New text-to-3D model , 3B multi-modal model on iPhone, Stable 3D, Sky Replacer, Runway's major update and more
Greetings and welcome to this week's AI Brews for a concise roundup of the week's major developments in AI.
In today’s issue (Issue #38):
AI Pulse: Weekly News & Insights at a Glance
AI Toolbox: Product Picks of the Week
AI Skillset: Learn & Build
🗞️🗞️ AI Pulse: Weekly News & Insights at a Glance
🔥 News
Luma AI introduced Genie, a generative 3D foundation model in research preview. It’s free during research preview via Discord [Details].
Nous Research released Obsidian, the world's first 3B multi-modal model family pre-trained for 4 Trillion tokens that runs locally on iPhones. Obsidian competes in benchmarks withWizardLM-13B and GPT4-X-Vicuna 13B and is based on CapybaraV1.9 [Details].
Phind has released a new model Phind Model V7 that matches and exceeds GPT-4's coding abilities while running 5x faster and having16k context [Details].
Runway released an update for both text to video and image to video generation with Gen-2, bringing major improvements to both the fidelity and consistency of video results [Link].
Stability AI announced [Details]:
Stable 3D (Private Preview): a tool to generate a draft-quality 3D model in minutes, by selecting an image or illustration, or writing a text prompt.
Sky Replacer: a tool that allows users to replace the color and aesthetic of the sky in their original photos with a selection of nine alternatives.
integration of Content Credentials and invisible watermarking for images generated via the Stability AI API.
Stable FineTuning (Private Preview)
Hugging Face released Zephyr-7B-β, a fine-tuned version of Mistral-7B that achieves results similar to Chat Llama 70B in multiple benchmarks and above results in MT bench [Details | Demo].
LangChain launched LangChain Templates - a collection of easily deployable reference architectures for a wide variety of popular LLM use cases [Details].
Nvidia unveiled ChipNeMo, a specialized 43 billion parameter large language model for chip design that can answer general questions related to chip design and write short scripts to interface with CAD tools [Details].
Together released RedPajama-Data-v2: an Open dataset with 30 Trillion tokens for training Large Language Models. It’s the largest public dataset released specifically for LLM training [Details].
Hugging Face released Distil-Whisper, a distilled version of Whisper that is 6 times faster, 49% smaller, and performs within 1% word error rate (WER) on out-of-distribution evaluation sets [Details].
Google Research and Google DeepMind present MetNet-3, the first AI weather model to learn from sparse observations and outperform the top operational systems up to 24 hours ahead at high resolutions. Google has integrated MetNet-3’s capabilities across its various products [Details].
Google DeepMind and Isomorphic Labs update on the next generation of AlphaFold: the new model greatly expands coverage of structure prediction beyond proteins to other key biomolecular classes. This paves the way for researchers to find novel proteins to eventually map biomolecular structures needed to design better drugs [Details].
Nolano Research and EleutherAI introduced Hi-NOLIN, first state-of-the-art open-source English-Hindi bilingual model built upon the Pythia model suite [Details].
Google is rolling out Immersive View for Routes in 15 cities, starting this week along with other AI-powered features in Maps. Immersive view combines Street view, aerial imagery, and live information like weather and traffic to give an aerial, photo-realistic preview of your planned Google Maps route [Details].
Perplexity announced two new models pplx-7b-chat and pplx-70b-chat, built on top of open-source LLMs and fine-tuned for chat. They are available as an alpha release, via Labs and pplx-api [Labs Link].
SlashNext's 2023 State of Phishing Report reveals a 1,265% increase in Phishing Emails since the launch of ChatGPT in november 2022, signaling a new era of cybercrime fueled by Generative AI [Details].
Google launches generative AI tools for product imagery to US advertisers and merchants [Details].
🔦 Weekly Spotlight
Three things to know about the White House’s executive order on AI [Link].
Developing a game Angry Pumpkins using GPT-4 for all the coding and Midjourney / DALLE for the graphics [Link].
Chatd: a desktop application that lets you use a local large language model (Mistral-7B) to chat with your documents. It comes with the local LLM runner packaged in [Link].
Teachers in India help Microsoft Research design AI tool for creating great classroom content [Link].
🔍 🛠️ AI Toolbox: Product Picks of the Week
ChatGot: Multiple AI assistants in one app. Supporting GPT4 (32K), Claude V2, Google PaLM 2, Llama, Midjourney etc. Free for now.
Outset: Train the AI interviewer to conduct, synthesize, and summarize interviews.
BeforeSunset AI: AI-powered workspace for professionals and teams. With AI assistant, you can automatically plan your day, make tasks actionable, and create subtasks.
📕 📚 AI Skillset: Learn & Build
A new free course by Andrew Ng: Generative AI for Everyone - Learn how Generative AI works, how to use it in professional or personal settings, and how it will affect jobs, businesses and society [Link].
Top 5 ways to create consistent characters with DALL-E 3 [Link].
Generative AI for Beginners - A 12 Lesson course by Microsoft on building Generative AI applications [Link].
Getting started with Llama - A Llama guide by Meta on how to access the model, hosting, how-to and integration guides [Link].
AI Brews is free, and your sharing it with a friend helps us grow. Thanks for your support and have a nice weekend! 🎉 Mariam.