TLDR AI 2024-05-24

Mistral & Harvey partnership 🤝, Anthropic Scaling Policy 📈, Google PaliGemma 🤖

Headlines & Launches

Mistral x Harvey Partnership (1 minute read)

Harvey, a legal startup, has partnered with Mistral. The announcement is low on details, but it's likely that they will develop a custom legal model together.

H raises $220M seed round (5 minute read)

H, which aims to build action models and was started by DeepMind scientists, raised a monster seed round. Another win for French AI talent.

Reflections On Our Responsible Scaling Policy (14 minute read)

Anthropic's Responsible Scaling Policy aims to prevent catastrophic AI safety failures by identifying high-risk capabilities, testing models regularly, and implementing strict safety standards, with a focus on continuous improvement and collaboration with industry and government.
Research & Innovation

Text-Video Retrieval (16 minute read)

Researchers have developed a new method, Global-Local Semantic Consistent Learning (GLSCL), to enhance text-video retrieval while significantly reducing computational costs.

Cohere's Aya Model (6 minute read)

The Aya project has 3 different models of increasing size and can converse in 101 languages, many of which are extremely low resource. This project is an incredible step for the open and general access research community.

Text-based Understanding of Proteins (20 minute read)

Researchers present ProtT3, a new framework designed to enhance text-based understanding of proteins by combining Protein Language Models (PLMs) with traditional Language Models (LMs). ProtT3 integrates a PLM for processing amino acid sequences and a language model for generating high-quality textual descriptions using a cross-modal projector called Q-Former.
Engineering & Resources

Enhancing Long-Term Memory for Language Models (GitHub Repo)

Streaming Infinite Retentive LLM (SirLLM) is a new approach that helps large language models maintain longer memory during extended dialogues.

Distributed inference on Llama cpp (GitHub Repo)

Llama cpp now supports distributed inference across multiple machines. It is limited to FP16 for now, but is a great step for open source deployment.

Remote Sensing Image Captioning (GitHub Repo)

This project introduces a new probabilistic diffusion model for Remote Sensing Image Change Captioning (RSICC), which describes environmental changes over time.

What I've Learned Building Interactive Embedding Visualizations (21 minute read)

An individual passionate about interactive embedding visualizations shares their refined workflow for creating such visualizations, which convey the intricate relationships between entities represented as points in high-dimensional spaces. The process involves data collection, building a co-occurrence matrix, sparsifying the matrix, embedding with PyMDE, and projecting to 2D for clear visualization. The creator utilizes web apps for the UI, advocating for their accessibility and GPU-accelerated rendering capabilities.

AI is the reason interviews are harder now (2 minute read)

This article discusses the rising complexity of technical interview questions, with companies now requiring candidates to solve harder problems in shorter times. It highlights the use of AI tools like Ultracode that can assist even non-technical users in passing such interviews. The piece suggests in-person interviews as a solution to ensure that candidates actually possess the programming skills needed for the job.

PaliGemma: Open Source Multimodal Model by Google (9 minute read)

Google has introduced PaliGemma, an open-source vision language model with multimodal capabilities that outperforms its contemporaries in object detection and segmentation. Optimized for fine-tuning on specific tasks, PaliGemma opens possibilities for custom AI applications and provides comprehensive resources for immediate use. PaliGemma demonstrates superior results in OCR and shows promise for various use cases when fine-tuned with custom data.
Quick Links

Krea AI (Product)

Generate an AI video from keyframe images based on users' unique styles, concepts, or products.

Casper Labs Announces AI Governance Solution, Prove AI (4 minute read)

Casper Labs has introduced Prove AI, developed with IBM, to bring transparency and auditability to enterprise AI applications.

ByT5 for Text Rendering (4 minute read)

Glyph ByT5 is a model that operates over bytes and has dramatically improved text rendering in synthesized images.
The most important AI, ML, and data science news in a free daily email.
Join 500,000 readers for