Post AtmYuaDBRB5SHS9U6y by peterainbow@beige.party
 (DIR) More posts by peterainbow@beige.party
 (DIR) Post #AtmTdcLljYgeDeG62C by cstross@wandering.shop
       2025-05-05T10:22:13Z
       
       0 likes, 1 repeats
       
       The LLMentalist effect: Large Language Models replicate the mechanisms used by (fake) psychics to gull their victims:  https://softwarecrisis.dev/letters/llmentalist/
       
 (DIR) Post #AtmUBfkZbBqabt6DHE by futurebird@sauropods.win
       2025-05-05T10:32:39Z
       
       0 likes, 1 repeats
       
       @cstross An expert on the Voynich manuscript whose content I enjoy was using an LLM for translation.  (He mostly posts interesting observations about the statistics of the symbols) So, not translating the manuscript itself by something in Turkish. Amid this he included the word 'Voynich' in one of the prompts to GPT and GPT says something flip about how annoying people who think they can translate the Voynich manuscript are. It was funny since it was so "on topic."
       
 (DIR) Post #AtmUQBhVbW9zl0aTS4 by futurebird@sauropods.win
       2025-05-05T10:35:16Z
       
       0 likes, 1 repeats
       
       @cstross The Voynich debunker chuckled about it. But... I was also unsettled. Because it seemed obvious to me that this sudden change in tone was triggered by the keyword "Voynich" and the fact that the vast majority of text the LLM had scanned would be people making just such comments to each other. It was exactly like "cold reading."
       
 (DIR) Post #AtmUWbgBvdai4B22ee by futurebird@sauropods.win
       2025-05-05T10:36:26Z
       
       0 likes, 1 repeats
       
       @cstross My Voynich nerd is a smart man. He even talked about how he was using GPT with care and just for translation and with an understanding that it could be flawed. But as we giggled about how the LLM "gets it about the Voynich frauds" we were all taken in!
       
 (DIR) Post #AtmUZKqIz568iU40Nk by futurebird@sauropods.win
       2025-05-05T10:36:56Z
       
       0 likes, 0 repeats
       
       @cstross The "psychic" notices chalk on my fingers and my conservative dress, bookish speech and guesses "teacher."Then says "They just don't fund schools enough!"I say "Damn right they don't" and really feel like this person understands.
       
 (DIR) Post #AtmUx1VsVCc1iWnvzU by sahqon@beige.party
       2025-05-05T10:41:10Z
       
       0 likes, 1 repeats
       
       @futurebird @cstross Somewhat unrelated but this could be why we who are masking constantly are immune to this kind of thing: they look at the mask and always guess wrong. And so we know they are full of shit. Might not be so obvious if they keep guessing right, which is the case for most people.
       
 (DIR) Post #AtmYuaDBRB5SHS9U6y by peterainbow@beige.party
       2025-05-05T11:25:32Z
       
       0 likes, 0 repeats
       
       @futurebird @cstross I have no doubt that they are using the same social psychologists threat actors to design the front end as they've been using for their social media platforms. It's already clear that these AI platforms are going to be the next gen of social media, see UK tech minister using it to find out social stuff to do and where, people won't be able to help themselves just like now with twitter etc when everyone can see how harmful it's being