Data Phoenix Digest - ISSUE 5.2024

Welcome to this week's edition of Data Phoenix Digest! This newsletter keeps you up-to-date on the news in our community and summarizes the top research papers, articles, and news, to keep you track of trends in the Data & AI world!

Be active in our community and join our Slack to discuss the latest news of our community, top research papers, articles, events, jobs, and more...

Call For Speakers

We regularly host webinars for our global AI and Data community of Engineers, Executives, and Founders. Being a speaker at our events is a great opportunity to share your technical expertise and knowledge with the Data Phoenix community.

To be considered for a speaker in our future events, please fill out this form.

Latest news

• OpenAI becomes the center of attention with the Sora early announcement
• Cohere For AI launches Aya, an LLM covering more than 100 languages
• Google announces multi-modal Gemini 1.5 with million token context length
• Otter.ai introduces Meeting GenAI, revolutionizing AI-powered tools for more effective meetings
• Anthropic tests Prompt Shield technology in the US to combat election misinformation
• The debut of Slack AI, enhancing workspace communication with AI capabilities
• NVIDIA's Chat with RTX offers AI PC owners the chance to create a personalized chatbot
• Guardrails AI unveils Guardrails Hub, a collaborative platform for AI 'validators'
• The UK-Ukraine TechExchange program accelerates defense-tech and agritech startups in support of Ukraine
• Microsoft announces a significant AI-focused €3.2 billion investment in Germany
• Kong Inc. introduces a new open-source AI Gateway featuring six plugins for enhanced AI and LLM usage
• Armilla AI celebrates a successful $4.5M seed funding round for AI product quality and performance
• Clarity AI secures $16M in seed funding to combat deepfakes and AI-generated synthetic media
• Zylon emerges with $3.2M raised in pre-seed funding, focusing on AI-powered workplace assistance
• Andrej Karpathy departs OpenAI, focusing on personal projects with a positive outlook
• Mindy announces a $6M seed round and the general availability of its AI-powered email assistant
• OpenAI's valuation soars over $80B, confirming its leading position in the AI industry
• LangChain lands a $25M round led by Sequoia Capital and launches LangSmith.

Find details and more news on our website.

Have some exciting AI news you'd love to share with the world? We'd be thrilled to hear from you! Feel free to send us your stories – we're always looking for fresh insights and can't wait to feature your news!

Summary of the top articles and papers

Articles

Milestones in Speaker Recognition
Speaker recognition has been a topic of interest for researchers for many decades. The recent advancements in AI have changed the landscape, and this article aims to provide a detailed review of these changes, as well as the historical context. Check it out!

Graph Neural Networks in TensorFlow
TensorFlow GNN 1.0 is a production-tested library for building GNNs at large scales. It supports both modeling and training in TensorFlow as well as the extraction of input graphs from huge data stores. Learn more about its role in modeling the relations between objects!

Building, Evaluating and Tracking a Local Advanced RAG System | Mistral 7b + LlamaIndex + W&B
RAG is an NLP technique that combines LLMs with selective access to knowledge to reduce LLM hallucinations. This article shows how to build a RAG system using a locally running LLM, which techniques can be used to improve it, and how to track the experiments.

Unlocking LLM’s Potential with RAG: A Complete Guide from Basics to Advanced Techniques
LLMs are increasingly powerful, yet often not ideal for handling new information. To improve this, experts created RAG, a method that helps LLMs search for extra information from the internet or other places when they need to answer questions. Dive into the topic!

Transformers from Scratch
This detailed article demonstrates how to create and train a transformer from scratch. It goes through each foundational element step by step and explains what is happening along the way. It is written in a Jupyter notebook so you can use it to run the code yourself.

Papers & projects

World Model on Million-Length Video And Language With RingAttention
RingAttention is a new technique that allows to scalably train on long sequences using a large dataset of diverse videos and books. This paper looks into how it can be used to help LLMs develop an understanding of both human textual knowledge and the physical world, enabling broader AI capabilities for assisting humans.

InseRF: Text-Driven Generative Object Insertion in Neural 3D Scenes
InseRF is a novel method for generative object insertion in the NeRF reconstructions of 3D scenes. Based on a user-provided textual description and a 2D bounding box in a reference viewpoint, it generates new objects in 3D scenes. InseRF is capable of controllable and 3D-consistent object insertion without requiring explicit 3D information as input.

Multi-Track Timeline Control for Text-Driven 3D Human Motion Generation
Recent advances in generative modeling have led to promising progress on synthesizing 3D human motion from text. However, using a single text prompt as input lacks the fine-grained control needed by animators. This paper describes an intuitive, yet fine-grained, input interface for users, allowing them to timeline control for text-driven motion synthesis.

MagicVideo-V2: Multi-Stage High-Aesthetic Video Generation
MagicVideo-V2 is a new tool that integrates the text-to-image model, video motion generator, reference image embedding module and frame interpolation module into an end-to-end video generation pipeline. Benefiting from these architecture designs, MagicVideo-V2 can generate an aesthetically pleasing, high-resolution video with remarkable fidelity.

En3D: An Enhanced Generative Model for Sculpting 3D Humans from 2D Synthetic Data
En3D is an enhanced generative scheme for sculpting high-quality 3D human avatars. The method described in the paper allows to develop a zero-shot 3D generative scheme capable of producing visually realistic, geometrically accurate and content-wise diverse 3D humans without relying on pre-existing 3D or 2D assets.