What they demonstrated as AI in 1984 today is considered business rules engine. There are plenty of them and they are not really considered "intelligent" (because they cannot learn by themselves).
On the other hand today we have really intelligent tools such as spam filters, spam bots, search engines, fraud detection tools, language translators, rating engines and many other that are important part of our lives.
I think it depends on who you ask, since it's somewhat of a subjective, philosophical question. There are definitely sub-areas of AI working on rule-engine type stuff, especially if it gets more complex, e.g. various stuff derived from Datalog or Prolog.
On the contrary side, there are actually people who consider a lot of spam-filtering and other statistical-ML tools to be more applied statistics than AI, though most people probably have more mixed views on it. To some extent they're just straightforwardly applying algorithms to get outputs with known mathematical properties from inputs. Sort of the way a sort algorithm takes in data and always gives you an output with specific properties, statistical algorithms take in data and always give you an output with specific pre-determined properties, which just happen to be more complex ones. But they never modify their own processes, for example, or change goals, or realize that they're making the same mistake over and over, or decide that this algorithm is obsolete and needs to be improved--- that only happens when the intelligence in the loop (the human statistician) updates the algorithm.
On the other hand, both approaches (rules-ish or statistics-ish) often produce results that 50 years ago would definitely have been considered to "require intelligence".
Yeah, it's entirely possible I don't understand what they were doing but that first example (the "Sunday" program), just seems like a lot of `if` statements.
