# Mar 25, 2024

- **Your salary is determined mainly by geography, not your skill level** (conclusions from the salary model built with 24k samples and 300 questions)  
  A **salary model** built from nearly 24,000 responses in the 2022 Kaggle Machine Learning & Data Science Survey reveals that **geography** significantly outweighs skill level, experience, or industry in determining salaries for data professionals. [Source](https://jobs-in-data.com/salary/data-scientist-salary)

- **Sora: First Impressions**  
  **Sora**, a tool introduced by OpenAI, is revolutionizing the creative process for visual artists, filmmakers, and designers by enabling the creation of **new and surreal ideas** previously deemed impossible.

- **New algorithm unlocks high-resolution insights for computer vision**  
  **FeatUp**, developed by MIT CSAIL, enhances computer vision by **upgrading the resolution of deep networks**, enabling them to capture both high- and low-level scene details simultaneously, akin to providing Lasik for computer vision systems.

- **“Emergent” abilities in LLMs actually develop gradually and predictably – study**  
  **Large language models (LLMs) like GPT-3 and LAMDA develop unexpected skills gradually and predictably**, not suddenly as previously thought, when measured with more nuanced metrics.

- **Pyenv – lets you easily switch between multiple versions of Python**  
  **pyenv** offers a seamless way to manage multiple Python versions on a per-user or per-project basis, without depending on Python itself or requiring shell integration.

- **PSChess – A chess engine in PostScript**  
  **PSChess** is a chess engine written in PostScript, designed to explore the potential of executing arbitrary code on printers and implementing complex logic within the PostScript language. [GitHub Repo](https://github.com/nst/PSChess)

- **ML interviews can be improved by focusing on candidate experience**  
  **ML interviews** can be improved by focusing on **candidate experience**, ensuring they are **less exhaustive** and more **reflective of real-world tasks**.

- **Is GPT-4 a Good Data Analyst?**  
  **GPT-4 demonstrates comparable performance to human data analysts** in end-to-end data analysis across various domains, as revealed by systematic comparative studies.

- **Tracecat – Open-source security alert automation / SOAR alternative**  
  **Tracecat** is an **open-source automation platform** aimed at security teams, offering a Tines / Splunk SOAR alternative with a focus on **enterprise-grade tools, AI infrastructure, and a practitioner-obsessed UI/UX**.

- **OpenAPI AutoSpec** is a **server proxy** that automatically generates **OpenAPI 3.0 specifications** for local websites or applications by capturing and documenting network requests in real-time.

- **Aleksa Godric's guide on securing a job at DeepMind** remains a topic of interest, especially for individuals applying deep learning in startups and aiming for roles at prestigious AI firms like DeepMind or Anthropic.

- **Variational Autoencoders (VAEs)** are perceived to be losing ground to **Diffusion models**, despite the latter being a specialized form of VAE.

- **Introducing Tamil Mistral: Opening Up New Language Possibilities with LLM Pretraining**  
  **Tamil Mistral** has been updated to better handle the **tokenization of Tamil words**, particularly 'uyir ezuthukal', by expanding its vocabulary from 32k to 50k and increasing the embedding size, leveraging a 25GB Tamil text corpus for pre-training.

- **Livebook** serves as a versatile tool in the Elixir ecosystem, enhancing workflows from learning to production with its notebook-style interface and support for reproducible, shareable, and extensible Elixir workflows.

- **Microsoft's acquisition strategy** mirrors its OpenAI approach with the **takeover of Inflection AI**, signaling a broader trend of Big Tech's aggressive expansion in the AI sector.