Whole-Brain AI

Empathi AI uses a large Bayesian Network of concepts in context to interpret natural language input and resolve ambiguity. Each context and concept effectively resides at the center of a constellation of related concepts that describe how the words fit together to represent things that are and things that happen. The symbolic modeling we use captures both obvious and nuanced phenomena with fuzzy reasoning where each discrete knowledge proposition has a confidence value that represents its likelihood in the given context. This is a whole-brained approach because it is holistic.

Neural network technologies are great at two dimensional problems like speech recognition and optical character recognition. As the size of the problem increases, the computational resources required explode. Witness Large Language Model (LLM) solutions. They are grotesquely expensive to build and run because they try to use simple patterns to achieve complex results. Artificial Neural Networks model left-brained thinking very well.

Our suite of tools will be used to add music, film and storytelling to the cognitive toolchest Empathi uses to provide insights to people with complex needs.

Language and real-world knowledge are inextricably connected, and neither functions well without the other. That is why natural language processing initiatives focused exclusively, or even primarily, on language structure have significant limitations. Grammar and syntax are necessary for understanding how language works, but they are rarely sufficient for determining a person's intent.

If the goal is language comprehension approaching human competence, then knowledge-based approaches that can resolve ambiguity based on meaning in context are essential. This calls for a universal theory of knowledge suitable for supporting knowledge-based language understanding — one that begins with defining concepts and the language symbols used to express them.

For a computer to be able to engage in a dialog with a human, it needs to have good language skills and large amounts of knowledge. To bring maximum value to the human in this interaction, it should adapt to the human rather than forcing the human to adapt to the computer.

In a way, Large Language Models implicitly replicate the parts of the body of human knowledge they are trained on, as patterns in a neural network. Empathi AI does so explicitly.

Large Language Models such as ChatGPT have an amazing capacity to learn things computers have never been able to achieve before. That capability comes at a huge cost, especially for outcomes specific to your business. Organizations with massive IT budgets can afford the ecosystem needed to train, manage and use LLMs. Smaller companies cannot.

Besides the cost of getting started before any business benefit is possible, there is another hidden cost. The “P” in GPT means that it learns everything at once and cannot add incremental knowledge without complete retraining.

Our solution works differently. Any source that can provide new knowledge such as a social media feed, streaming log data from manufacturing equipment or transactions from an ERP can be ingested and added to the central knowledge catalog incrementally. The metadata catalog supports self-service analytics and search. We provide curation tools so Subject Matter Experts or data stewards can ensure that new knowledge is classified correctly. Once classified correctly (or understood) it can be matched with inquiries immediately and used in verbal and visual insights for qualitative and prescriptive intelligence.

The Knowledge Catalog contains proprietary knowledge controlled by each customer and never shared. The Catalog works with the General DLU Knowledge Model to bring in Empathi's deep conceptual knowledge of the natural associations between all the digital assets a business uses to compete and grow. As not all users can access all data, the metadata model of the Denims Catalog uses tags for least privilege access management.