into the mouths of well-known politicians or show famous people engaged in invented fictitious pornographic acts. Open source AI could be harnessed to distribute fake information in political campaigns and thus pervert the democratic process.
On the other hand, should the power of AI be under the control of a handful of mega-corporations that will control its use and deployment? Will government regulation ensure that small players simply won’t have the funds to conform to a regulatory system that the big guys will be able to take in their stride?
Will AI make data analysts redundant?
(Substack)
“How can data visualisation, UX design, and AI work together so we can all benefit (and keep our jobs at the same time)?” asks Sarah Gilligan in her Substack Data Dispatch. Her conclusion is a positive one and is that data analysts need to find ways to use AI to augment their work rather than to replace them. Read more here.
Introducing the Multi-Chord Diagram: Visualizing Complex Set Relationships
(Medium - paywall)
Chord diagrams have only recently impinged on my consciousness (I guess I may have come across them before and thought, “Very pretty but what are they?”).
In this article, Nick Gerend, explains the background to Multi-Chord diagrams and shows us how to implement them from scratch in Python.
The diagrams show how data flows from one place to another, or indeed to more than one other place and are a little like a Sankey diagram but with the origins and destinations arranged around a circle.
You won’t find many implementations in standard Python visualization libraries but Gerend shows us how to use the library that he created to use them in our apps without the fuss of understanding the implementation!