Did a quick scan of this article, and I’m not that impressed. It looks like an expert system (which have been around since the 70s) with JMS to facilitate message passing. No big deal. There are a lot of technologies that have been refined over the years that are peripheral to true Artificial Intelligence (AI). Some (to name a few) are speech recognition, natural language understanding (to a limited degree), pattern recognition, and image processing. Some approaches, like Genetic Programming, can even emulate creativity by evolving new programs to solve complex problems.
But, in the end, the tough nut to crack in AI is being able to incorporate and interpret new knowledge in an existing knowledge base. This involves being able to take an observation (no matter how simple), understand its context and successfully relate it to other facts and rules. Doing this will involve creating new rules and/or intermediate (inherent) facts to be able to use the new knowledge in a flexible manner. You can incoporate the fact by itself, but unless you integrate it correctly, it will be brittle and become just so much garbage in memory. It would be as if you saw a piece of paper on the floor and just noted that fact without giving it context or trying to analyze what the paper was about. As such, you might mistake it as scrap paper when, in fact, it was part of a critical document.
This type of learning is really the Holy Grail of AI and it is the hardest thing to do. Humans do it all the time, we just have no clue how. In the end, in order to understand a little about something, you have to know a lot about everything. Humans gradually pick up and assimilate this knowledge during childhood. We have no such patience with computers.
If computers ever really learn how to learn, then watch out.
Skynet.
There was a scientist who was asked, “Will there ever be a computer as intelligent as a human?” He answered, “Yes, but only for a few moments”.
Can it be made to work with binary logic and in anything close to real time?