This is the multi-page printable view of this section. Click here to print.

Return to the regular view of this page.

News

DfocusGPT Copyright

DfocusGPT leverages LLMs for everyday enterprise workflows.
By Jungho Lee (@DfocusSns) |

Copyright information for DfocusGPT - Korea Copyright Commission

Copyright

LLM History and People

A brief look at the history of LLMs.

In October last year, key figures in AI gathered at MIT. See the discussion: Research on Intelligence in the Age of AI.

Deep Learning History

Video overview (Korean)

DfocusGPT Data Flow

Basic question-answer flow in DfocusGPT.

Below is a basic sequence for answering user questions with company knowledge and LLM.

sequenceDiagram
    autonumber
    loop History

    User->>Chatbot: Question

    alt History exists
        Chatbot->>Chatbot: Generate Reinforced Query (Q + History)
    end

    Chatbot->>Company KB: Search (Reinforced Q)
    Company KB->>Chatbot: Results
    
    Chatbot->>LLM: Prompt (Q + KB + History)
    LLM->>Chatbot: Answer

    Chatbot->>User: LLM answer + KB
    Note right of Chatbot: Store Q&A

    end

Details may vary by options.