Why AI Companies Need to Stop Giving Their Features Human Names

At a developer conference in San Francisco this week, Anthropic unveiled a new capability for its AI agent platform called “Dreaming.” The feature allows agents to scan their own activity logs after completing multi-step tasks—like browsing websites or reading files—and identify patterns that could improve future performance. It’s part of a broader push to make AI agents more self-sufficient, but the name is causing some unease.

Anthropic’s blog post explains: “Together, memory and dreaming form a robust memory system for self-improving agents. Memory lets each agent capture what it learns as it works. Dreaming refines that memory between sessions, pulling shared learnings across agents.” The reference to Philip K. Dick’s *Do Androids Dream of Electric Sheep?* is hard to miss, but the problem goes beyond literary allusions.

Since the chatbot boom of 2022, AI vendors have consistently borrowed terms from human cognition. OpenAI launched a “reasoning” model that needed “thinking” time. Startups market chatbots with “memories” that recall personal details like a user’s neighborhood or dietary preferences. These aren’t technical descriptions—they’re branding choices that blur the line between machine processes and human experience.

Research published in *AI & Ethics* warns that anthropomorphizing AI distorts how we judge its capabilities and reliability. When users treat a chatbot as having a personality or inner life, they may over-trust it or project qualities that simply aren’t there. Anthropic’s own constitution even encourages Claude to embody “virtue” and “wisdom,” and the company employs a philosopher to guide its values.

Tech leaders would do well to revisit the cautionary tales they borrow from. In Dick’s novel, the protagonist is crushed to discover that a creature he believed was alive is just a machine. The same lesson applies today: no matter how human the feature names sound, these tools remain fundamentally mechanical. It’s time to call them what they are.

Source: Wired gear