The Daily Index — Friday, October 27, 2023 · TLDR Index

Across today’s newsletters

Today’s Index

TLDR

NEXT.JS 14

This article goes over the changes in Next.js 14. The update features partial prerendering, stable Server Actions, a free course, and more. There are several breaking changes and some features have been deprecated.

nextjs.org · 16 min

TLDR AI

IMAGE DIFFUSION MODEL TRAINED ON OPEN DATA

Most popular diffusion models like Stable Diffusion have been trained on a wide variety of data, but much of that data is subject to different copyright restrictions. This new model from MosaicML is trained purely on Creative Commons data and achieves similar performance to SD 2 while requiring only a fraction of the training data.

arxiv.org · 24 min

TLDR AI

4-BIT QUANTIZATION IN LLMS

This research introduces LLM-FP4, a novel method to compress large language models by converting their weights and actions to 4-bit floating-point values after training.

arxiv.org · 17 min

TLDR AI

EVALUATING GPT-4V'S OPTICAL CHARACTER RECOGNITION ABILITIES

The paper examines how well GPT-4V, a Large Multimodal Model, performs in various Optical Character Recognition (OCR) tasks, from reading scenes and handwritten texts to understanding complex document structures.

arxiv.org · 18 min

TLDR AI

THE CONVERGENCE OF DATA AND SOFTWARE ENGINEERING IN THE AGE OF AI

AI needs data to work, meaning data functions are getting closer to the core of the product-building process and moving away from internal business intelligence.

tomtunguz.com · 2 min