# May 7, 2024

## Kan: Kolmogorov-Arnold Networks
- **Kolmogorov-Arnold Networks (KANs)**, inspired by the **Kolmogorov-Arnold representation theorem**, propose **learnable activation functions on edges** instead of fixed ones, eliminating linear weights in favor of **spline-parametrized univariate functions**.

## Apple introduces M4 chip
- The **M4 chip**, utilizing **second-generation 3-nanometer technology**, significantly enhances the **iPad Pro's performance and efficiency**, featuring a **new display engine** for the Ultra Retina XDR display and a **10-core CPU and GPU** for advanced computing and graphics capabilities.

## The best way to have complex discussions?
- **CQ2** offers a structured approach to **complex discussions**, addressing common issues like impulsive responses and lack of organization found in traditional platforms.

## A 100x speedup with unsafe Python
- Achieving a **100x speedup** in numpy code by utilizing "unsafe Python" techniques, the method involves manipulating memory layout and strides of numpy arrays to optimize performance.

## [D] Kolmogorov-Arnold Network is just an MLP
- **Kolmogorov-Arnold Networks** can be represented as **Multi-Layer Perceptrons (MLPs)** by incorporating specific **repeats and shifts** before the ReLU activation function.

## We've been put in the vibe space
- **The integration of LLMs into user interfaces** is challenging pre-existing user expectations by blending search and recommendation experiences, leading to a **"vibe space"** where interactions are less defined by traditional search paradigms.

## Gradient descent visualization
- **Gradient Descent Viz** is a desktop application designed to **visualize various gradient descent methods** in machine learning, aiming to foster intuitive understanding among users of all expertise levels.

## New iPad Pro with M4 chip and Apple Pencil Pro
- Apple's **new iPad Pro** features a **breakthrough Ultra Retina XDR display** with **tandem OLED technology**, offering **unprecedented visual experience** and **brightness levels**.

## Eli Bendersky: Faster XML Stream Processing in Go (2019)
- **Eli Bendersky introduces a new Go module, [gosax](https://github.com/eliben/gosax),** for faster XML stream processing by leveraging C's performance advantages.

## Deterministic Quoting: Making LLMs safer for healthcare
- **Deterministic Quoting** is a technique developed by Invetech to ensure **LLMs quote verbatim from source material**, addressing the challenge of hallucinations in healthcare applications.

## ScrapeGraphAI: Web scraping using LLM and direct graph logic
- **Scrapegraph-ai** is an **open-source library** designed for **AI-powered web scraping**, enabling users to **scrape thousands of web pages** efficiently with minimal setup.

## [D] Stack Overflow partnership with OPEN AI
- **Stack Overflow's partnership with OpenAI** may formalize access to the platform's extensive knowledge base, following **speculations** that OpenAI had already utilized Stack Overflow data for training ChatGPT.

## Microsoft Working on 'Far Larger' In-House AI Model
- Microsoft is developing **MAI-1**, an **AI model with 500 billion parameters**, aiming to outperform competitors like Google and OpenAI.

## [D] Llama 3 Monstrosities
- A **120B Instruct variant of Llama 3** was created by merging it with itself, resulting in a model with duplicated layers, showcasing a trend in crafting **Frankenstein models**.

## YC’s founder-matching service helped medical records AI startup Hona
- **Hona**, a **GenAI medical records startup**, was founded by friends **Danielle Yoesep and Adam Steinle**, later joined by AI specialist **Shuying Zhang** through **Y Combinator's Co‑Founder Matching Platform**, aiming to **summarize patient medical records** for doctors.