Spot the robot
Now that we live in the world Open AI has made, how can you be sure that I, Catherine-with-a-C, wrote this sentence? The answer is simple: artificial intelligence doesn’t produce language as idiosyncratic as “Catherine-with-a-C.” Large language models — LLMs — work by choosing the most likely next word or next piece of a word, so what they write is predictable, by definition. It can’t not be. So how can you be sure that a human did not write this: “The Bee Sting” is an extraordinary novel that weaves a compelling narrative around the complexities of family dynamics, personal failures, and the journey toward redemption. Paul Murray has crafted a story that is both heart-wrenching and heartwarming, capturing the essence of human resilience in the face of life’s challenges. The characters are intricately developed, each with their flaws and virtues, making them incredibly relatable. . . . These paragraphs were written by ChatGPT. The giveaway is that they do not contain a single word, phrase, sentence, or observation that’s idiosyncratic or unique. The language is hackneyed, the content clichéd. Unfortunately, given the fact that humans deploy plenty of hackneyed language and predictable observations themselves, it’s not easy to tell the two apart. That’s the problem. Anyone who reads for any purpose other than entertainment needs to know the difference between human prose and text generated by an AI. Teachers reading student papers, admissions officers reading applications, customers reading product reviews—even scientists reading peer-reviewed papers and judges reading briefs: a lot of us need to know which is which. It’s self-defense. That’s why we’ve created Spot the Robot. Spot the Robot is a daily “perception puzzle” that, over time, will teach you to distinguish human from AI.
Information-integration learning
To do this, we use information-integration learning. Information-integration learning is the system babies use to learn the difference between cat and dog, pigeons use to learn the difference between Monet and Picasso, and experts in any field use to acquire expert intuition. It’s our perceptual-and-category learning apparatus.3
Perceptual Learning
Perceptual learning requires two elements: 1) repetition and 2) immediate feedback. Perceptual learning does not require conscious rules and analysis, though conscious rules and analysis don’t hurt (see 2 kinds of learning), and you’ll develop both over time. Once you do, you’ll use conscious analysis a lot, especially when intuition doesn’t give you a clear answer. But in the beginning, all you need to do is read a passage, guess whether it’s AI or human, then immediately look at the answer. A new perceptual category will begin to form inside your brain. Try it: Spot the Robot #1
