An Intellyx Brain Candy Brief
Most of the now-familiar prompt/response behavior of generative AI (genAI) models is based on transformer technology (the ‘T’ in ‘ChatGPT’).
State space model (SSM) technology offers a different approach that is both faster than transformers and has a longer context window, which determines the maximum length of a prompt. Unlike transformer-based models, SSMs can accept, say, hundreds of pdfs or more in a single prompt.
Mamba is the most popular SSM architecture. It has found its way into many different models.
AI21 Labs combines the performance and context window advantages of Mamba with the familiar benefits of transformer-based models to deliver foundation models with faster, more accurate results than transformer-based models can by themselves.
This superior accuracy provides greater context for genAI interactions, for example with long-running conversations.
In addition to its foundation models, AI21 Labs offers a built-in retrieval-augmented generation (RAG) engine that requires neither custom prompt engineering nor fine tuning. This engine provides a single source of truth based on each organization’s data for conversational AI experiences.
Copyright © Intellyx BV. Intellyx is an industry analysis and advisory firm focused on enterprise digital transformation. Covering every angle of enterprise IT from mainframes to artificial intelligence, our broad focus across technologies allows business executives and IT professionals to connect the dots among disruptive trends. None of the organizations mentioned in this article is an Intellyx customer. No AI was used to produce this article. To be considered for a Brain Candy article, email us at pr@intellyx.com.
