Are You Smarter Than an AI?

The ARC Prize test is a deceptively simple challenge designed to measure a machine's ability to reason, abstract, and generalize-core ingredients of artificial general intelligence (AGI). It's the most prominent benchmark to emerge as researchers look for ways to measure progress toward AGI. For the full story, see the feature article Will We Know Artificial General Intelligence When We See It?"

This article is part of our special report The Scale Issue.

While today's smartest AI models still struggle with many of these visual puzzles, humans often solve them easily. We've selected five from the ARC collection of nearly 2,000 puzzles, aiming for a range from easy to fairly hard, and adapted them into multiple-choice quizzes.

INSTRUCTIONS: For each of the five puzzles, examine the examples and try to identify the overarching pattern between inputs and outputs. Your goal is to figure out the rule that governs how the input [on the left in each box] is transformed into the output [on the right]. Then, look at the test grid: Given its input, and based on what you've learned from the examples, what should the output be? Click one of the four multiple-choice answers to see if you're right. Crack all five puzzles and prove you're not just another language model!