He never says any of these problems are fundamentally insurmountable. Just that you need to take any meta-data with a large grain of salt. Which, well, is exactly what Google does. It uses certain metadata, but doesn't trust it absolutely.

Also, he mentions Google in this essay; this was written after the emergence of Google, when Google was trouncing all of these other services that tried to use metadata embedded in webpages or tried to classify all websites into some big hierarchy.