Post

Perhaps the most disturbing aspect of working with AI systems is discovering how similar the human computer is to the AI token prediction model. For example, look at behavioral economic studies on anchoring effects. You can prime human responses in a strikingly similar way as one can prime an LLM.