When robots can't riddle: What puzzles reveal about the depths of our own minds
undefined 24, 2024
At Amsterdam's Vrije Universiteit, assistant professor Filip Ilievski challenges AI with riddles to improve its common sense reasoning. While AI excels at pattern recognition, it struggles with abstract thinking, as seen when GPT-4 misinterpreted a riddle about a character named Mable. Ilievski's research reveals gaps in AI’s reasoning abilities, especially in tasks like the Wason selection problem, where GPT-4 failed to identify the correct cards to turn over, highlighting its limitations compared to human intuition.
Read full article
