An interesting experiment with LLMs

I’ve been reading Muskism: A Guide for the Perplexed by Quinn Slobodian and Ben Tarnoff). And so has another reader of this blog who had a terrific idea for an experiment. He (or she) fed the text of the book into eight of the best-known language models — ChatGPT, Claude, Grok, DeepSeek, Gemini, Copilot, Manus and Perplexity. And they were all asked the same question: “How would the 19th century philosopher Karl Marx analyze this 21st century phenomenon?” The individual responses were then fed as sources into Google’s NotebookLM which was asked to provide an audio commentary on them. Here’s an AI-generated transcript of the NotebookLM podcast created from the sources. Think about your morning for a second. You wake up and maybe you check a social media feed that’s entirely curated by this opaque algorithm. Right, something you have zero control over. Exactly. Or you get into a car that is constantly beaming telemetry data back to some server farm somewhere. You pay for your coffee…