If Computers Could Write

I have many titles.  Gentleman scientist.  Consultant.  Husband.  Dad.  Some are self-applied (the advantage of being my own boss), and some are earned.  One that I am proud of, and take seriously, is the title of “writer”.  Writing well is not easy, and I have the somewhat old-fashioned idea that I should only write if I have something worthwhile to say.  So when I do write something, be it a blog post or a textbook, I take some pride in it.

But what if, in today’s world of high performance computing and Jeopardy-winning algorithms, a computer could be taught to do what I am doing now – to write?  Simpler than the full-blown Turing Test, a writing computer certainly seems possible.  But could a computer catch my interest?  Inform me and intrigue me?  Keep me reading?  What might the result be like?

Based on empirical evidence, I know the answer.

Crap.

Computers can’t write worth crap.

Granted, this is my opinion, and I suspect that Dr. Philip M. Parker would disagree.

Professor Parker is an economist who describes himself as a pioneer in “automated authoring processes”.  His work on computer authoring has resulted in one patent (US Patent #7,266,767, Method and apparatus for automated authoring and marketing) and over 200,000 book titles (more than 100,000 of which are available on Amazon).  He says he has authored hundreds of thousands of poems using graph theory (I don’t even want to know what that means).  Many of his titles use the “Webster” name to give it an imprint of authority, though the Webster name is in the public domain and in fact means nothing.

The basic idea is simple:  create an application-specific template, fill it in with web-searched data, then apply some automated copy-editing rules.  Combine this with print-on-demand, and viola.  A hundred thousand books on Amazon (a large portion of which, I suspect, have never been read by a human).

I ran across this interesting and bizarre idea while searching on Amazon recently and coming across a title that intrigued me:  Microlithography: Webster’s Timeline History, 1975-2007.  The title sounded great, but the author was unfamiliar to me.  What could it be?  Since I have a lithography timeline of sorts on my website (http://www.lithoguru.com/scientist/lithohistory.html), I wanted to know.  It cost me $28.95 to find out, and I am now on a mission to make sure that no one else will have to waste their time and money the way I did.

I’m not sure what I thought a “Timeline History” was, but in Dr. Parker’s automated hands it is simply an ordered list of publications containing the keyword (Microlithography, in this case).  And not a very good list, either.  The formatting varies from entry to entry, with each item largely unidentified (Is it a book?  A journal article?  A Master’s Thesis? A conference proceedings?) and often with insufficient information to actually find the item without Google’s help.  To get a feeling for what is there, here are some stats.

The book has 347 entries, of which 22 are duplicates.  The majority of the unique entries are patents (325, 70%), most of which include abstracts but none of which include patent numbers.  Without these entries, the book would only be a few pages long.  The rest are books, journals and conference proceedings (67, 20%), technical reports found on webpages (15, 5%), MS and PhD theses (10, 3%), individual peer-reviewed articles (5, 1.5%), and an encyclopedia entry (1, 0.3%).

What am I to make of these numbers?  Are there really only 5 peer-reviewed articles on microlithography between 1975 and 2007?  Only 10 MS and PhD theses?  I have in my office far more than 67 books, conference proceedings and journals on microlithography.  And what about the patents?

While patents make up the majority of the entries, 325 is closer to the number of microlithography patents issued in a few months, rather than over a 32 year period.  A quick search of patents issued between 1975 and 2007 (using Google Patents) with the keyword “microlithography” turned up 7,300 patents.  There are 2,860 patents with microlithography in the title.  If you add “photolithography” to the keyword search, there are 29,900 hits, rising to 33,100 when “optical lithography” is added to the keywords.  I’m not sure what value the 325 patents (less than 1% of the total) contained in this little book might provide a reader.

The bottom line is this: Microlithography: Webster’s Timeline History is a waste of time, a waste of money, and a waste of print-on-demand paper. I suspect that the full range of Philip M. Parker’s computer-generated books have equal value.

But hey, I got a blog post out of it.  And it was entirely human-written.