New AI Research Prompts a Reflection on How We Talk to AI

Summary

Felipe Coimbra reflects on the way frustration changes his language when he works with AI. He describes writing comments such as “these all look horrible” when an output misses important details, while recognizing that such wording communicates irritation more clearly than a useful diagnosis. A research review posted on September 29 examines anthropomorphism, or the tendency to treat nonhuman systems as human, and asks how language shapes people’s understanding of AI and the roles they assign to it. Coimbra connects that question to TaskClaw, his project for turning rough ideas into clearer plans and focused tasks for AI coding tools: he argues that specific feedback about crowded layouts, uneven spacing, or unsuitable images is more useful than a general expression of anger. He also discusses an October 1 preprint on artificial consciousness, which argues that behavioral similarities provide only limited evidence that a system has conscious experience. A human-sounding response would therefore not establish that an AI feels hurt or disappointed. The author’s practical conclusion is independent of whether AI has feelings: people can keep high standards while naming the actual problem, removing unnecessary hostility, and pausing when an exchange starts to feel like a fight. He sees this as a way to examine the communication habits being practiced on the human side of the screen.