🥟 Chao-Down #8: A new sorting record, cheaper to train Stable Diffusion, and an explosion in Generative AI funding

Today’s stories are all about new breakthroughs in costs enabled by advances primarily in software. As the cost of computation goes down, we’ll start to see new capabilities and applications begin to emerge as what was once thought impossible becomes possible.

Exciting times indeed.

-Alex, our resident Chaos Coordinator.

What happened in AI? 📰

Ray breaks the $1/TB barrier as the world’s most cost-efficient sorting system (anyscale.com)

Training Stable Diffusion from Scratch Costs <$160k (mosaicml)

The CEO of a $1.6 billion biotech says his bet on AI is paying off with faster, cheaper research that could reshape the industry (Business Insider)

Generative AI funding exploded over the past 2 years - (Insider Intelligence Trends, Forecasts & Statistics)

Always be Learnin’ 📕 📖

• Building Data and ML Applications with HuggingFace Spaces (Canva)

• Everything is Connected: Graph Neural Networks (Arxiv)

  • “In many ways, graphs are the main modality of data we receive from nature. This is due to the fact that most of the patterns we see, both in natural and artificial systems, are elegantly representable using the language of graph structures. Prominent examples include molecules (represented as graphs of atoms and bonds), social networks and transportation networks.”

Projects to Keep an Eye On 🛠

FlashConv: Speeding up State Space Models (link)

𝗗𝗦𝗣 (Demonstrate–Search–Predict) - A framework for composing retrieval models and language models into powerful pipelines that tackle knowledge-intensive tasks. (Github)

The Latest In AI Research 💡

Batch Prompting: Efficient Inference with Large Language Model APIs (arxiv)

A Watermark for Large Language Models (arxiv)

The World Outside of AI 🌎

Google’s Most Serious Anti-Trust Lawsuit To-Date (Platformer)

A comprehensive recap of CES 2023 - More Business, Less Spectacle (Learning By Shipping)

One Last Bite 😋