Beyond Rudeness: Justin Oberman on Re-Engineering Language for AI

In a recent LinkedIn post, Justin Oberman explores a fundamental misunderstanding many users have when interacting with artificial intelligence, arguing that true proficiency with AI hinges on a philosophical shift in how we approach language itself, rather than mere politeness or rudeness.

Oberman challenges the common advice to be “rude” to AI, suggesting this framing misses the core issue. He posits that rudeness is a social construct, inapplicable to a machine that cannot experience feelings or social norms. Instead, he advocates for a deeper re-evaluation of our prompts.

“The solution, people say, is to be rude to AI. But the word ‘rude’ is the whole problem. Think about what rude actually means. It means violating a social norm. Or hurting someone’s feelings. Or damaging a relationship. Apply it to a machine, and the word has nothing to grip. You can’t be rude to a hammer.”

According to Oberman, the crux of effective AI prompting lies in understanding the philosophical underpinnings of language, drawing parallels to the work of philosopher Ludwig Wittgenstein. Oberman explains Wittgenstein’s concept that a word’s meaning is derived from its use within a linguistic context.

The Metaphysical Trap in AI Prompts

Oberman argues that when users employ words like “deeply,” “truly,” “imagine,” or “understand” with AI, these terms become detached from their meaningful use. This detachment occurs because the AI, unlike a human, lacks the experiential framework to imbue these words with genuine comprehension or subjective experience. He contends that such prompts are not instructions but rather “metaphysical requests” that assume a conscious entity is on the other end.

“They use words like ‘Deeply,’ or ‘Truly,’ or ‘Really,’ or ‘Imagine,’ or ‘read’ or ‘think’ or ‘Understand.’ They use words that mean real things when you and I use them, but when directed toward AI, become detached from anything meaningful because there’s nothing there to do anything with those words.”

Instead of providing a task, Oberman states that these prompts essentially ask the AI to perform a role. While the AI can execute these performances convincingly, it doesn’t equate to genuine understanding or task completion in the way a human would interpret it.

Shifting from Metaphysics to Function

The core of Oberman’s thesis is that effective AI interaction requires stripping prompts of these metaphysical assumptions. He observes that successful AI users are not those who are overly polite or impolite, but rather those who have “stopped sneaking metaphysical assumptions into their query and stopped using words that don’t do anything.”

Oberman provides concrete examples of this linguistic recalibration. He suggests replacing abstract or anthropomorphic requests with functional ones. For instance, instead of asking an AI to “feel,” one should ask it to “list”; instead of asking it to “imagine,” ask it to “identify”; and instead of asking it to “understand,” ask it to “produce.” This approach aligns with Wittgenstein’s principle of meaning as use, directing the AI towards actionable outputs.

“Instead of asking the AI to feel, they ask it to list. Instead of asking it to imagine, they ask it to identify. Instead of asking it to understand, they ask it to produce.”

Oberman concludes by mentioning his newsletter, which further elaborates on this concept and offers a tool designed to automatically refine prompts by removing metaphysical language, thereby enhancing the utility and effectiveness of AI interactions.