 | I'm currently studying the social implications of AI. Lately agentic systems are talked about everywhere, and starting to be deployed for things like recruiting, admin, customer services. My understanding is that these systems are often brittle and used in tasks poorly suited to generative AI I built House of IFs as an experimental project; it applies Mesopotamian omen logic (IF weird sign > THEN outcome) to AI. Every day, an AI agent scans current news to construct a new omen. It links today's events to similar sign-and-outcome patterns from recent history. The project is both an experiment in "agentic" AI and a critique of how AI makes arbitrary patterns feel convincing. It has a shared memory system, tool-use loops, RAG with embeddings, ... One thing I found was how difficult it is to keep the chatbot accurate, even when it is given precise sources. It really tries to embellish, infer or fill gaps to answer questions. The site is available at: I’d love feedback on one main thing: Does it make you think (differently) about how AI works and is used today? [link] [comments] |
Predictive technology: computational social science