'Swift' AI drone beats human champions, Watermarking tool for AI-generated images, a16z Open Source AI Grant program, Runway's Creative Partners Program 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 #29):
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
Researchers introduce ‘Swift’, the first autonomous vision-based drone that beat human world champions in several fair head-to-head races. This marks the first time that an autonomous mobile robot has beaten human champions in a real physical sport [Details].
Generative AI updates from Google Cloud Next event:
General availability of Duet AI in Google Workspace [Details].
SynthID - a tool for watermarking and identifying AI images generated by Imagen (Google’s text-to-image diffusion model). It embeds a digital watermark directly into the pixels of an image, making it invisible to the human eye, but detectable for identification, without reducing the image quality [Details].
AlloyDB AI for building generative AI applications with PostgreSQL [Details].
Vertex AI’s Model Garden now includes Meta’s Llama 2 and TII’s Falcon — and pre-announcement of Anthropic’s Claude 2 [Details].
Model and tuning upgrades for PaLM 2, Codey, and Imagen. 32,000-token context windows and 38 languages for PaLM 2 [Details].
Style Tuning for Imagen - a new capability to help customers align their images to their brand guidelines with 10 images or less [Details].
Launch of fifth generation of its tensor processing units (TPUs) for AI training and inferencing [Details].
Meta AI released CoTracker - a fast transformer-based model that can track any point in a video [Hugging face | GitHub].
WizardLM released WizardCoder 34B based on Code Llama. WizardCoder-34B surpasses GPT-4, ChatGPT-3.5 and Claude-2 on HumanEval Benchmarks [Details].
Meta AI introduced FACET (FAirness in Computer Vision EvaluaTion) - a new comprehensive benchmark dataset for evaluating the fairness of computer vision models for protected groups. The dataset is made up of 32K images containing 50,000 people, labeled by expert human annotators [Details].
Allen Institute for AI launched Satlas - a new platform for exploring global geospatial data generated by AI from satellite imagery [Details].
A new generative AI image startup Ideogram, founded by former Google Brain researchers, has been launched with $16.5 million in seed funding. Ideogram's unique proposition lies in reliable text generation within images [Details].
a16z announced a16z Open Source AI Grant program and the first batch of grant recipients and funded projects [Details].
Runway AI announced Creative Partners Program - provides a select group of artists and creators with exclusive access to new Runway tools and models, Unlimited plans, 1 million credits, early access to new features and more [Details].
OpenAI has released a guide for teachers using ChatGPT in their classroom—including suggested prompts, an explanation of how ChatGPT works and its limitations, the efficacy of AI detectors, and bias [Details].
DINOv2, a self-supervised vision transformer model by Meta AI which was released in April this year, is now available under the Apache 2.0 license [Details | Demo].
Tesla is launching a $300 million AI computing cluster employing 10,000 Nvidia H100 GPUs [Details].
Inception, an AI-focused company based in the UAE unveiled Jais, a 13 billion parameters open-source Arabic Large Language Model (LLM) [Details].
Google announced WeatherBench 2 (WB2) - a framework for evaluating and comparing various weather forecasting models [Details].
Alibaba launched two new open-source models - Qwen-VL and Qwen-VL-Chat that can respond to open-ended queries related to different images and generate picture captions [Details].
OpenAI disputes authors’ claims that every ChatGPT response is a derivative work [Details].
DoorDash launched AI-powered voice ordering technology for restaurants [Details].
OpenAI launched ChatGPT Enterprise. It offers enterprise-grade security and privacy, unlimited higher-speed GPT-4 access, longer context windows for processing longer inputs, advanced data analysis capabilities and customization options [Details].
OpenAI is reportedly earning $80 million a month and its sales could be edging high enough to plug its $540 million loss from last year [Details].
🔦 Weekly Spotlight
How 3 healthcare organizations are using generative AI [Link].
The A.I. Revolution Is Coming. But Not as Fast as Some People Think [Link].
LIDA by Microsoft: Automatic Generation of Visualizations and Infographics using Large Language Models [Link].
Curated collection of AI dev tools from YC companies, aiming to serve as a reliable starting point for LLM/ML developers [Link].
Beating GPT-4 on HumanEval with a Fine-Tuned CodeLlama-34B [Link].
🔍 🛠️ AI Toolbox: Product Picks of the Week
Cursor: an AI-powered code editor. Cursor is a fork of VSCode for pair-programming with AI.
You: The AI-powered search and chat app has introduced a new plan for students and teachers — access unlimited AI chat + search, unlimited AI writing generations, unlimited AI art generations, supercharged with GPT-4 and Stable Diffusion XL at $6.99/month (a valid school email required).
Ideogram: A generative AI app based on a new foundation AI model that turns text into images. It supports many fonts and text styles and supports a diverse set of image style tags like Illustration, Typography, Poster, 3D Render, Architecture, Graffiti etc.
📕 📚 AI Skillset: Learn & Build
FAQ/Guide by OpenAI for educators and students using ChatGPT [Link].
How do domain-specific chatbots work? An Overview of Retrieval Augmented Generation (RAG) [Link].
Deploying Open Source LLMs for RAG with SageMaker [Link].
Thanks for reading and have a nice weekend! 🎉 Mariam.