General A.I. is science fiction. For any A.I. to be accurate the problem must be confined, the right core capabilities selected, and then trained within a specific domain
Domain experts define the problem, data, and desired outputs for a product
Technologists determine the right mix of components needed to produce these outputs
Domain experts train the system with a small set of examples that constitute a small percentage of all available data
The expert examples form a semantically rich training data-set the AI consumes to learn about the domain initially
With periodic expert feedback and continual hypothesis testing, the system keeps improving its understanding of the domain
Over time, the accuracy increases and ultimately the value to end users
Creating the domain layer and plugging it into the A.I. is a fast process. This is only made possible due to the modular design of the core capabilities and a robust scalable process.