Posts

Showing posts from April, 2025

CatGPT: Finally, Someone Understands Meow

Image
See more of my AI co-creations

Eco-nomy Class

Image
See more of my AI co-creations  

Watching Weather Through Charts

Observing various weather aspects through charts offers a clearer understanding of seasons and climate compared to analyzing raw data. Datawrapper has an option to update charts, maps, and tables automatically , without republishing.  To handle continously updated data, Datawrapper has a free data hosting service . It can pick up your original data hosted on any external data source like Github or Google Sheets and copies the CSV to its own server at certain intervals (every minute on the first day, then every hour the next 29 days). The chart above is an experiment to see how it works for my data.  By hovering on one of the interactive chart, you can see tooltips in the other charts. Related: Zomato Weather Union: Real-Time Weather Data for 60 Indian Cities via Public API

This Week I Learned - Week #17 2025

Image
This Week I Learned -  *  Code golf - It’s a sport (of sorts) where people attempt to reduce the number of characters they use in source code in order to solve a problem. It tends to result in comically terse, virtually unreadable lines of code, but it still gets the job done.  *  Gergely Orosz on AI coding tools - These tools really feel like a massive multiplier for experienced devs - those of us who have it in our head exactly what we want to do and now the LLM tooling can move nearly as fast as my thoughts!  *  Google's Machine Learning Crash Course - 12 modules, 15 hours * Developers can ask GitHub Copilot questions in Indian languages like Hindi, Kannada or Telugu and get answers in the same script * How Americans spend the day - a data visualization: * The Bechdel test asks whether a work of fiction features at least two women who have a conversation about something other than a man. It is named after Alison Bechdel, who credited her friend L...

Bharat Ratna Legends in 5 Charts

Image
After having it on my wishlist for a long time, I published my first Kaggle Dataset on Bharat Ratna Awardees (1954-2024)  with an accompanying notebook, " Bharat Ratna Legends in 5 Charts " As a Kaggle Experimenter, I was glad to have AI assistants by my side to help me build this timeline chart among other things.  Did you know, the average age of recipients honored during their lifetime (17 of the 53 have been awarded posthumously) is 79 years! Sachin Tendulkar received it at just 41, while Dr. Dhondo Keshav Karve was honored at the incredible age of 100!  The longevity also speaks about the passion, discipline and dedication that these Bharat Ratna legends possess.

This Week I Learned - Week #16 2025

Image
This Week I Learned -  *  A practical
guide to building agents from OpenAI * "AI is a powerful genie that does what we ask it to do, which is not necessarily what we actually want. Every story about genies revolves around the inability of those given the magic wishes to wish for the right thing. The art of asking is everything. That is, the future belongs to those who are exercising the intelligence and insight that AI itself does not have." - Tim O’Reilly * Folksonomy is a user-generated system for classifying and organizing online content using metadata like electronic tags. Also known as social tagging, folksonomies are collections of user-defined metadata. They evolve naturally as users create or store content on specific sites and identify what they believe the content is about. "Tag clouds" highlight various identifiers and their frequency of use on a folksonomy site. This grassroots classification method, similar to blogs and wikis, is subject to vandalism a...

This Week I Learned - Week #15 2025

Image
This Week I Learned -  * Never a dull day in the AI world - Offering sophisticated multimodal features, Meta has introduced Llama 4, their newest open-weight AI models comprising Scout, Maverick, and Behemoth. The models are based on a novel "mixture of experts" (MoE) architecture. Llama 4 is trained to operate with both text and visual inputs like photographs or videos. It uses a technique called early fusion, meaning it processes all input types through one combined system from the beginning. This lets it be trained all at once on a combination of text, image, and video data. Scout can operate effectively on one high-end graphics card—the NVIDIA H100. While Maverick has 17 billion active parameter models with 128 experts, Llama 4 Scout with the same 17 billion active parameters is constructed using 16 experts. In artificial intelligence, parameters are similar to brain cells in that they enable the model to learn from data, interpret language and make informed dec...

The Financial Jungle - Animal Metaphors

Image
See more of my AI co-creations

Maps: Lesson and Listen Series by Bill Shander

Image
In Episode #9 of Bill Shander's Data Visualization, Data Storytelling, and Information Design - Lesson and Listen Series on LinkedIn Learning, he interviews Dr. Kenneth Field, a Senior Product Engineer at ESRI. Ken is the author of 2 books - Cartography (2018) and Thematic Mapping (2021) He also blogs ( cartoblography.com ) and tweets ( @kennethfield ) about mapping. Key takeaways from the talk: Maps are more than navigation tools: While primarily used for navigation, maps also provide valuable insights into culture, technology, and non-geographic data. Thematic maps convey data: Thematic maps, like choropleths, use visual elements (color, shading) to represent data values across geographic regions. Choropleths have limitations: Choropleth maps can be misleading because they emphasize the size of geographic areas, which may not accurately reflect the data being represented (e.g., electoral votes vs. land area). Cartograms offer an alternative: Cartograms, especially grid...

Top Prompter, Top Performer

Image
See more of my AI co-creations

This Week I Learned - Week #14 2025

Image
This Week I Learned -  * OpenAI is planning to release its first open-weight language model with reasoning capabilities. An open-weight language model's trained parameters or weights are publicly accessible, which can be used by developers to analyze and fine-tune the model for specific tasks without requiring original training data. These language models are different from open-source models, which provide access to the complete source code, training data and methodologies. * Playlist of the 5-Day Gen AI Intensive Course with Google * A single human genome's raw data alone demands roughly 200 GB. This is all necessary to comprehend gene variants and their locations at the subatomic level. Statistically, roughly 0.1% of gene locations differ between any two people, creating a broad spectrum of genetic makeup. Understanding these variations helps pinpoint the causes (and target specific mutations) of genes (disordered sequences causing diseases), influencing disease susceptibi...

HOW TO Manage Web Video Autoplay

Image
If you find videos playing automatically when you visit a webpage annoying, there is a browser setting in most browsers like Edge, Firefox or Vivaldi to block such videos.  Here's how you can enjoy a quieter browsing experience: For Firefox: Open Firefox and click on the menu button (three horizontal lines) in the top-right corner. Select Settings. In the Settings window, click on Privacy & Security in the left-hand menu. Scroll down to the Permissions section. Find the Autoplay setting and click on Settings next to it. Choose Block Audio and Video to stop videos from playing automatically. For Vivaldi: Open Vivaldi and click on the Vivaldi menu (Vivaldi logo) in the top-left corner. Select Settings. In the Settings window, go to the Websites tab. Scroll down to the Media Autoplay section. Choose Block to prevent videos from playing automatically. For Microsoft Edge: Open Edge and click on the menu button (three horizontal dots) in the top-right corner. Select Settings. Click o...

Jeans may fade and rip, but bad genes last forever

Image
See more of my AI co-creations