• 15 Posts
  • 686 Comments
Joined 2 years ago
cake
Cake day: December 9th, 2023

help-circle




  • KEVIN Birmingham’s new book about the long censorship fight over James Joyce’s Ulysses braids eight or nine good stories into one mighty strand.

    It’s about women’s rights and heroic female editors, the First World War, anarchism and modernism, tenderness and syphilis, moral panic and about the Lost Generation and the tent it pitched at Sylvia Beach’s Paris bookstore. It isolates a great love story, that of Joyce and Nora Barnacle, one that comes with a finger-burning side order of some of the most cheerfully filthy correspondence in literary history.

    https://www.scotsman.com/arts-and-culture/the-battle-to-publish-james-joyces-ulysses-1531186

    And what a quest it was. “Ulysses” was illegal to own in most of the English-speaking world for more than a decade. It was banned, burned, debated, smuggled, and finally legalized following a 1933 court ruling. In Birmingham’s highly readable and erudite book, he infuses this story with drama, reminding us that the right to express oneself can never be taken for granted.

    Readers will quickly realize the immense scope of “The Most Dangerous Book.” Modernism, obscenity, the power once held by postal authorities, vice squads, 19th century English law, Joyce’s sex life and health problems, The Lost Generation, early literary magazines, Wall Street lawyers, the suffrage movement, anarchy in America, and even the Enlightenment are all seamlessly woven into this most fascinating tapestry.

    https://www.wbur.org/news/2014/06/13/kevin-birmingham-ulysses




  • No, the AI was most certainly trained on the same stack overflow posts as humans would manually search out in the past.

    Thus the effective difference is precisely that between an active attempt to understand and blindly copying since the AI is specifically there to introduce a stochastic opaqueness between truth (i.e. sufficiently curated training data) and interpretation of truth.

    There is a context to stackoverflow posts and comments that can be analyzed from many different perspectives by the human brain (who posted the question with what tone, do the answers/comments tend to agree, how long ago was it posted etc…), by definition the way LLMs work they destroy that in favor of a hallucinating-yet-authoritative disembodied voice.