In his blog, Paul Romer mentions a book ‘Why information grows’ and take some time to review how differently the physicist and the economist may think about that resource called information and its storage.
In his demonstration, codified information (i.e. information that can be stored using hardware devices) does not have the same characteristics than information that is stored in our brain (people like you and me would probably call that human knowledge).
What is disputable is that Paul Romer assumes that human stored data cannot be extracted unless you use torture or other painful methods. Every body in the security field would probably object as there is a engineering science which has this very goal of extracting data from people without them to realize it and suffer at the time the extraction is made: that’s called social engineering.
Although it doesn’t invalidate the demonstration in its entirety, it would change some of it by saying than brain-stored information can be extracted at a cost which is different from other storage, and more importantly at a cost that greatly varies from one situation to another and is therefore likely to invalidate the economical model in edge situations.