Generative AI for 3D objects from 2D videos, Top-ranked open source LLM royalty-free for commercial use, 3600+ AI jobs in JPMorgan and more
Greetings and welcome to this week's AI Brews - your thoughtfully curated guide to AI products, learning resources and a concise roundup of the week's impactful news. Our goal? To provide a balanced selection in the rapidly evolving AI landscape, keeping you well-informed without the information overload. We value your feedback - don't hesitate to reply to this email with suggestions on how we can make this better for you. Thanks!
In today’s issue:
AI Pulse: News, Insights and Social Spotlight of the Week
AI Toolbox: Product Picks of the Week
AI Skillset: Learn & Build
🗞️ AI Pulse: News, Insights and Social Spotlight of the Week
🔥 News & Insights
The recently released open-source large language model Falcon LLM, by UAE’s Technology Innovation Institute, is now royalty-free for both commercial and research usage. Falcon 40B, the 40 billion parameters model trained on one trillion tokens, is ranked #1 on Open LLM Leaderboard by Hugging Face [Details | Open LLM Leaderboard].
Neuralangelo, a new AI model from Nvidia turns 2D video from any device - cell phone to drone capture - into 3D structures with intricate details using neural networks [Details].
In three months, JPMorgan has advertised 3,651 AI jobs and sought a trademark for IndexGPT, a securities analysis AI product [Details].
Google presents DIDACT (Dynamic Integrated Developer ACTivity), the first code LLM trained to model real software developers editing code, fixing builds, and doing code review. DIDACT uses the software development process as training data and not just the final code, leading to a more realistic understanding of the development task [Details].
Japan's government won't enforce copyrights on data used for AI training regardless of whether it is for non-profit or commercial purposes [Details].
‘Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.’ - One sentence statement signed by leading AI Scientists as well as many industry experts including CEOs of OpenAI, DeepMind and Anthropic [Details].
Nvidia launched ‘Nvidia Avatar Cloud Engine (ACE) for Games’ - a custom AI model foundry service to build non-playable characters (NPCs) that not only engage in dynamic and unscripted conversations, but also possess evolving, persistent personalities and have precise facial animations and expressions [Details | YouTube Demo].
OpenAI has launched a trust/security portal for OpenAI’s compliance documentation, security practices etc. [Details].
Nvidia announced a new AI supercomputer, the DGX GH200, for giant models powering Generative AI, Recommender Systems and Data Processing. It has 500 times more memory than its predecessor, the DGX A100 from 2020 [Details].
Researchers from Nvidia presented Voyager, the first ‘LLM-powered embodied lifelong learning agent’ that can explore, learn new skills, and make new discoveries continually without human intervention in the game Minecraft [Details].
The a16z-backed chatbot startup Character.AI launched its mobile AI chatbot app on May 23 for iOS and Android, and succeeded in gaining over 1.7 million new installs within a week [Details].
Microsoft Research presents Gorilla, a finetuned LLaMA-based model that surpasses the performance of GPT-4 on writing API calls [Details].
OpenAI has trained a model using process supervision - rewarding the thought process rather than the outcome - to improve mathematical reasoning. Also released the full dataset used [Details | Dataset].
WPP, the world's largest advertising agency, and Nvidia have teamed up to use generative AI for creating ads. The new platform allows WPP to tailor ads for different locations and digital channels, eliminating the need for costly on-site production [Details].
PerplexityAI’s android app is available now, letting users search with voice input, learn with follow-up questions, and build a library of threads [Link].
Researchers from Deepmind have presented ‘LLMs As Tool Makers (LATM)’ - a framework that allows Large Language Models (LLMs) to create and use their own tools, enhancing problem-solving abilities and cost efficiency. With this approach, a sophisticated model (like GPT-4) can make tools (where a tool is implemented as a Python utility function), while a less demanding one (like GPT-3.5) uses them [Details].
Google’s Bard now provides relevant images in its chat responses [Link].
🔦 Social Spotlight
Paragraphica - a camera without lens [Twitter thread].
Andrew Ng announces three 3 new Generative AI courses (free) [Twitter thread].
A 2-minute introduction to the fundamental building block behind Large Language Models: Text Embeddings [Twitter thread ].
8 use cases for quick development (<30 lines of code) using LangChain [Twitter thread link].
🔍 🛠️ AI Toolbox: Product Picks of the Week
AskEdith
AskEdith is an AI-powered analytics tool that lets you ask questions from your data in natural language. It then generates and execute SQL and produces the relevant graphs. It integrates with 100+ different data sources.
Recraft
A generative AI platform where you can generate and edit vector art, icons, 3d images and illustrations. Multiples styles and presets are included. It is free and even allows commercial use of the generated images.
TranslateAudio
TranslateAudio lets you translate your video into multiple languages in your own voice. It can also translate to multiple languages at once and can automatically upload to YouTube channels.
📕 AI Skillset: Learn & Build
AI at Zapier: How we use artificial intelligence to streamline work [Link].
6 ways to use the Zapier ChatGPT plugin [Link].
Vertex AI Embeddings for Text: Grounding LLMs made easy [Link].
Generative AI short courses (free) from DeepLearning.AI [Link].
Build 'DrawIt' a Sketch to Image App using AI (Next.js, Chakra UI, Leap API) [Link].
Thank you for reading AI Brews! If you have any thoughts, questions or just want to say hello, please don't hesitate to hit reply. Mariam