Deep Learning Weekly: Issue 419

This week in deep learning, we bring you Anthropic launches Claude for Chrome in limited beta, How a 20-Year-Old Algorithm Can Help Us Understand Transformer Embeddings, and a paper on Memento: Fine-tuning LLM Agents without Fine-tuning LLMs.

You may also enjoy ByteDance Seed Open-Sources VeOmni, Unlocking Any Modality Model Training, A scalable framework for evaluating health language models, a paper on Memp: Exploring Agent Procedural Memory, and more!

As always, happy reading and hacking. If you have something you think should be in next week's issue, find us on Twitter: @dl_weekly.

Until next week!

Industry

Anthropic launches Claude for Chrome in limited beta

Anthropic has begun testing a Chrome browser extension that allows Claude to take control of users’ web browsers.

ByteDance Seed Open-Sources VeOmni, Unlocking Any Modality Model Training

To advance research and application of omni-modal LLMs, the ByteDance Seed team has unveiled and open-sourced VeOmni, a PyTorch-native omni-modal training framework.

Stanford study finds AI has reduced availability of entry-level programming jobs

A new Stanford study suggests that the number of entry-level programming jobs in the U.S. has declined significantly since the launch of ChatGPT.

Google rolls out image-to-video capability to Google Vids powered by Veo 3

Google is updating its AI-enabled video app Google Vids to make it more accessible and powerful for teams to generate and edit video content.

MLOps & LLMOps

101 real-world gen AI use cases with technical blueprints

A guide that contains 101 architectural blueprints for various generative AI use cases.

8-bit Rotational Quantization: How to Compress Vectors by 4x and Improve the Speed-Quality Tradeoff of Vector Search

A technical blog post about the 8-bit Rotational Quantization method, which compresses vectors by 4x, speeds up vector search, and improves search quality by combining random rotations with scalar quantization.

JUDE: LLM-based representation learning for LinkedIn job recommendations

A post introducing JUDE, LinkedIn's production platform for generating and serving high-quality, fine-tuned LLM embeddings for job recommendations.

A Practical Guide for Choosing the Right Vector Database for Your AI Applications

A comprehensive guide providing a practical decision framework for choosing the right vector database for AI applications.

Learning

How a 20-Year-Old Algorithm Can Help Us Understand Transformer Embeddings

An insightful blog post from the Stanford AI Lab explaining how a modified 20-year-old algorithm, KSVD (specifically DB-KSVD), can be effectively scaled to understand transformer embeddings.

A scalable framework for evaluating health ...