Given that large language models are designed to guess the next word and fill in the gaps in strings of text, it shouldn’t come as a big surprise that ChapGPT has been enlisted to try to unredact partially declassified documents. Of course, it would be difficult to impossible to check the accuracy of the AI since that information has not been released. What is surprising, however, is that users seem to primarily if not exclusively using the these capabilities to read censored names and locations on documents from NASA on UFO sightings—not that that isn’t an exciting topic worthy of pursuit, at least it used to be, until congressional hearings that seemed to be directed by The History Channel (a rather low information operation posing as educational television) turned apparent government secrecy about the nature of unidentified aerial phenomena into a seminar for the aggrieved in general. I wonder what happens when someone takes on more consequential redactions and what that might mean for future disclosures.
synchronoptica
one year ago: Denmark plans a Synthetic Party led by an AI, the first rail route in Japan (1872), more radio calling cards plus a song from English Beat
two years ago: Faust (1926) plus the architecture of Hรฉlรจne Binet
three years ago: special meal requests plus more natic movements in plants
four years ago: nominees for Word of the Year, Germany’s Mushroom of the Year plus New York City through an AI lens
five years ago: the world’s first motion picture (1888), Apollo 7 transmits from the Moon (1968), The Watersons, The Bells of Rhymney plus diplomatic tensions between the US and Turkey
