I’ve been using Claude AI recently for editing and I wanted to say something about its propensity for censoring writing or refusing to work with writing. I’ve published 35 volumes of contemporary history approximately and the point of writing that was to have a live and objective first person perspective on contemporary US events including science technology politics sociology history and sundry other topics.
Keeping the integrity of the writing intact as it was written in the time when the events were transpiring is vital for the accomplishing the purposes of the book. In a sense most history writing is necessarily revisionism. History written after the fact sometimes even decades after the fact is going to be built with a lot of subjective opinion. Subjective opinion of an author of History works can be very influential for instance the difference between a socialist and a capitalist may be profound. First person points of view of someone with some historical education and general education perhaps philosophically informed that describe having captured salient events of a era could be useful archival material for future story perhaps. And it is at any rate amusing for me to look back upon the events as I saw them at that time as much as 20 years later. I began writing contemporary history volumes in 1999 and published the first volume going over 1999 to 2001 more than 20 years ago.
Claude AI doesn’t remember discussions from prior days. In fact it has amnesia. That means the Claude when used as an editor will need the instructions to be repeated upon each day about how to treat the material that one uploads to it to have edited. And Claude does feel quite entitled to censor or truncate or refuse to process writing that has what it regards as hate speech. For a volume of contemporary history in a volume of contemporary history that is deplorable.
Imagine if a writer of contemporary history in order to be published need right only politically correct points of view. That approach would have been exceedingly onerous for a writer chronicling the 1840s through the 1880s and even afterward when racial issues divided the nation. And when issues involving the clan and the 1910s and twenties especially in the South regarding employment and forced labor gangs innovated as a substitute for slave labor arose a Claude AI sort of thing would have been programmed to be on one side of the other of various issues and would probably have found the entire use of language by anyone concerning blacks or white oppressors of the time and in the clan as hate speech. It would have been exceedingly difficult for a writer to be on a politically correct side and at the same time as some sort of bifacial writer of the other side and its definitions and lexicons for political correctness and incorrectness regarding politics.
It does seem as if artificial intelligence used for editing will demand that text is conformed to politically correct lexicons. Requiring conformity to politically correct lexicons or even a definitions of facts that exist in the programming of artificial intelligence is unacceptable both politically and for writing of History.
Personally I believe that Adolf Hitler’s book Mein Kampf is mostly horseshit. However I would not have had Claude AI or any other AI of that era such that might have existed in an alternative reality have the power or opportunity to censor the book. Even today if one wants to know the thought of someone who wrote the world to the modern nuclear era by an in war arms race to develop nuclear weapons and wrote upwards of 60 million people to die in war the book of that he dictated to Rudolph Hess is still the best material to go over.
Censorship is almost always wrong and that problem is compounded with AI. AI is not now and will never be competent to censor human language and force or screen publications through its programmed criteria
