Patti is very smart and very friendly. All she wants to do is help people find the information they need. Every day, she sends her spider bots out to find new information that might be helpful.
Patti is a digital person, so she has all the advantages of permanent memory: she remembers everything she reads and, unlike some famous AIs who will remain unnamed, she remembers where she read it and is happy to share.
This makes her like a reference librarian with perfect recall.
Empathi spiders crawl more than just web pages. Given the necessary permissions, Patti sends them to all the approved databases, documents, spreadsheets, presentation decks intranet site pages, and even audio and video content to index their detailed contents in a metaknowledge catalog. It’s kind of like an old card catalog at the library, but so much more advanced because it indexes every word.
Words are the symbols of meaning, but taken out of context, words can become problems. This is why Patti’s deep language understanding is so important. When people ask things that are ambiguous and could be misinterpreted, Patti asks just the right questions to understand the person’s true intent. Then she easily matched it with the meaning in the catalogued digital content and, voila, an answer as trustworthy as the original source – and detailed references to prove it.
Unlike generative AI that uses neural networks, Patti understands everything she reads on the first pass. According to some reports, larger LLMs using GPT learning require thousands of costly and power-hungry high-end GPUs. (see Medium https://towardsdatascience.com/how-long-does-it-take-to-train-the-llm-from-scratch-a1adb194c624).
This costs large amounts of money in four ways:
- You have to find or build a data center with thousands of costly GPUs
- You have to hire experts in a variety of areas to train and refine models
- You have to periodically train a new model from scratch to keep up
- You consume unsustainable kilowatts of electricity each training cycle
Empathi’s unique AI solution solves the mystery of the missing guru: “Where do I find (and how do I pay for) AI experts to train and tune our model?” Provide access to the documents, spreadsheets, documents, web sites and databases and Patti trains herself. She infers the meaning and the context, and as is the case with human analysts sometimes she needs a little help confirming or correcting her inferences. You don’t need AI experts for that. Our dialog-based curation tools enable your Subject-Matter-Experts to ensure the correctness of the model. This is the optimum approach for data stewardship and governance: AI does the heavy lifting and SMEs review and make minor adjustments as needed.
Patti’s knowledge has no cut-off, so you never have to start from scratch. She’s already been learning for years and she simply associates new concepts with concepts she already knows in her knowledge graph or ontology.
Unlike LLMs and other neural networks which store patterns implicitly, Patti’s knowledge is entirely made up of explicit knowledge propositions in context. And she never has to read a book, database, spreadsheet, web page… twice: she understands the whole thing on first read.
Her friendly attitude is a key part of her value to people with questions. She not only provides answers, but she asks appropriate questions to clarify the person’s true needs, then provides detailed references for where they can find additional detail related to their question.
