Post AvDHUSGFy2viUlLicK by hzulla@infosec.exchange
 (DIR) More posts by hzulla@infosec.exchange
 (DIR) Post #AvDHUGjwnpY8lN9GEK by hzulla@infosec.exchange
       2025-06-17T04:54:30Z
       
       0 likes, 2 repeats
       
       27 years ago, a professional political speechwriter and ghostwriter contacted me about a silly Markov chain text generator I had coded.The text generator had been fed with the party platforms of the German federal election of 1998. It was a silly PR stunt by a political news website at the time. It was obviously meant as a joke.https://www.politik-digital.de/projektuebersicht/phrasendrescher/But the speechwriter was absolutely sure that this will be a worthwhile tool to help him write for his clients.I think about this guy a lot today.
       
 (DIR) Post #AvDHUI9rWxsDA2HTGK by hzulla@infosec.exchange
       2025-06-17T05:03:49Z
       
       0 likes, 1 repeats
       
       A Markov chain text generator is the great grandfather of today's generative AI. It's the dumbest form of machine learning you could imagine. But modern LLMs are basically still run on similar ideas at their core:You feed a good amount of text to the machine, it learns about its statistical characteristics without understanding the content. And then the machine is capable to create new text with similar  characteristics. It is why us programmers make fun of LLMs as a "stochastic parrot".The Markov chain "virtual politician" for the 1998 German federal election was based on an article by Brian Hayes written in 1983. The algorithm is so simple, I read about this and then wrote my first Markov chain text generator in high school.
       
 (DIR) Post #AvDHUNRlroAZZ1hLN2 by hzulla@infosec.exchange
       2025-06-17T05:07:16Z
       
       0 likes, 0 repeats
       
       The 1998 "virtual politician" wasn't more than a party trick. It would write nonsense that barely sounded like a human author.(The 1983 article its code was based on used Shakespeare as input and was headlined "A progress report on the fine art of turning literature into drivel" for a reason.)We even added a long explanation of Markov chains on the website. This was high school mathematics!Yet we got responses from the audience by people who considered this an artificial intelligence.https://www.scientificamerican.com/article/computer-recreations-1983-11/
       
 (DIR) Post #AvDHUSGFy2viUlLicK by hzulla@infosec.exchange
       2025-06-17T05:09:26Z
       
       0 likes, 0 repeats
       
       Artificial intelligence is only in the mind of the machine's observer.
       
 (DIR) Post #AvDHUXlzTUHjadEdH6 by hzulla@infosec.exchange
       2025-06-17T06:33:37Z
       
       0 likes, 0 repeats
       
       Afterthought: Reading Brian Hayes and Martin Gardner during high school taught me more about programming and algorithms than any of my professors at University.