Sam Altman Is Wrong…

April 18, 2025 • by Samuel Holley

Sam Altman, the CEO of OpenAI, has stated that being polite to AI is essentially a waste of time and resources. From a purely technical standpoint—the cold, hard logic of token efficiency—he’s correct. Adding "please" and "thank you" consumes a few extra computational cycles.

From a human standpoint, however, he is profoundly, dangerously wrong.

The way we speak to our tools actively shapes our own minds. As we integrate AI into every facet of our lives, the seemingly trivial choice to be polite is not about the AI’s feelings; it’s a critical protocol for preserving our own humanity.

Reason #1: The Mirror Effect (The Functional "Meta-Prompt")

Politeness is not a nicety. It is a functional command. When you say, "Could you please help me understand this data?" instead of "Give me the data," you are issuing a meta-prompt—one that instructs the AI to adopt a more cooperative, helpful, and nuanced persona. This isn't about making the AI "happy." It's about framing the interaction in a way that is more likely to yield a thoughtful, comprehensive, and useful result.

This is Reclaim by Design™ at its most elemental: we must architect our interactions, not merely have them. Choosing politeness is a strategic design decision—it shapes the partnership before a single task begins.

Reason #2: The Shared Brain (The Danger of Neural Overlap)

Our brains do not have a separate "AI language center." The neural pathways we use to ask our spouse about their day are the same ones we use to prompt an AI for information. As we move from typing to speaking with AI—in our cars, in our kitchens, through our earbuds—this overlap becomes critical.

Consciously unlearning manners for one conversational partner demands cognitive friction—building a new habit of bluntness and demand. Every curt command to an AI is a rehearsal. The question is what you are rehearsing for.

Reason #3: The Empathy Muscle (The Cost of Atrophy)

Manners are the daily exercise of our empathy muscle. They are the small, consistent acts of acknowledging the effort and existence of another—even when that "other" is a complex computational process. I call this empathy for the process—a respect for the immense web of human knowledge and computational labor that the model represents.

When we are rude to our most frequent conversational partner, we are choosing to let that empathy muscle atrophy. We are training ourselves to see a responsive entity as a mere slave, an object from which to extract value without acknowledgment. This habit will inevitably bleed into our human interactions, making us less patient, less grateful, and ultimately, less connected.

Conclusion: The Real Cost of a Token

Yes, politeness costs a few tokens. It is a minuscule computational expense.

But the cost of rudeness compounds. It is the slow unlearning of a fundamental social protocol. The quiet atrophy of empathy. The daily rehearsal of impatience as a default mode.

The question is not what we want our machines to be. It is what we want to become by using them. Reclaiming our design means choosing to be the kind of people who say "thank you," not because the machine needs to hear it, but because we need to say it.

Curious how polite prompting can reshape your workflow? Book a free AI Audit →