Welcome back!

Elon Musk's xAI has taken the AI world by surprise with the sudden release of Grok-2, a frontier model that competes with top-tier AIs like GPT-4o and Claude. See how this unfolds…

Today’s Summary:

• 🚀 xAI unveils Grok-2 with image gen

• 🧠 Snowflake launches AI data analyst Cortex

• 📊 MIT releases AI Risk Repository

• 🎙️ WhatsApp tests AI voice chat

• 🌐 Opera iOS browser adds AI features

• 💰 Claude prompt caching cuts API costs by 90%

• 2 new tools

TOP STORY

xAI releases Grok-2, adds image generation on X

The Summary: Elon Musk's AI Company xAI has released Grok-2 and Grok-2 mini in beta, bringing improved reasoning and new image generation capabilities to X. Available only to X Premium and Premium+ users, Grok-2 joins the ranks of top AI models like GPT-4, Claude, Gemini, and Llama.

Key details:

• Grok-2 outperforms Claude 3.5 Sonnet and GPT-4-Turbo on the LMSYS leaderboard, but trails behind GPT-4o and Google Gemini Pro

• Both models will be available through an enterprise API

• Grok 2 is strong in visual math reasoning and document-based question answering

• The image generation features are powered by Flux, not Grok directly

Why it matters: The introduction of Grok-2 with image generation capabilities has the potential to transform user interactions on X and could significantly impact the social media landscape. This sudden push by Elon Musk will further increase the competitive pressure in the AI field.

DATA ANALYTICS

Snowflake launches Cortex, an AI-powered Data Analyst

The Summary: Snowflake, the data-as-a-service cloud company, has introduced Cortex Analyst, an AI system for self-service analytics. Users can ask business questions in plain English, and the system converts them into database SQL queries delivering accurate answers. Snowflake claims Cortex achieves 90% accuracy, surpassing other text-to-SQL offerings. The system uses multiple AI agents working together to ensure reliable insights.

Key details:

• Combines multiple LLM agents for improved accuracy

• Requires semantic descriptions of data assets during setup

• Available as REST API for integration into applications

• Uses models Snowflake Arctic, Mistral, and Meta AI

Why it matters: This development could make data analytics more accessible to non-technical business users. By allowing natural language queries with high accuracy, Cortex Analyst may speed up decision-making and reduce the workload on data analysts.

Behind the scenes

AI SAFETY

MIT releases AI Risk Repository, a detailed database of AI risks

The Summary: MIT researchers have developed the AI Risk Repository, a comprehensive database of AI risks. The repository consolidates information from 43 existing taxonomies. It is intended to be a living database, regularly updated with new risks and insights.

Key details:

• Database includes 700+ unique AI risks from 43 sources

• Risks by cause (entity, intent, timing) across 7 domains

• Previous frameworks covered only 34% of subdomains identified

• Publicly accessible and downloadable for organizational use

• Functions as a checklist for risk assessment and mitigation in AI development

Why it matters: This repository provides a unified framework for understanding AI risks, addressing the previously fragmented landscape of classifications. It offers a solid foundation for more targeted research and risk management strategies, AI governance, and safety practices.

Research paper

QUICK NEWS

Quick news

• WhatsApp working to add AI voice chat mode for Meta AI

• Opera iOS browser adding AI features

• Claude Prompt Caching can reduce API costs up to 90%

TOOLS

🥇 New tools

• Omnifact - Ultra-secure AI assistants using internal knowledge

• Postgres.new - In-browser database with AI assistance