No. It’s another kind of Internet-enabled AI.
The current excitement in AI is all about machine learning and that name perfectly explains its appeal: machines learning to behave ‘intelligently’ without being explicitly programmed. But machine learning is only able to perform specific tasks for which large volumes of high quality learning-by-example data is available (or can be generated by simulated trial-and-error). Successful machine learning projects “follow the data” and ignore the many applications for which sufficient quality learning data is not available.
Product support is a great example of this. Support organisations have large knowledge bases of past questions and answers, but staff expertise and answer quality varies widely and training a machine learning AI on horrible (low quality) knowledge yields horrible results. Using machine learning to add flexible natural language interfaces to low quality knowledge bases is even worse.
eXvisory is another kind of Internet-enabled AI: capturing expertise (rather than experience) in language-like logic networks. Like neural networks (the precursor to machine learning) logic networks (aka expert systems) fell out of favour before the Internet, because they required specialist programmers and suffered from combinatorial complexity. But just as Internet-scale data and compute have reopened the flood gates for machine learning, modern Internet capabilities are also enabling a fresh take on logic networks.
eXvisory provides graphical web editors and a palette of logic nodes to enable regular programmers or senior support staff (who develop the scripts for junior support staff) to encode existing support expertise into maintainable logic networks, overcoming the complexity problems that plagued expert systems. These deep logic networks automatically generate conversational AI support assistants that act as curated guides through the best support articles on the Internet, while leveraging easy web access, automated translation and social sharing.
Copyright © 2019 eXvisory | version