my hunch is that Cyc still has about 20% coverage for the probability density of meaning in text... just not quite enough.
Just increasing the size by 5x won't do because it's a problem like modelling human hair, at some point the snags, knots and frizz will just drive you insane.
That was my blog article. We are getting better tools for knowledge management such as RDFS/OWL, lots of linked data sources, perhaps (Open)Cyc, and companies like Google demonstrating great things that can be done with statistical NLP to solve tough problems like machine translation.
All that said, I think we have a long ways to go before we can build 'real' AI real world knowledge based systems that understand natural language text.
Just increasing the size by 5x won't do because it's a problem like modelling human hair, at some point the snags, knots and frizz will just drive you insane.