This Week I Learned - Week #29 2024

This Week I Learned - 

* Discourse is an open source platform for community discussion & an alternative to Slack. Open Food Facts uses Discourse, Slack & Wikis for community discussions.

* Guidance for writing text descriptions (alt text) for digital images when creating HTML content

* pyspark.ai is an English SDK for Apache Spark. It takes English instructions and compile them into PySpark objects like DataFrames. Its goal is to make Spark more user-friendly and accessible, allowing you to focus your efforts on extracting insights from your data.

* Enterprise Architecture on a Page v2.1 [PDF]

* IBM's watsonx.ai supports open-source models from Mistral, Saudi Data and AI Authority (SDAIA) & Meta. It includes Mistral's popular Mixtral series of open-source mixture-of-experts models

* Claude is now available as an Android app.

* Many conversational AI platforms and chatbots require the use of voice datasets. The Vaani project by Bhashini, led by IISc-ARTPARK and Google, is open-sourcing 16,000 hours of speech recordings from 80 districts and an extensive 150,000 hours from one million individuals across 773 districts in India. This vast dataset is available for any organization interested in developing speech models for Indian languages and dialects, allowing them to download and utilize the data in alignment with their business goals.

* Large language models (LLMs) are central to generative AI applications; however, they are not trained on enterprise data, which can lead to “hallucinations” when addressing business-specific questions. Retrieval-augmented generation (RAG) can solve this problem by combining LLMs (like ChatGPT) with AI Vector Search on private business data. RAG allows AI Vector Search to provide LLMs with relevant, real-time business data, to generate accurate responses without training the LLM on that data. RAG with Oracle Database also benefits from advanced security (so only user-visible data is used for augmentation), sophisticated SQL to enforce business rules, plus a full-featured text pipeline for document processing. With Oracle Database 23ai you can use AI Vector Search to complement public LLMs with your own private data. 

* LangSmith is a unified platform for developing, testing, deploying, and monitoring LLM applications. The LangSmith platform builds on top of LangChain.

* AI Use Cases with Oracle Gen AI service. [PDF]

* Why test-driven development and pair programming are perfect companions for GitHub Copilot

* For startups, Google is expanding access to its AI models like Gemini and Gemma, introducing new language tools from Google DeepMind India, and enhancing the software development process with AI-powered features. In a bid to democratise AI for Indian developers, Google is focusing on three key areas with transformative potential in India—multimodal, multilingual, and mobile. 

* A major deficiency of the AIDA model and other hierarchical models is the absence of post-purchase effects such as satisfaction, consumption, repeat patronage behaviour and other post-purchase behavioural intentions such as referrals or participating in the preparation of online product reviews.

* UTF-8 has a lot of glyphs that look like other characters. Using glyph sets in content breaks inline translation or browser translation making it inaccessible to those using screen readers or translation tools. The power of the web is that text can be converted, indexed and made accessible to others. When you garble your message to achieve a certain look or stand out in a busy timeline, you block out quite a lot of people who further down the line could become important multipliers or customers.

* Social media is all about engagement. But there’s a fine line between encouraging interaction and being downright manipulative. Interaction bait is killing social media - Christian Heilmann 

* The AIDA model, comprising attention, interest, desire, and action, is the cornerstone of numerous successful marketing strategies. Introduced by Elias St. Elmo Lewis, an American advertising and sales pioneer, in 1898, this formula parallels the captivating elements of a skilled storyteller. Since stories tend to be more memorable than mere facts and figures, storytelling serves as an effortless and potent method to pique curiosity and connect with an audience.

Source: Hubspot

* In the bestselling self-help book Atomic Habits, author James Clear talks about habit stacking. To build a new habit easily, he suggests pairing it with a long-established habit you already have.

* There are more guns than people in America. More than 32% of the people apparently own a firearm.

* In the preface of his novel Breakfast of Champions, or Goodbye Blue Monday,  Kurt Vonnegut states that he tends "to think of human beings as huge, rubbery test tubes, too, with chemical reactions seething inside."