# Dec 5, 2024

## Genie 2: A large-scale foundation world model
- **Genie 2** is a groundbreaking **foundation world model** that generates diverse, action-controllable 3D environments from a single image prompt, enabling limitless training scenarios for AI agents.

## GenCast predicts weather and the risks of extreme conditions with state-of-the-art accuracy
- **GenCast** is a new AI ensemble model that enhances weather predictions, achieving **superior accuracy** over the European Centre for Medium-Range Weather Forecasts (ECMWF) by outperforming it on **97.2%** of tested targets, especially for extreme weather events.

## VectorChord: Store 400k Vectors for $1 in PostgreSQL
- **VectorChord enables the storage of 400,000 vectors for just $1**, offering a cost-effective solution that outperforms competitors like Pinecone and pgvector by a factor of 6x and 26x, respectively.

## PaliGemma 2: Powerful Vision-Language Models, Simple Fine-Tuning
- **PaliGemma 2** enhances vision-language capabilities, allowing models to generate **detailed captions** and recognize complex inputs like chemical formulas and chest X-rays, as outlined in the [technical report](https://arxiv.org/abs/2412.03555).

## [R] ICLERB: A better way to evaluate embeddings and rerankers for in-context learning
- The **ICLERB** benchmark introduces a novel evaluation framework that utilizes **Direct Preference Optimization (DPO)** to better assess the effectiveness of embeddings and rerankers in **in-context learning** scenarios, moving beyond traditional relevance metrics.

## How to pack ternary numbers in 8-bit bytes
- **Efficient packing of ternary numbers** into 8-bit bytes achieves **1.6 bits per trit**, resulting in **99.06% efficiency** compared to perfect packing, which is crucial for optimizing data storage in machine learning models like BitNet b1.58.

## Exploring inference memory saturation effect: H100 vs. MI300x
- The benchmark reveals that **NVIDIA's H100** suffers from **memory saturation** under large prompts, leading to a **51% cost advantage** for AMD's **MI300x** when using a single replica on 8xMI300x compared to two replicas on 4xMI300x.

## Bringing the Instructions to the Data
- **LLVM IR can optimize SQL query execution** by compiling queries into efficient machine code, significantly improving performance, as demonstrated with the NYC Taxi dataset where vectorized LLVM IR execution reduced runtime to **74 µs** from **340 µs** using naive methods.

## [D] Daily Paper Discussions - FlashAttention 3
- **FlashAttention-3** enhances performance on Hopper GPUs through **Producer-Consumer Asynchrony**, which optimally divides tasks to utilize GPU resources and minimize delays.

## [R] ReVersion: Learning Relation Prompts from Images for Controlled Diffusion Generation
- **ReVersion** innovatively learns and transfers **visual relationships** using diffusion models, focusing on **interaction** rather than mere appearance through relation prompts and specialized sampling techniques.

## 🤗Welcome PaliGemma 2 – New vision language models by Google
- **PaliGemma 2** introduces enhanced vision language models from Google, featuring **new pre-trained models** with **3B, 10B, and 28B parameters**, offering flexibility in input resolutions of **224x224, 448x448, and 896x896** for diverse applications.

## Google DeepMind at NeurIPS 2024
- **Google DeepMind** will showcase **over 150 new papers** at NeurIPS 2024, highlighting advancements in **adaptive AI agents**, **3D scene creation**, and **LLM training** methodologies.

## NVIDIA NIM on AWS Supercharges AI Inference
- **NVIDIA NIM microservices** are now integrated into AWS services, enabling **faster AI inference** and reduced latency for generative AI applications, enhancing deployment efficiency for developers.

## 2025 Predictions: Enterprises, Researchers and Startups Home In on Humanoids, AI Agents as Generative AI Crosses the Chasm
- **Generative AI is projected to generate $1.3 trillion in revenue by 2032**, as enterprises and startups increasingly adopt multimodal models to enhance innovation and efficiency across various sectors.

## 🤗“How good are LLMs at fixing their mistakes? A chatbot arena experiment with Keras and TPUs
- **LLMs can effectively fix mistakes** when prompted, as demonstrated in a chatbot arena experiment using Keras and TPUs, where models like **Gemma 2 9B** and **Llama 3.1 8B** consistently produced correct API calls after receiving feedback.