July 2024 Summaries

AI21 Labs and Snowflake have announced the integration of Jamba-Instruct, a foundation model capable of processing long context prompts effectively without sacrificing performance, into Snowflake Cortex AI. This integration allows users to access Jamba-Instruct within their secured and governed Snowflake AI Data Cloud using a simple SQL statement. The long context window provided by Jamba-Instruct opens up use cases such as long document summarization, call transcript analysis, information extraction, reasoning over long documents, and many-shot prompting.

AI21 Labs' Jurassic-X is a Modular Retrieval-Augmented Generation (MRKL) system that combines multiple models to perform complex tasks, such as answering questions about current events or performing calculations. It can access external databases and APIs for real-time information, making it more versatile than traditional language models like GPT-3. Jurassic-X is designed to be transparent by displaying which expert answered which part of the question and presenting intermediate steps taken. This system has various applications in industries such as finance, healthcare, retail, and customer support.

AI21 and AGI House hosted a hackathon focused on Jamba's 256K context window, attracting over 100 Bay Area developers. The event featured insightful speaker sessions and intense coding sessions, with winning projects showcasing innovative uses of AI and long context capabilities. One project, FBI AGI, developed an AI conversational agent for addressing online harassment on social media platforms. Another project, Excel Mamba, aimed to automate spreadsheet analysis in financial fields. The hackathon highlighted the transformative potential of AI in tackling real-world problems and fostered a collaborative environment among developers and innovators at AGI House.