Where is Artificial General Intelligence?

We expected Artificial General Intelligence (AGI) to follow quickly when we formed an artificial intelligence team and lab in the late 1980's.  Surely once we started to build incremental intelligence into the expert systems of the day they would soon be interconnected into 'general' intelligence that would guide the enterprise forward.  Yet that did not happen. Five years later our AI team was disbanded.  Despite some successes, they were all in focused, limited domains.   We invested in companies  that looked like they would become giants of AI, like Teknowledge.  But these companies also soon sank into limited domains.   Even successes like IBM's Watson are mashups of the ideas we used then.  What happened?  Why have there been no advances for AGI  in decades?

Discussed in Kurzweil AI.   Good thoughts.  One answer:  The hardware has not caught up yet.  Despite all the suggestion that the brain's neural methods would soon be replicable, it has not happened yet.