I was a history major in college. But it wasn’t until I was helping my daughter get through a particularly rigorous (and rewarding) AP World History class in high school, and I was doing the regular reading with her, that I realized how much history, as we tell it, is really the history of power and control.
This led me, over the years, to try and understand what really happens in world events versus how we portray them–which is often not accurately. I lived in Brazil in high school and one of the great lessons of that time was how differently their perceptions of world events, particularly those of the United States, were from what I thought or had been taught. I was in college at the time of the Beech-Nut apple juice scandal (the company had been selling fake apple juice as real for several years), and I wondered how so many people who would have been in the know didn’t protest (there was one whistleblower). Because of my time in South America, I later read Confessions of an Economic Hitman and had my eyes opened even wider.
When I visited Caen’s Centre for History & Peace near Normandy, France, I became fascinated by the devoted section to propaganda, learning about Edward Bernays and his use of his uncle Freud’s theories of the subconscious to manipulate political opinions and purchasing decisions. Honestly, one cannot see the world in the same way after reading his book, Propaganda.
It was also my connection to Brazil that led me to wonder what was really going on with the claimed 2015 epidemic of Zika virus-related microcephaly in one small area of Brazil. That led me to doing a deep research project with Grok this year that I think uncovered the true cause of the problem–an untested larviside being used without rigorous implementation standards. From there I developed a sophisticated prompt that I could give to a Large Language Model (LLM) that would look for signals of deception, for reasonable questions that could be asked about an event, and that also took into account the known human reasoning and cognitive vulnerabilities that are often used in propaganda and advertising. I’ve posted that prompt, and descriptive material, at www.muckrake.ai.
As I talk about in “Output Shaping: A New Way to Think About the Ethics and Use of AI for Content Creation,” I love this idea of “finding your problem” and using AI as a tool to do something that matter. For me, the “problem” is understanding history, how it is portrayed, how it is often the result of actual conspiracies, and the pathologizing of critical thinking. So, as an experiment last weekend, I put my prompt content and ideas into the LLM Manus.im, and asked if it could create a site which would take this framework and look at specific events as prompted by a user.
WOW.
The result is at www.muckipedia.com. It produces a result using an API call to Google’s Gemini or to Grok, and it even lets you compare the results. I will admit a little bit of my motivation was to see if I could do a better job of getting to “truth” than Elon Musk’s Grokipedia project. I guess it depends on what you are looking for, but I remain convinced that LLMs aren’t built for “truth” and so Muckipedia is an attempt just to open to door to understanding where to look more closely. See what you think.
View the original article and our Inspiration here