What is Deep Dive and how does it work?

Deep Dive essentially is AI document analysis on steroids. It excels at analyzing long documents and returning detailed, comprehensive answers to questions.

Deep Dive is a proprietary algorithm that builds a second enhanced memory for the AI.

A chat must be either fully Deep Dive enabled or not. If you'd like only a subset of documents to be Deep Dive enabled, please create another chat with just those documents and Deep Dive enable that chat.

After a chat is Deep Dive enabled, Socrates automatically determines whether to use the typical memory or second enhanced memory to answer the query. The typical memory performs better with focused queries while Deep Dive performs better with broader queries. Deep Dive queries typically take longer to process than non-Deep Dive queries.

You can toggle this auto-Deep Dive mode in the keyboard and force a chat to only use Deep Dive or force it to NOT use Deep Dive, assuming the chat is Deep Dive enabled.

How does it work?
Typical Q&A AI systems look through the data using one method (typically RAG) before returning a response. We not only use RAG but also a combination of other methods to return the best response.

It does cost us a significant amount on the backend to implement this combination of methods, which is reflected in the Socrates pricing plans.