top of page

Executive Branch Directives Concerning AI

Keyword searching. We’ve all been doing it for a long time, from searching for words in documents to the arrival of AltaVista, Yahoo, Google, Bing, and other search capabilities.  Modern search engines have graduated from keyword to algorithmic searching.  These search engines seek to provide a list of links based on what is most typically selected from the results of previous similar keyword searches made by other people. The product of keyword searching isn’t knowledge per se but, rather, a long, often disjointed list of potential documents or sites containing information that we can then work to translate into knowledge. We’ve all been doing that too.

 

When we use AI, and in this case generative AI, we don’t keyword search. With AI, we engage in a dialog. We pose a query; we ask a normally constructed question, a prompt. The AI responds by digesting the content of all the data it has gathered or has been provided, much like we do after we’ve sorted through all of the documents and links identified in a keyword search. The AI then returns to us knowledge extracted from the data in the form of a normally constructed answer.

 

The query function on this page makes available the aggregate knowledge and perspective expressed within the totality of Executive Branch directives concerning AI. It also somewhat shows just how easy the right platform can make deploying and utilizing the stupendous power of AI, as it took very little effort to add this knowledge source to the PixelRain website. These documents are all in the public domain, and there are no violations of the right to use in this environment.

 

Please explore.

Retrieval Augmented Generation (RAG) 

What is RAG?

Retrieval-Augmented Generation (RAG) enhances large language models (LLMs) by combining text generation with relevant data retrieval. It involves vectorizing data and turning it into numerical representations to quickly find and use specific information from unstructured data sources like documents or from structured data as well. This approach reduces hallucinations and provides more accurate, focused answers compared to non-RAG methods like ChatGPT.

RAG Example

To showcase the power of RAG and its benefits in providing information, the embedded chatbot interface (on this “AI Directives + RAG” page) is linked to multiple AI mandates and standards from the U.S. Executive Branch and federal agencies, including NIST (see Source Documents below for links to the information). These documents are vectorized, allowing users to access specific knowledge embedded within them when interacting with the RAG tool via the chatbot. Furthermore, the PxR platform is designed to run RAG, Advanced-RAG, and other LLM-enhanced tools, providing governance over who can access which tools and what data these tools can provide.

Example Data Sources

Data sources used by this RAG tool include:

bottom of page