Stable Video , Speech-to-Speech, Llama Packs, 40x faster speech-to-text, 200K token context window, ZipLoRA 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 #41):
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
Stability AI released Stable Video Diffusion, a latent video diffusion model for high-resolution text-to-video and image-to-video generation. [Details | Paper].
Microsoft Research released Orca 2 (7 billion and 13 billion parameters), open-source models created by fine-tuning the corresponding LLAMA 2 base models on tailored, high-quality synthetic data. Orca 2 significantly surpasses models of a similar size, even matching or exceeding those 5 to 10 times larger, especially on tasks that require reasoning [Details].
Researchers from Google andUIUC present ZipLoRA, a method to cheaply and effectively merge independently trained style and subject LoRAs in order to achieve generation of any user-provided subject in any user-provided style [Details Implementation ].
Inflection AI, the startup behind the chatbot Pi, announced that it has completed training of Inflection-2 claiming it to be the 2nd best LLM in the world [Details].
Anthropic updated and released Claude 2.1 having 200K token context window, a 2x decrease in hallucination rates and system prompts. It is available now via API, and is also powering the chat interface at claude.ai for both the free and Pro tiers [Details].
Researchers from UC Berkeley released Gorilla OpenFunctions, an open-source function calling model. Gorilla OpenFunctions is a drop-in open-source alternative. Given a prompt and API, Gorilla returns the correctly formatted function call [Details].
Deepgram introduced Nova-2 model for speech-to-text which delivers +18% accuracy than Nova-1 & over 36% accuracy than OpenAI Whisper large while being 5-40x faster compared to alternatives [Details].
LlamaIndex introdcded Llama Packs — a community-driven hub of prepackaged modules and templates to making building an LLM app for any use case easier [Details].
Google is open sourcing Project Guideline, a platform for computer vision accessibility [Details].
Google’s Bard AI chatbot can now answer questions about YouTube videos [Details].
Amazon aims to provide free AI skills training to 2 million people by 2025 with its new ‘AI Ready’ program which includes eight new and free AI and generative AI courses and AWS Generative AI Scholarship to 50,000 students globally with access to a new generative AI course on Udacity [Details].
SynthID, a tool by Google DeepMind for watermarking and identifying AI-generated content, can now watermark AI-generated music and audio [Details].
xAI’s chatbot ‘Grok’ will launch to X Premium+ subscribers next week [Details].
🔦 Weekly Spotlight
AI Exploits: A collection of real world AI/ML exploits for responsibly disclosed vulnerabilities [Link].
A timeline of the OpenAI saga with CEO Sam Altman [Link].
RAGs: a Streamlit app by LlamaIndex to create and customize your own RAG pipeline and then use it over your own data — all with natural language [Link].
🔍 🛠️ AI Toolbox: Product Picks of the Week
AI Speech to Speech Converter by Eleven Labs: a new voice conversion tool that lets you turn the recording of one voice to sound as if spoken by another. It lets you control the emotions, tone, and pronunciation beyond what's possible with TTS prompts alone.
Magnific AI: Magnific enhances images to higher resolutions, letting you dictate the upscaling with descriptions and controls like 'Creativity', which adjusts the AI's generated details.
Screenshot to Code: an open-source app that converts a screenshot to HTML/Tailwind CSS. It uses GPT-4 Vision to generate the code and DALL-E 3 to generate similar-looking images.
📕 📚 AI Skillset: Learn & Build
Intro to Large Language Models by Andrej Karpathy. This video is based on the slides of a talk he gave recently at the AI Security Summit [Link].
Getting started with Llama by Meta - Quick setup and how-to guide [Link].
Quality and Safety for LLM Applications - A free short course by DeepLearning.AI [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.