AI Search Engine for RAG & AI Agents
PLUS: New AI video generation model, Best opensource model for tool-use
Today’s top AI Highlights:
Groq’s Llama-3 model for advanced tool use and function calling
AI search engine built specially for RAG apps and AI agents
Cohere toolkit now builds AI apps with Interactive HTML and Multi-step Tool-use
New text-to-video model that generates 8 seconds-long videos in 1080p
Python and React AI assistant powered by Claude 3.5 Sonnet
& so much more!
Read time: 3 mins
Latest Developments 🌍
Groq’s Llama-3 Tops Function Calling Benchmarks 🏆
Groq has unveiled two new open-source language models, Llama-3-Groq-70B-Tool-Use and Llama-3-Groq-8B-Tool-Use, specially designed for advanced tool use and function calling. These models are available on GroqCloud Developer Hub and Hugging Face. They are released with the same permissive style license as the original Llama 3 models.
Key Highlights:
Unprecedented Performance - Llama-3-Groq-70B-Tool-Use tops the Berkeley Function Calling Leaderboard (BFCL) surpassing all models with a 90.76% overall accuracy. Llama-3-Groq-8B-Tool-Use achieves an 89.06% accuracy, securing the 3rd position on the leaderboard.
LLM Routing Strategy: Groq suggests a hybrid approach where developers can implement an LLM routing system using the Llama-3-Groq Tool-Use models for function calling or API tasks, and a general-purpose model like Llama 3 70B for other language-based requests.
Open-Source and Accessible: Both the models are readily available via the Groq API using the model IDs “llama3-groq-70b-8192-tool-use-preview” and “llama3-groq-8b-8192-tool-use-preview.”
AI Search Engine for RAG Apps & AI Agents
Exa is a powerful search engine designed for AI developers. Unlike traditional keyword-based search engines, Exa leverages advanced neural search capabilities and a vast, constantly updated index of high-quality web content. This makes it particularly well-suited for RAG apps for retrieving highly specific content, identifying semantically similar pages, and powering research automation tools.
Exa has announced a major upgrade with its 1.5 release, delivering substantial improvements across its platform.
Key Highlights:
Smarter Model - Exa 1.5 is 3x larger than its predecessor and trained with new methods like Matryoshka Representation Learning. It can understand more complex and nuanced search queries to give accurate results, especially when searching for niche information.
Expanded Index - Exa 1.5 features an upgraded index with high-value data types, including scientific research papers, company information, news articles, online writing, and even tweets.
Hybrid Search (Phrase Filters) - Exa 1.5 introduces hybrid search to combine neural search with keyword matching for highly targeted results. For example, search for “discussions about AI” and filter for mentions of “Elon Musk.”
Auto Search with Google Fallback - This intelligent feature automatically determines the best search approach for optimal results. If neural search is insufficient, it defaults to Google keyword search.
AI Apps - Exa API is ideal for a range of tasks, including RAG applications. It integrates seamlessly with tools like LangChain, Typescript, OpenAI, CrewAI, and LlamaIndex.
New Cohere Toolkit Feature for Quick Prototyping and Smarter AI Apps 🧠
Cohere Toolkit is an opensource collection of pre-built components for developers to build and deploy RAG applications quickly. Cohere has expanded this toolkit with new features including HTML rendering, configurable authentication, and multi-step tool use for creating sophisticated AI assistants.
Key Highlights:
AI-powered HTML Generation - You can now ask Command R models to generate interactive HTML applications directly within the Chat UI. With simple text prompts, the model will generate HTML code for basic web components, such as forms, tables, and layouts.
Security with Authentication: You can set up access permissions using email/password authentication, Google OAuth, or OpenID Connect. This ensures secure access to deployed toolkits, especially when dealing with sensitive data sources requiring individual user permissions.
Multi-Step Tool Use for Complex Queries: Cohere has integrated its multi-step tool use capability, previously only available via API, into the toolkit. When the model is given a list of tool definitions, it generates a plan of action and decides which tools to use, populates the required parameters, and defines the order of operations.
Quick Bites 🤌
London-based company Haiper has released its new video generation model Haiper 1.5 which generates 8-second-long videos from text or image prompts. It can even extend your prior 2 and 4-second videos to 8 seconds, just like Luma Labs Extend feature.
Below is a comparison of videos generated by Haiper 1.5, Luma Labs, and Runway GEN-3, for the same text prompt.
”Dragon-toucan walking through the Serengeti.”Not just this, Haiper also has an integrated upscaler that can upscale videos to 1080p in a single click. (Source)
Menlo Ventures and Anthropic have launched the Anthology Fund with a $100 million fund to invest in early-stage AI companies. The fund will provide startups with $100,000 in funding and $25,000 in credits for using Anthropic’s models. (Source)
Microsoft’s Designer app is now available on iOS and Android. The app includes features like AI image editing, background removal, and a variety of templates, and integrates with Microsoft apps like Word and PowerPoint. (Source)
Anthropic has released Claude app for Android. It works just like Claude on iOS and the web. Pick up and continue conversations with Claude across web, iOS, and Android apps. It also supports multimodal inputs, language translation, and advanced reasoning with Claude 3.5 Sonnet. (Source)
😍 Enjoying so far, share it with your friends!
Tools of the Trade ⚒️
ProctorAI: Monitors your computer screen and alerts you if it detects procrastination. It takes periodic screenshots and analyzes them with multimodal AI models like Claude 3.5 Sonnet or GPT-4o to ensure you’re focused on your tasks.
langgraph_streamlit_codeassistant: AI assistant that integrates Python execution capabilities with React component rendering on the fly, offering a comprehensive environment for data analysis, visualization, and interactive web development.
Pieces: On-device AI coding assistant that provides contextual solutions to complex development tasks and manages workflow information. It processes data offline, integrates with popular tools, and uses advanced LLMs to generate code and explain concepts.
Awesome LLM Apps: Build awesome LLM apps using RAG for interacting with data sources like GitHub, Gmail, PDFs, and YouTube videos through simple texts. These apps will let you retrieve information, engage in chat, and extract insights directly from content on these platforms.
Hot Takes 🔥
the biggest barrier to fine tuning LLMs is not cost or modeling or systems expertise anymore, but in collecting high-quality data. it’s hard to do for custom tasks. people try to use gpt 4 as a data generator, which seems ok at a glance but is full of random mistakes at scale ~
Shreya ShankarA lot of SF schizos are just upset that th new X does not prioritize SF schizo content. Discovering that the world doesn’t revolve around you and your niche neurosis can be pretty unsettling. ~
Bojan Tunguz
Meme of the Day 🤡
😅 Math Olympiad becomes easier for AI; Common sense is still hard.
That’s all for today! See you tomorrow with more such AI-filled content.
Real-time AI Updates 🚨
⚡️ Follow me on Twitter @Saboo_Shubham for lightning-fast AI updates and never miss what’s trending!
PS: I curate this AI newsletter every day for FREE, your support is what keeps me going. If you find value in what you read, share it with your friends by clicking the share button below!