This article has many earmarks of an AI type who has just suffered from a stroke.
The history of AI is embarrassing. Too many researchers with lots of ego and little fundamental justification shooting off at the mouth. The field is/was valid, but from a public perspective, it was poisoned by idiots. That it attracted more than its share of eccentric cranks didn't help. Research hasn't stopped, but most everyone keeps busily quiet now that nobody pays attention any more.
That said, a LOT of progress has been made in recent years with respect to AI, fundamental progress, to the extent that we've done more in the last five years than the previous forty. While there isn't any fanfare, the puzzle is finally starting to unravel, thanks in large part to some important mathematical breakthroughs that allowed the fundamental problem to be characterized. But since most researchers are gun shy these days, you won't hear about any kind of AI until it is extremely mature. That the current state of the art defies intuition and really can't be easily explained to anyone but other mathematicians in the appropriate fields despite being extremely elegant doesn't help either. At least things like "neural nets" could be summed up in a catchy soundbites.
I will say, as a mathematician tangentially associated with the AI community, that the capabilities developed in the last few years far exceed what you probably think is possible, and that it uses algorithms/technologies that essentially no layman has ever heard of.
P.S. The speed of your processor has little to do with why your computer is so stupid. That "we don't have computers fast enough" is a common misperception; CPU speed has nothing to do with intelligence except in the most pathological case. The bottleneck in hardware performance is elsewhere.
After the AI people went underground, a new breed of structural engineering students arose in the 1990's. They were promising a revolution in engineering analysis where expert systems would guide a design engineer through a project. The engineer would interact with the program answering queations, whatever that meant, and the program would knowingly come up with an optimized design all the while meeting the needs of complex set of engineering rules. Graduate students were writing their theses on this subject. Then the term fuzzy logic made it into the lexicon of structural engineering academics. All of a sudden, these proponents appear to have gone underground. Im still working with a stupid finite matrix analysis program that has an up-to-date windows interface. As time wears on, I become increasing impressed with how much our intuition guides our lives and thinking and how little that rigorous logic no matter how fuzzy is ever applied.