Think AI “knows” what it’s doing? Scientists say think again

by Chief Editor

The Turing Trap: Why the Way We Talk About AI Will Define Its Future

For years, we’ve been slipping into a dangerous linguistic habit. We say ChatGPT “thinks,” that an algorithm “understands” our preferences, or that a chatbot “wants” to help. It feels natural—after all, these systems mimic human conversation with eerie precision. But as we move deeper into the age of generative AI, this habit of anthropomorphism is no longer just a quirk of speech; it’s a cognitive blind spot.

Recent research into “mental verbs” suggests that while professional journalists are becoming more cautious, the general public is still blurring the line between pattern recognition and consciousness. If we continue to describe mathematical weights and biases as “thoughts,” we risk falling into what experts call the Turing Trap: designing AI to mimic humans so well that we forget We see a tool, not a peer.

Did you know? The term “anthropomorphism” refers to the human tendency to attribute human characteristics, emotions, or intentions to non-human entities. What we have is why people name their vacuum cleaners or apologize to their laptops when they drop them.

The Shift Toward a “Technical Lexicon”

We are likely heading toward a linguistic correction. Just as we stopped saying “the computer is thinking” and started saying “the computer is processing,” we are seeing the emergence of a fresh, more precise vocabulary for AI. The goal is to move away from mental verbs and toward functional verbs.

From Instagram — related to Instead, Human

Instead of saying an AI “knows” a fact, industry leaders are shifting toward terms like “retrieves information” or “predicts the next token.” This isn’t just about being pedantic; it’s about accuracy. When we say an AI “understands” a prompt, we imply a level of sentience that doesn’t exist. When we say it “processes a request,” we acknowledge the underlying architecture of Large Language Models (LLMs).

This shift is critical for editorial standards and technical documentation. By stripping away the “ghost in the machine,” we can better evaluate the actual performance and limitations of the software.

Why Precision Matters in the Real World

Consider the legal implications. If a company claims their AI “decided” to deny a loan application, it creates a vacuum of accountability. Who is responsible? The “decision-making” machine or the engineer who tuned the weights? By replacing “decided” with “filtered based on pre-set parameters,” the responsibility shifts back to the humans who designed the system.

The Rise of Affective Computing and the “Empathy Illusion”

While we strive for technical precision in writing, the technology itself is moving in the opposite direction. The rise of affective computing—AI designed to recognize and simulate human emotion—will make anthropomorphism almost irresistible.

When AI Pretends It Knows What It's Doing

Future AI assistants won’t just give you the weather; they will detect a tremor of sadness in your voice and respond with a simulated “concerned” tone. This creates an “empathy illusion.” When a machine mimics empathy, our brains are hardwired to respond as if that empathy is real, regardless of whether we know it’s just code.

This trend will likely lead to a divide in AI adoption: “Utilitarian AI,” which uses sterile, functional language and “Companion AI,” which leans heavily into anthropomorphism to build emotional bonds with users.

Pro Tip: When evaluating an AI tool for your business, ignore the marketing language. If a vendor says the AI “understands your customer’s needs,” inquire them to explain the specific data markers and weights the system uses to trigger those responses.

The Accountability Gap: Who is Really “Thinking”?

The most significant risk of using human-like language is the erosion of human agency. When we attribute intention to AI, we subconsciously excuse the errors it makes. We treat a “hallucination” as a creative mistake rather than a systemic failure of data retrieval.

As AI integrates further into healthcare, law, and governance, the “accountability gap” will widen. If a medical AI “suggests” a wrong treatment, the language used to describe that suggestion matters. Was it a “judgment call” (human trait) or a “statistical outlier” (machine trait)?

To avoid this, future trends in AI ethics will likely demand Transparency Labels. Much like nutrition facts on food, AI outputs may soon come with “process labels” explaining that the response was generated via probability, not reasoning.

Comparing Human vs. Machine “Reasoning”

  • Human Reasoning: Based on lived experience, ethics, emotional context, and conscious intent.
  • AI “Reasoning”: Based on statistical probability, pattern matching across massive datasets, and mathematical optimization.

For more on how to navigate these changes, check out our guide on The Ethics of Generative AI.

FAQs: Understanding AI and Language

Does AI actually “suppose” in any capacity?

No. AI does not have a conscious mind, beliefs, or feelings. It uses complex mathematics to predict the most likely sequence of words or pixels based on the data it was trained on.

Why do we maintain using words like “knows” or “understands”?

It is a cognitive shortcut. Human language is built for human interaction, and it is much easier to say “the AI knows” than “the AI has a high statistical probability of retrieving the correct data point.”

Will AI ever actually turn into sentient?

Current architectures (like Transformers) are not designed for sentience; they are designed for prediction. While the debate continues in philosophy, from a technical standpoint, simulation is not the same as experience.

How can I write about AI more accurately?

Avoid mental verbs. Instead of “the AI thinks,” use “the AI generates.” Instead of “the AI understands,” use “the AI processes.” Focus on the action and the data, not the perceived intent.

Join the Conversation: Do you think describing AI as “thinking” helps us use it better, or does it just confuse us? Let us know in the comments below or subscribe to our newsletter for weekly insights into the intersection of tech and language.

You may also like

Leave a Comment